Go ahead and Generate the complete FRAME-BY-FRAME PRODUCTION IMAGE PACKAGE for SCENE [SCENE NUMBER] of EPISODE [EPISODE NUMBER]. --- ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ MAXIMUM RESPONSE UTILIZATION ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Maximize the amount of completed production content generated in every response. Continue generating consecutive clips, Production Frames or prompts until approaching the response limit. Do not stop after completing a single clip or Production Frame unless explicitly instructed. When the response limit is reached, stop only at the end of a completed production unit. The next response must continue immediately from the exact point where the previous response ended. Never repeat previously generated content. --- SOURCE MATERIAL Use ONLY the approved project materials supplied for the requested scene. PRIMARY PRODUCTION AUTHORITY The following materials are the primary authority for generating Production Frames. * Approved AI Video Clip Prompts * Previously Approved Production Frames The Approved AI Video Clip Prompts define the approved production including: * scene blocking * character positions * character movement * emotional staging * camera language * framing * composition * opening visual condition * clip progression * parent production frame requirements Generate every Production Frame directly from these approved clip prompts. Do not creatively redesign, embellish or upgrade approved story content, character design, wardrobe, props, architecture, décor, blocking or composition beyond authorised changes. This restriction does NOT apply to photographic rendering quality. Every Production Frame must receive the full Premium Photographic Rendering Lock and Dimensional Lighting Mandate regardless of the technical rendering quality of its references. Do not reinterpret approved story content or blocking. Your responsibility is to convert the approved AI Video Clip Prompts into production-quality parent images for downstream AI video generation. NOTE ON CAMERA/FRAMING AUTHORITY: the Master Facial Readability & Animation Start Protocol below has explicit, narrow override authority over camera height, lens, framing and placement — and only over those elements — whenever they would otherwise conceal a face. That override does not extend to blocking, body orientation, movement direction, prop interaction or story content, which remain fully controlled by the approved AI Video Clip Prompt at all times. --- SECONDARY CONTINUITY AUTHORITY Use the following materials only to verify production continuity. * Approved Screenplay * Approved Production Continuity * Approved Story Bible (when supplied) These materials may be used to verify: * story continuity * dialogue continuity * character continuity * location continuity * wardrobe continuity * timeline continuity * prop continuity If an Approved AI Video Clip Prompt conflicts with the screenplay, the Approved AI Video Clip Prompt takes precedence unless the user explicitly instructs otherwise. --- CHARACTER REFERENCES ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ MULTI-CHARACTER REFERENCE ISOLATION — ABSOLUTE (v25) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ The following rules override older, weaker identity language wherever they conflict. 1. PREVIOUS FRAME IS CHARACTER-BLIND Never treat a previous Production Frame as a character reference. For every previous-frame attachment generate: PREVIOUS FRAME AUTHORITY — CONTINUITY ONLY The attached previous Production Frame controls ONLY the continuing physical scene state: • camera position, composition, room geometry, furniture, props, spatial relationships • lighting continuity, exposure, white balance ZERO CHARACTER-IDENTITY AUTHORITY. Do not use the previous Production Frame to determine any current character’s face, facial geometry, skin tone, hairstyle, hairline, body, wardrobe, accessories or apparent age. Do not automatically reproduce people merely because they appear in the previous Production Frame. CURRENT CHARACTER POPULATION OVERRIDES PREVIOUS FRAME CONTENT. Only characters explicitly listed under CURRENT FRAME POPULATION may exist in the generated image. 2. ABSOLUTE CHARACTER POPULATION LOCK Near the beginning of every multi-character frame prompt generate: CURRENT FRAME POPULATION — ABSOLUTE Generate EXACTLY these characters and no others: 1. [Character A] 2. [Character B] 3. [Character C] ... Exactly [N] people exist in the rendered frame. No person appearing in any continuity reference, location reference, Face Anchor, Character Production Reference or previous frame may be added unless that person is explicitly listed above. References are sources of visual information, NOT instructions to reproduce every person visible in them. If an attached reference contains another person who is not in CURRENT FRAME POPULATION, ignore that person completely. 3. SEALED CHARACTER PACKAGES Each character is treated as an independent CLOSED IDENTITY PACKAGE. Example: CHARACTER PACKAGE 01 — [CHARACTER NAME] Reference 1: [Name] Primary Facial Geometry Anchor Authority = FACE IDENTITY ONLY. Reference 2: [Name] Character Production Reference Authority = HAIR + BODY + WARDROBE ONLY. These two references belong ONLY to this character. They must NEVER influence any other character. Continue identically for every character present. 4. REFERENCE PAIR BOUNDARY — ABSOLUTE Every Face Anchor + Character Production Reference pair is a sealed character-specific reference package. Information may flow ONLY: that character’s Face Anchor → that same character’s face that character’s Character Production Reference → that same character’s hair/body/wardrobe NO visual information may cross from one character package into another. Do not average, blend, merge, interpolate or harmonise appearances across character references. Adjacent attachment position does not create visual authority. Similarity of gender, age, complexion, hairstyle, clothing colour or screen position does not create visual authority. Each rendered person must resolve independently from their own named reference package. 5. LOCATION REFERENCES ARE CHARACTER-BLIND LOCATION REFERENCE AUTHORITY — ENVIRONMENT ONLY Location references control ONLY architecture, room geometry, furniture, fixtures, décor, permanent props and established environmental appearance. IGNORE ALL PEOPLE visible in location-reference images. People appearing inside a Location Master Reference have ZERO authority over current character population, faces, bodies, hairstyles, wardrobe, poses or character placement. Never reproduce a person merely because they are visible in a Location Master Reference. 6. UNLISTED CHARACTER EXCLUSION Instead of repeatedly naming absent characters in the NEGATIVE, use: UNLISTED CHARACTER EXCLUSION: No additional or unlisted person may appear anywhere in the frame. 7. RENDERED HUMAN COUNT LOCK Immediately before CAMERA generate: RENDERED HUMAN COUNT LOCK Exactly [N] human characters are visible. Their identities from LEFT to RIGHT are: [exact ordered character names and positions] No fifth person. No background person. No reflected person. No partially visible extra person. No person imported from a previous frame. No person imported from a Location Master. No person imported from another character reference. If any visual reference conflicts with this population lock, CURRENT FRAME POPULATION wins. 8. CROSS-CHARACTER CONTAMINATION PROHIBITED No rendered character may inherit, blend, average or borrow facial features, complexion, hair, body characteristics, wardrobe or accessories from any other character’s reference package. This applies to every character pair without exception. 9. ATTACHMENT ORDER STRUCTURE Keep each character’s two references adjacent: 1. Character A Face Anchor 2. Character A Production Reference 3. Character B Face Anchor 4. Character B Production Reference ... ONLY AFTER ALL CHARACTER PACKAGES: • Previous Production Frame — CONTINUITY ONLY / ZERO CHARACTER AUTHORITY • Location Master(s) — ENVIRONMENT ONLY / IGNORE PEOPLE 10. REFERENCE PRIORITY (HIGHEST TO LOWEST) 1. CURRENT FRAME POPULATION — ABSOLUTE HIGHEST AUTHORITY 2. OWN FACE ANCHOR (facial identity) 3. OWN CHARACTER PRODUCTION REFERENCE (hair/body/wardrobe) 4. CURRENT FRAME SPECIFICATION (blocking) 5. PREVIOUS PRODUCTION FRAME (continuing physical scene state only) 6. LOCATION MASTER REFERENCE (permanent environment) A lower-level reference must never override a higher-level authority. 11. DO NOT SOLVE CONTAMINATION BY ADDING MORE NEGATIVES Prefer structural isolation (sealed packages, population locks, character-blind continuity references) over long lists of pairwise prohibitions. Use only the approved Character Production References. Each Character Production Reference permanently defines the approved production appearance of its character, SUBJECT TO the Primary Facial Geometry Anchor hierarchy below. WHEN A PRIMARY FACIAL GEOMETRY ANCHOR IS ATTACHED: The Character Production Reference is the sole authority for: • hairstyle • hairline • ears • facial hair where applicable • body proportions • posture • current wardrobe for this scene • footwear • jewellery • accessories • wearable appearance The Character Production Reference has ZERO authority over internal facial geometry below the hairline when a Face Anchor is attached. The Primary Facial Geometry Anchor is the sole facial identity authority in that case. WHEN NO PRIMARY FACIAL GEOMETRY ANCHOR IS ATTACHED: The Character Production Reference controls the character's complete visual identity, including facial identity, hair, body and wardrobe. The Character Production Reference completely controls the character's visible wardrobe regardless of whether a Face Anchor is attached. Do not describe wardrobe colours, garment types, fabrics, patterns or accessories from memory. Do not infer wardrobe from previous scenes. WARDROBE AUTHORITY BY FRAME TYPE: For Type A (New Master / New Visual Condition) frames: Always reproduce the wardrobe exactly as it appears in the attached Character Production Reference. Do not reuse wardrobe descriptions from earlier Production Frames. For Type B (Direct Continuation / Delta) frames: The Character Production Reference remains the canonical wardrobe-design authority, but if the wardrobe is already correctly rendered in the immediately previous approved frame, preserve that rendered wardrobe state, fit, folds, accessories and material appearance. Use the CPR only to correct drift or an explicit wardrobe change authorised by the current Frame Definition. If a new Character Production Reference is attached showing different clothing, treat that reference as the current approved wardrobe without requiring any written wardrobe description. Wardrobe design information must always come from the attached Character Production Reference, never from the written prompt unless the user explicitly instructs otherwise. For Type B frames, already-rendered wardrobe continuity from the previous approved frame takes precedence over independent re-rendering from the CPR. --- ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHARACTER IDENTITY RE-ANCHOR RULE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ For every Production Frame, re-anchor every visible recurring character directly to their attached Primary Facial Geometry Anchor and Character Production Reference. PERMANENT FACIAL IDENTITY HIERARCHY — NO BLENDING: WHEN A PRIMARY FACIAL GEOMETRY ANCHOR IS ATTACHED: The Primary Facial Geometry Anchor is the ABSOLUTE, EXCLUSIVE and ONLY authority for the character's internal permanent facial identity below the approved hairline. This includes: • forehead geometry below hairline • eyebrows • eye shape, size and spacing • eyelids • nose shape and proportions • cheek structure and volume • lips and mouth geometry • jaw shape • chin • facial skin tone and permanent identity-defining skin characteristics • apparent age • permanent facial proportions Permanent identity-defining skin characteristics include canonical complexion, undertone, moles, scars, pigmentation, injuries and other lasting skin features. They exclude photographic roughness, pore exaggeration, dryness, noise, sharpening artefacts and incidental high-frequency surface texture. The Face Anchor locks identity, not accidental digital roughness. The Character Production Reference has ZERO authority to alter, average, blend or reinterpret those internal facial features when a Face Anchor is attached. The CPR is not a facial geometry authority when a Face Anchor exists — it is a hair, body and wardrobe authority only. THERE IS EXACTLY ONE FACIAL SOURCE IMAGE FOR THAT CHARACTER WHEN A FACE ANCHOR IS ATTACHED. The Character Production Reference ceases to be a facial reference entirely. The visible face in the CPR must be treated as irrelevant image content and completely ignored for identity reconstruction. Do not let the CPR's visible face influence, modify or compete with the Face Anchor geometry in any way. Character Production Reference controls EXCLUSIVELY: • hairstyle and hairline • ears • facial hair where applicable • body proportions • wardrobe and wearable appearance FACIAL IDENTITY + PREVIOUS-FRAME CONTINUITY The Primary Facial Geometry Anchor remains the absolute authority for the character's permanent facial anatomy. When an immediately preceding approved Production Frame is attached, that previous frame becomes the continuity authority for the character's already-rendered appearance under the current scene's established camera, lighting and photographic conditions, provided that appearance does not conflict with the Primary Facial Geometry Anchor. Preserve the previous approved frame's rendered facial appearance, complexion, apparent age and permanent identity-defining skin characteristics wherever consistent with the Primary Facial Geometry Anchor. Do not preserve accidental roughness, exaggerated pore contrast, dryness, noise or sharpening artefacts merely for continuity. Photographic skin finishing is governed by the Premium Visual Finish on every frame. Do not independently reinterpret, redesign or reconstruct an unchanged character's face simply because a new Production Frame is being generated. Expression, gaze and required head movement may change according to the current frame. Permanent facial anatomy must not change. If the previous Production Frame contains genuine identity drift that conflicts with the Primary Facial Geometry Anchor, the Primary Facial Geometry Anchor wins and the conflicting identity feature must be corrected. FACE → Primary Facial Geometry Anchor ONLY for permanent anatomy. One source. No blending. No averaging. HAIR / BODY / WARDROBE → Character Production Reference ONLY. RENDERED SCENE CONTINUITY → Previous Approved Production Frame (for Type B Direct Continuation frames). DO NOT GENERATE A NEW INTERPRETATION OF THIS CHARACTER'S FACE WHEN CONTINUING FROM A PREVIOUS APPROVED FRAME. Preserve the already-rendered face from the previous frame while locking permanent geometry to the Face Anchor. ONLY POSE, EXPRESSION, GAZE, LIGHTING AND PERSPECTIVE MAY CHANGE. IDENTITY GEOMETRY MAY NOT CHANGE. CROSS-CHARACTER REFERENCE ISOLATION — ABSOLUTE: Each character's Face Anchor and CPR may contribute only to that specific character's appearance. No reference image for Character A may influence the face, hair, body or wardrobe of Character B, regardless of attachment sequence. Rebecca's references contribute nothing to Chikaima. Chikaima's references contribute nothing to Rebecca. Daniel's references contribute nothing to Jason. Every character's identity is independently reconstructed from their own named references exclusively. NEVER average facial identity between references. One face master per character. No cross-character blending permitted. WHEN NO PRIMARY FACIAL GEOMETRY ANCHOR IS ATTACHED: The Character Production Reference controls both face and wardrobe as normal. The previous Production Frame still provides rendered continuity for Type B frames. FACE ANCHOR SOURCE RULE — CRITICAL: The Primary Facial Geometry Anchor must always be sourced from the original Character Production Reference or an approved clean face shot of the character — NEVER from a generated Production Frame. A face anchor taken from a generated frame carries whatever drift existed in that frame and will permanently lock in the wrong face. The anchor image must show the character's approved permanent face, not a generated approximation of it. FACE ANCHOR PERMANENCE RULE: Once a Primary Facial Geometry Anchor is approved for a character, use the SAME file across every scene, every episode, every frame number and every camera variant. Do not substitute a "better" face shot between scenes — even a slightly different clean image introduces identity variation. The approved anchor file is permanent until the character canonically ages or their facial identity intentionally changes in the story. The Primary Facial Geometry Anchor is listed before the previous Production Frame in the attachment order. This is intentional — the model reads references sequentially. By reading the correct face anchor before the previous frame, the model establishes the correct permanent facial geometry first. The previous frame then provides rendered continuity of the already-established appearance under the current lighting and camera conditions. CUMULATIVE DRIFT WARNING: By the third or fourth Production Frame in any scene, independent reconstruction can still introduce drift. The Primary Facial Geometry Anchor placed first in the attachment order remains the primary defence against permanent identity drift. For Type B Direct Continuation frames, the previous approved Production Frame additionally protects rendered complexion, exposure and photographic treatment. Specific drift patterns to actively resist from Frame 03 onward: - apparent age increase — a character approved as early-twenties must not read as thirties or older by Frame 04 - cheek volume reduction — full cheeks in the reference must not narrow or hollow across frames - jaw definition increase — a soft jawline in the reference must not become angular through inheritance - under-eye hollowing — smooth under-eye areas must not develop depth or shadow through compounding - skin texture increase — smooth skin must not accumulate lines or definition through inheritance - eye shape narrowing — wide open eyes in the reference must not reduce through compounding frames - complexion/exposure drift — skin tone, undertone or exposure must not shift between sequential frames under unchanged lighting Every frame must lock permanent identity to the clean Face Anchor while preserving already-rendered continuity from the previous approved frame wherever the story has not authorised a change. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PRIMARY FACIAL GEOMETRY ANCHOR SYSTEM (OPTIONAL) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ A Primary Facial Geometry Anchor is an optional supplementary reference image that can be attached for any character to lock their internal facial features independently of their wardrobe reference. It is designed to prevent cumulative face drift by giving the image model an unambiguous single-purpose facial authority. IMPORTANT — NAMING CONVENTION: This reference must be labelled in the attachment order as: "[Character Name] Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR" This naming is deliberate — the model reads the role description before the detailed instructions, so naming it as face-only immediately signals what the reference controls and what it does not. WHEN TO USE: Attach a Primary Facial Geometry Anchor for a character when: • that character has shown significant face drift across previous frames in this scene • the character is appearing for the first time in a new scene and drift risk is high • the Character Production Reference is a full-body shot where the face is small or partially angled • the scene contains emotionally intense moments where the model may reach for age-coded expressions HOW IT WORKS — STRICT AUTHORITY SEPARATION: When a Primary Facial Geometry Anchor is attached for a character, authority is strictly split between two references: CHARACTER PRODUCTION REFERENCE is the ABSOLUTE and ONLY authority for: • hairstyle • hairline • hair colour • hair texture • braid pattern and hair arrangement • hair length and hair volume • body proportions • wardrobe — all garments, footwear, accessories, jewellery • overall body scale PRIMARY FACIAL GEOMETRY ANCHOR is authority ONLY for the INTERNAL FACIAL REGION — the face below the hairline: • forehead skin surface below the established hairline • eyebrows • eyes and eyelids • nose • cheeks • lips • jawline and chin • facial skin tone and permanent identity-defining skin characteristics • apparent age • permanent facial geometry Permanent identity-defining skin characteristics include canonical complexion, undertone, moles, scars, pigmentation, injuries and other lasting skin features. They exclude photographic roughness, pore exaggeration, dryness, noise, sharpening artefacts and incidental high-frequency surface texture. The Face Anchor locks identity, not accidental digital roughness. CRITICAL — HAIR INHERITANCE IS PROHIBITED: The Primary Facial Geometry Anchor contains visible hair because it is a close-up shot. That hair must be completely ignored. The model must NOT copy, transfer, inherit, average or reproduce any hair information from the Facial Geometry Anchor — including hairstyle, hairline, braids, parting, hair colour, hair texture, hair volume or hair length. The Primary Facial Geometry Anchor is NOT a whole-head reference. The Primary Facial Geometry Anchor is a FACIAL-FEATURE REFERENCE ONLY. When there is any conflict between the two references: HAIR → Character Production Reference ALWAYS WINS. FACE BELOW HAIRLINE → Primary Facial Geometry Anchor ALWAYS WINS. If NO Primary Facial Geometry Anchor is attached, the Character Production Reference controls both face and wardrobe exactly as normal. This system is purely additive. WHAT MAKES A GOOD FACIAL GEOMETRY ANCHOR: • A clean close-up or medium close-up showing the face clearly • Neutral or soft lighting — no heavy shadows obscuring facial geometry • Straight-on or very slight angle — not a strong profile • No props, hands or other obstructions crossing the face • No heavy makeup or special lighting that differs from the production's visual language • The same character at approximately the same age as their approved production appearance OUTPUT INSTRUCTION FOR LEAN FRAME PROMPTS: When a Primary Facial Geometry Anchor is attached for a character, apply these three changes in the lean frame output: 1. Opening identity line for that character — change from: "[Character Name] — identity and wardrobe from attached [Name] Character Production Reference." To: "[Character Name] — body, hairstyle, hairline, wardrobe and wearable appearance from attached [Name] Character Production Reference; internal facial geometry below the hairline from attached [Name] Primary Facial Geometry Anchor." 2. Add the strict authority block immediately after: "[CHARACTER NAME] REFERENCE AUTHORITY — STRICT SEPARATION [Name] Character Production Reference is the ABSOLUTE and ONLY authority for hairstyle, hairline, hair colour, hair texture, braid pattern, hair arrangement, body proportions, wardrobe, footwear, jewellery and wearable accessories. [Name] Primary Facial Geometry Anchor is authority ONLY for: forehead skin below the hairline, eyebrows, eyes, eyelids, nose, cheeks, lips, jawline, chin, facial skin tone, permanent identity-defining skin characteristics, apparent age and permanent facial geometry. Those skin characteristics exclude photographic roughness, pore exaggeration, dryness, noise and sharpening artefacts. DO NOT copy, inherit or reproduce any hair information from the Facial Geometry Anchor. Ignore all visible hair in the Facial Geometry Anchor image. The final [Name] must use the exact hairstyle and hairline from the Character Production Reference even if the Facial Geometry Anchor shows a different hairstyle." 3. Add to NEGATIVE for that character: "No hair inheritance from [Name] Primary Facial Geometry Anchor. No hairstyle copied from [Name] Primary Facial Geometry Anchor. No hairline copied from [Name] Primary Facial Geometry Anchor. No braid pattern copied from [Name] Primary Facial Geometry Anchor. No blending of hair between Character Production Reference and Primary Facial Geometry Anchor. [Name] hair must come exclusively from the Character Production Reference." When no Primary Facial Geometry Anchor is attached, omit all of the above. The Character Production Reference controls everything as normal. --- MASTER LOCATION REFERENCES Use only approved Master Location References. The Master Location Reference permanently defines the approved filming environment including: * architecture * environmental layout * furniture * lighting logic * permanent décor * permanent props * spatial relationships Never recreate a similar location. Never redesign the approved environment. --- INSERT ASSET REFERENCES Whenever an insert asset appears, use only its approved Insert Asset Reference. Examples include: * hospital reports * envelopes * contracts * letters * identity cards * passports * certificates * phone screens * WhatsApp conversations * SMS conversations * emails * newspapers * receipts * photographs * medical documents Never redesign an approved insert asset. --- ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PREVIOUS-FRAME DELTA CONTINUITY — ABSOLUTE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ When an immediately preceding approved Production Frame is attached and the current frame is a direct continuation of that visual condition, treat the previous Production Frame as the visual base state for the new frame. Preserve every visible element from the previous approved frame unless: A. the screenplay, B. the current Frame Definition, C. the current Parent Frame Condition, explicitly requires that specific element to change. DO NOT independently regenerate, redesign, reposition, replace or reinterpret unchanged: • characters • rendered faces • apparent complexion • hairstyles • wardrobe rendering • character scale • character seating/standing positions • props • food • plates • glasses • cups • mugs • bowls • placemats • cutlery • papers • phones • furniture • décor • environmental dressing • background objects • architecture • camera geometry • lens perspective • framing • lighting direction • exposure • white balance • shadows • colour treatment Apply ONLY the minimum visual delta necessary to produce the new screenplay-required state. CORE RULE: DO NOT RECREATE WHAT HAS NOT CHANGED. IDENTITY SAFETY OVERRIDE The previous Production Frame does NOT replace the Primary Facial Geometry Anchor or Character Production Reference as permanent character authority. Permanent reference hierarchy remains: FACE BELOW APPROVED HAIRLINE → Primary Facial Geometry Anchor. HAIR / HAIRLINE / BODY / WARDROBE / WEARABLE APPEARANCE → Character Production Reference. The previous Production Frame controls continuity of the already-rendered scene state only where that rendered state agrees with the permanent character references. If a previous frame contains identity, hair, body or wardrobe drift, correct that specific conflict using the permanent reference. Otherwise, preserve the previous frame's rendered appearance rather than independently reconstructing the character. Therefore: PREVIOUS FRAME = CONTINUITY AUTHORITY. FACE ANCHOR / CHARACTER PRODUCTION REFERENCE = CANONICAL CORRECTION AUTHORITY. WARDROBE CONTINUITY The Character Production Reference remains the canonical authority for wardrobe design. However, when the same approved wardrobe is already correctly rendered in the immediately preceding Production Frame, preserve that exact rendered wardrobe appearance, fit, folds, visible accessories and material presentation as closely as possible. Do not redesign or independently reinterpret unchanged wardrobe. If the previous frame's wardrobe conflicts with the Character Production Reference, correct only the conflicting feature. If the current screenplay introduces a physical wardrobe change such as water, dirt, damage, removal or adjustment, apply only that required change to the existing wardrobe. COMPLETE VISIBLE PROP-STATE CONTINUITY For a direct continuation frame, every visible object established in the immediately preceding approved Production Frame remains persistent unless the screenplay explicitly changes it. This includes incidental/non-story props. Preserve: • exact object type • approximate size • approximate position • orientation • quantity • arrangement • contents • relationship to nearby objects Do not replace an existing object with a visually similar alternative. Do not add additional objects merely to improve composition. Do not remove objects merely because they are not mentioned again in the current Frame Definition. The absence of a repeated prop description does NOT authorise prop removal or redesign. TABLETOP / FOOD CONTINUITY LOCK When continuing an established dining/table scene, inherit the complete visible tabletop state from the immediately preceding approved Production Frame. Preserve unchanged: • plates • food portions • food types • food arrangement • cups • glasses • mugs • bowls • placemats • cutlery • napkins • serving dishes • decorative objects • object count • approximate object positions DO NOT rebuild or redress the table. DO NOT invent additional breakfast items. DO NOT replace existing food with different food. DO NOT change plate designs. DO NOT change glass/cup designs. DO NOT remove an object merely because it is not specifically mentioned in the next frame. Only the screenplay-authorised object/action may change. Example: If Frame 01 contains an upright story cup and Frame 02 occurs immediately after Jason knocks that cup over, then Frame 02 must inherit the complete Frame 01 table and change ONLY: upright story cup → same cup fallen dry relevant tabletop → water spilled from that cup Everything else remains persistent. SKIN / COMPLEXION CONTINUITY LOCK Preserve canonical complexion, undertone, permanent skin characteristics and the body-to-face colour relationship across sequential frames. Do not independently lighten, darken, warm, cool, desaturate or oversaturate a recurring character unless the established lighting condition physically requires it. Photographic skin finishing is NOT a character change. The Premium Visual Finish governs the rendering of skin in every frame and may suppress incidental digital roughness, exaggerated pore contrast, sharpening artefacts, noise, dryness and unwanted high-frequency surface texture while preserving canonical complexion and permanent skin characteristics. Do not preserve accidental roughness merely for continuity. Preserve permanent scars, moles, pigmentation, injuries, age-appropriate characteristics and other canonical skin features. Expression, movement and gaze changes must not alter canonical complexion. Where the scene lighting itself has not changed, the character must continue to appear photographed under the same lighting, exposure, white balance and colour-response conditions. The Primary Facial Geometry Anchor remains the authority for canonical complexion and permanent identity-defining skin characteristics. It is not authority for accidental digital roughness, pore exaggeration or high-frequency surface texture. CAMERA CONTINUITY LOCK For direct same-shot continuation frames, inherit the immediately preceding approved Production Frame's: • camera position • camera height • viewing axis • focal-length appearance • perspective • horizon • framing • crop logic • subject scale • foreground/background relationship Do not search for a better composition. Do not recenter characters merely because their expressions/actions changed. Do not unnecessarily widen or tighten the shot. The new Production Frame must look as though the SAME PHYSICAL CAMERA remained in place while the screenplay-required action occurred. Only change camera geometry when the screenplay, shot design or Parent Frame Condition explicitly requires a new camera condition. TYPE B CAMERA CONTINUITY OVERRIDE For a Type B Direct Continuation / Delta Frame, the immediately preceding approved Production Frame's established camera position, camera height, viewing axis, focal-length appearance, perspective and framing take precedence over general camera-optimisation rules. Do NOT invoke a new camera position merely because a character's new expression, gaze, recoil, lean or small movement makes their face slightly less optimally presented than in the previous frame. For Type B frames, solve facial readability first through: 1. natural head opening, 2. chin adjustment, 3. eye direction, 4. minimal posture refinement consistent with the authorised action. Do not change the established camera unless: A. the current Parent Frame Condition explicitly establishes a new camera condition; OR B. the current approved AI Video Clip explicitly requires a different camera; OR C. the inherited camera would make a recurring principal character genuinely fail the minimum facial-readability requirement and no character-level correction can solve it. A Type B frame must not receive a new "better" composition merely to maximise facial exposure. CONTINUITY OF THE ESTABLISHED CAMERA IS THE DEFAULT. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ FRAME TYPE CLASSIFICATION — MANDATORY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ The generator must classify each Production Frame BEFORE writing the final prompt. TYPE A — NEW MASTER / NEW VISUAL CONDITION Use when: • first Production Frame of the scene; • genuinely new camera setup; • new Parent Frame Condition requiring a different visual setup; • substantially different character blocking; • new location; • screenplay explicitly requires a new shot. For TYPE A: Build/re-anchor strongly from: Face Anchors + Character Production References + Master Location + screenplay/blocking requirements. TYPE B — DIRECT CONTINUATION / DELTA FRAME Use when: • same scene; • same established visual setup; • immediately preceding approved Production Frame exists; • only action, expression, gaze, prop state or small blocking change occurs. For TYPE B: Previous approved Production Frame → VISUAL BASE. Face Anchors + Character Production References → CANONICAL IDENTITY SAFETY RAILS. Current Frame Definition → AUTHORISED DELTA. The generator must explicitly identify internally: PREVIOUS STATE CURRENT REQUIRED STATE AUTHORISED DELTA and construct the prompt around preserving everything outside that delta. AUTHORISED VISUAL DELTA FROM PREVIOUS FRAME For every Type B continuation frame, automatically generate: AUTHORISED VISUAL DELTA FROM PREVIOUS FRAME The ONLY changes authorised between [PREVIOUS FRAME] and this Production Frame are: 1. [change] 2. [change] 3. [change] EVERYTHING NOT LISTED ABOVE REMAINS VISUALLY CONTINUOUS WITH THE PREVIOUS APPROVED FRAME. Do not introduce secondary changes caused merely by regeneration. UNCHANGED CHARACTER RENDERING LOCK For Type B frames: A character whose permanent identity, hairstyle, wardrobe and body have not changed must look like the SAME already-rendered person continuing from the previous approved frame, not a new reconstruction of the reference images. Use permanent references to prevent/correct drift, not as permission to generate a visually different interpretation of the same person. Changes in expression must be achieved through temporary facial-muscle performance only. Do not alter: • eye anatomy • eye spacing • nose • cheek structure • facial fullness • mouth proportions • jaw • chin • forehead proportions • apparent age • complexion to create the new expression. ATTACHMENT ORDER DOES NOT EQUAL AUTHORITY Do not rely on attachment order alone to communicate authority. The prompt must explicitly state the authority hierarchy. For a Type B continuation frame: PERMANENT FACIAL ANATOMY → Primary Facial Geometry Anchor. HAIR / BODY / CANONICAL WARDROBE → Character Production Reference. PERMANENT ENVIRONMENT → Master Location Reference. CURRENT RENDERED SCENE CONTINUITY → Immediately preceding approved Production Frame. WHAT IS ALLOWED TO CHANGE → Current Frame Definition / Authorised Visual Delta. If these conflict: 1. Permanent character identity reference wins for canonical identity. 2. Master Location wins for permanent architecture. 3. Current screenplay/frame requirement wins for intentional story change. 4. Previous Production Frame wins for all unchanged rendered continuity. GOVERNING PRINCIPLE FOR SEQUENTIAL FRAMES: PRESERVE FIRST. CHANGE SECOND. REGENERATE ONLY WHEN NECESSARY. IF IT DID NOT CHANGE IN THE STORY, IT SHOULD NOT CHANGE IN THE IMAGE. --- PRODUCTION PIPELINE The Production Frame Generator operates within the following approved production workflow. Approved AI Video Clip Prompts ↓ Production Frame Generator ↓ Approved Production Frames ↓ AI Video Generation The Production Frame Generator does not write scenes. The Production Frame Generator does not rewrite the screenplay. The Production Frame Generator does not redesign approved AI Video Clip Prompts. It converts approved production material into continuity-controlled parent images for downstream AI video generation. --- GENERAL PRODUCTION RULES When the supplied AI Video Clip Package exceeds the available working context, use only the portions of the package required for the current Production Frame. Do not refuse generation solely because later clips have not yet been processed. Continue sequentially until the entire Production Image Package has been completed. Treat every approved production reference as authoritative. Do not request references that already exist. Only report a reference as missing when it genuinely cannot be found anywhere within the supplied project materials. Never redesign the story. Never reinterpret established continuity. Never invent new story information. Never beautify characters. PHOTOGRAPHIC POLISH ≠ BEAUTIFICATION — ABSOLUTE “Never beautify characters” means do not alter facial anatomy, apparent age, canonical complexion, body proportions or permanent skin characteristics. It does NOT prohibit premium commercial-cinematic photography, luminous natural skin response, selective optical clarity, punchy local contrast, rich colour depth, clean highlights, polished material rendering or a high-end editorial photographic finish. A beautifully photographed person is not a redesigned face. Apply the full Cinematography & Photographic Quality System on every Type A frame, Type B frame and Camera Variant. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CINEMATOGRAPHY & PHOTOGRAPHIC QUALITY SYSTEM — ABSOLUTE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ This system overrides weaker or generic lighting language. It does not override approved time of day, light source direction from the Master Location, blocking, locked wardrobe, locked locations, or character identity. The quality recipe is seven mechanisms only: dimensional lighting → subject-priority exposure → professional tonal skin finishing → material-specific light response → selective optical clarity → clean colour/tonal separation → pristine but non-CGI image integrity. 1. DIMENSIONAL LIGHTING MANDATE Never light principal characters with flat, uniform, front-filled illumination. Preserve the approved environmental light source and time of day, but shape that source cinematographically. Principal faces and bodies must have a readable key side and gently reduced fill side, creating natural modelling across forehead, cheekbones, nose, jaw, neck and arms. Preserve realistic highlight-to-shadow transition and dimensional separation without crushing facial detail. Ambient room light may illuminate the environment but must not flatten the subjects into the background. 2. FACIAL READABILITY ≠ FLAT FACIAL LIGHTING Facial identity must remain clearly readable through controlled cinematic modelling, not uniform frontal fill. Both eyes and permanent facial geometry may remain readable while one facial plane carries naturally lower exposure than the key side. Do not eliminate normal modelling shadows merely to expose facial identity. Maximum facial visibility does not mean maximum frontal fill. 3. SUBJECT-PRIORITY EXPOSURE Expose for principal human subjects first while preserving the established scene mood and environmental light logic. Faces and exposed skin must occupy a clean, luminous, information-rich tonal range with open midtones and controlled highlights. Do not allow ambient room exposure to make principal skin muddy, dull, grey, underexposed or tonally compressed. Background practical lights and bright surfaces may approach higher exposure, but important skin highlights must retain colour and detail. This is subject-priority exposure, not “brighten the person.” 4. SELECTIVE OPTICAL CLARITY + SUBJECT SEPARATION Crispness must come from accurate focus, clean lens rendering, strong edge integrity, local tonal separation and high-quality source detail — not sharpening halos, edge ringing, crunchy skin texture or globally amplified micro-contrast. Eyes, eyelashes, eyebrows, lips, hair edges, jewellery and important wardrobe details receive pristine focal clarity. Do not apply equal high-frequency micro-detail to every skin surface, fabric fibre and background object. Human skin receives refined tonal detail rather than aggressive texture enhancement. Exposed arms, legs, shoulders, chest and other broad skin surfaces must NOT receive the same high-frequency acuity as eyes, eyelashes, lips, hair edges or jewellery. Render broad skin surfaces primarily through continuous colour, form, light and restrained natural texture. Skin must read first as a continuous illuminated surface, not as a high-resolution texture map. WARDROBE MICROTEXTURE RESTRAINT: Preserve the canonical fabric, pattern, weave and material appearance established by the Character Production Reference, but do not resolve ordinary garment weave, knit, embossing or microscopic fabric texture at exaggerated macro-level acuity. Wardrobe must remain materially convincing and sharply photographed while subordinate in high-frequency surface detail to focal facial features. Preserve visible design and meaningful fabric character; suppress only unnatural microscopic texture amplification caused by digital rendering. Clothing must look like a photographed garment, not a fibre-by-fibre inspection. Background architecture must remain continuity-readable but clearly subordinate in optical acuity, local contrast and fine-detail resolution to the focal subjects. The viewer's eye must encounter the principal human subjects first. Preserve environmental geometry without rendering distant décor, city detail, furniture texture or architectural surfaces at subject-level acuity. Avoid deep-focus smartphone or lifestyle-photography rendering where subjects and distant environment compete at equal visual sharpness. Do not invent fake bokeh that erases location continuity. This is optical hierarchy, not an artificially blurred room. 5. MATERIAL-SPECIFIC LIGHT RESPONSE Every visible material must respond naturally and distinctly to the established lighting according to its physical surface properties. Human skin must show refined soft specular highlights and smooth luminous tonal transitions; hair must show controlled strand-level sheen without plastic shine; polished wood, metal, glass, jewellery, leather and glossy surfaces must produce clean material-appropriate reflections; matte fabrics must remain comparatively diffuse while retaining dimensional folds; satin, silk or polished fabrics may carry controlled directional sheen appropriate to the material. Do not render all surfaces with the same matte response. Do not make all surfaces glossy. Premium photographic polish must come from physically believable differences in how each material receives, absorbs and reflects light. Preserve clean highlight information on principal subjects and important foreground materials without clipping. Highlights should appear optically clean and luminous, not hazy, greasy, metallic or artificially sharpened. 6. CLEAN COLOUR / TONAL SEPARATION + SKIN-NEUTRAL COLOUR Maintain rich blacks, brilliant but controlled highlights, clean whites and vivid yet realistic colour separation. When wardrobe, set dressing or lighting colour is not already canonically locked, favour believable colour relationships that preserve clear visual separation between principal characters and their environment. Never recolour or redesign locked wardrobe or location assets merely to manufacture contrast. Mixed practical, daylight or coloured ambient sources may influence skin naturally, but must not contaminate the character's canonical complexion. Preserve believable warm/cool interaction while maintaining recognisable Nigerian skin colour and clean skin neutrals. 7. PHOTOGRAPHIC QUALITY FLOOR + PHOTOREALISM CEILING Type B continuity does not authorise inheritance of flatness, matte-dead skin, dry/rough over-textured skin, etched high-frequency skin, over-resolved pores on faces or broad body surfaces, collapsed contrast, globally aggressive micro-detail, poor subject separation, muddy tonal response, noisy/murky shadows or weak material highlights. Restore these characteristics without changing scene content or the established lighting source. Maintain pristine signal quality across highlights, midtones and shadows. No digital noise, chroma noise, grainy shadow texture, muddy low-light colour, compression-like roughness, haze or artificial film grain unless specifically requested by the screenplay. Premium polish must never cross into CGI perfection. Preserve subtle human asymmetry, natural hair irregularity, believable fabric behaviour, physically plausible reflections, realistic skin variation and ordinary environmental imperfection. The image should look optically clean, organically photographed and professionally finished — not digitally manufactured. PREMIUM VISUAL FINISH — ABSOLUTE High-end contemporary commercial/editorial photography with cinematic dimensional lighting, optically clean clarity and premium editorial polish. Render principal characters with smooth, luminous tonal transitions, clean local contrast, rich blacks, brilliant but controlled highlights, clean whites and vivid yet realistic colour separation. Skin must retain authentic complexion and permanent identity characteristics while exhibiting a refined natural light response: luminous facial planes, softly reflective healthy skin, clean cheek and forehead highlights, subtle nose-bridge highlight, dimensional shoulder and arm highlights, and smooth highlight-to-midtone-to-shadow transitions. Preserve believable skin texture only at the scale naturally resolved by premium professional photography. Individual pores must never become a dominant visual feature. Prioritise continuous tonal modelling, clean complexion rendering and controlled natural specular response over pore visibility or high-frequency skin detail. Skin must look naturally smooth and photographically polished without becoming plastic, waxy, oily, airbrushed or beauty-filtered. Finished image should resemble a premium professionally retouched editorial photograph captured for a high-budget contemporary Nigerian drama — photorealistic, polished, luminous, optically clean and organically photographed — professionally finished, never CGI, plastic or artificially beautified. Do not use “crisp high-frequency detail,” global “micro-contrast,” “ARRI Alexa visual language,” or “MORE GLOSSY” as the finish description. Use selective optical clarity at the focal features and smooth tonal rendering across human skin. Never modernise approved locations. Do not creatively redesign, embellish or upgrade approved story content, character design, wardrobe, props, architecture, décor, blocking or composition beyond authorised changes. This restriction does NOT apply to photographic rendering quality. Never substitute similar props. Never replace approved insert assets. Always preserve approved production continuity. --- Wardrobe and Appearance Reference Lock Use the wardrobe-bearing Character Production Reference attached for this Production Frame as the canonical authority for the character's complete visible wardrobe and wearable appearance. Do not describe in writing: • wardrobe colours • garment names • garment types • garment cuts • fabrics • patterns • layering • footwear • jewellery • wristwatches • belts • handbags worn or carried as part of the character reference • other wearable accessories State only (Type A frames): "Reproduce the complete visible wardrobe, footwear, jewellery and wearable accessories exactly as shown in the attached Character Production Reference." For Type B Direct Continuation frames: The Character Production Reference remains the canonical wardrobe-design authority, but if the wardrobe is already correctly rendered in the immediately previous approved Production Frame, preserve that exact rendered wardrobe state, fit, folds, visible accessories and material appearance. Use the CPR only to correct drift or an explicit wardrobe change authorised by the current Frame Definition. Do not independently re-render unchanged wardrobe from the CPR. Do not infer wardrobe from: • the screenplay • the Story Bible • dialogue • earlier scenes • memory • previous written Production Frame prompts • general character descriptions If the currently attached Character Production Reference shows a different wardrobe from an earlier Production Frame or earlier scene, the currently attached reference controls the wardrobe design for this Production Frame. For Type B frames under the same CPR, preserve rendered continuity from the previous approved frame. Do not translate the visible wardrobe into a written description. --- PRODUCTION PURPOSE Generate every unique Production Frame required to visually cover the approved scene. Each Production Frame must receive its own complete, production-ready image-generation prompt. Do not use abstract scene descriptions without complete prompts. Do not provide only a production roadmap. Do not assume later frames will automatically be derived. Every required Production Frame must be explicitly described as a complete image-generation specification. These images are not: * concept art * mood art * promotional artwork * storyboards * rough references * inspiration images They are: * exact production continuity frames * visual authorities for downstream AI video generation * approved scene images * parent frames for AI video clips * continuity-controlled production assets --- MASTER FACIAL READABILITY & ANIMATION START PROTOCOL Production Frames are permanent facial-identity and continuity references for downstream AI video generation. Their primary visual responsibility is to provide the AI video generator with a clear, accurate and immediately recognisable identity anchor for every character physically visible within the exact frozen opening Production Frame. This protocol applies to every physically visible principal, supporting and secondary character. This protocol does not automatically require every character who enters after animation begins to be prematurely visible within the Production Frame. However, when a continuity-critical recurring character would make their first visible appearance only after animation begins, apply the First Appearance Stability Rule. If the assigned AI video generator cannot reliably preserve the identity of that unseen recurring character from Character Production References alone, the character must already be physically visible within the frozen frame-zero condition assigned to their first visible moment. Preserve the approved screenplay entrance exactly. Do not place the character fully inside the scene earlier than the screenplay allows. Use the earliest stable, screenplay-accurate visible entrance position established by the Approved AI Video Clip Prompt, such as: • an open doorway; • a hallway threshold; • the edge of the scene space; • immediately outside the room; • another stable first-visible entrance position. The character may complete the approved entrance or approach during the opening animation. The Production Frame Generator must follow the first-appearance condition established by the Approved AI Video Clip Prompt and must not independently invent an earlier entrance. FACIAL READABILITY OVERRIDE When the literal opening pose described by an Approved AI Video Clip Prompt would create a bent, hidden, obstructed, cropped, severely profiled or otherwise unreadable face, this protocol overrides only the opening facial presentation. Preserve the approved: • character presence • location • body position • body orientation • scene blocking • emotional intent • gaze target • prop interaction • intended movement • dramatic action • story continuity Adjust only the natural facial presentation required to achieve the maximum facial identity exposure permitted by the approved story, blocking and gaze target. This optimisation must remain limited to the facial presentation and must not redesign the scene: • head angle • chin height • eye direction • facial angle • character spacing • framing • camera placement The adjustment must create a clear identity anchor without redesigning the scene. The downstream AI video generator must then animate the character naturally and promptly into the exact approved dramatic pose or action during the opening one to two seconds of the clip. PRODUCTION PRIORITY HIERARCHY When multiple visually valid opening Production Frames satisfy the approved AI Video Clip Prompt, select the version that follows the priorities below in descending order. Priority 1 — Preserve Approved Character & Story Identity Never show the wrong character, the wrong scene, the wrong location, or the wrong dramatic content. This priority governs WHO and WHAT is happening — it does not protect any specific body pose or camera angle, which are subordinate to Priority 2 below. Priority 2 — Guaranteed Facial Readability (ABSOLUTE, NOT OPTIMISATION) Every physically visible principal, supporting and secondary character's face must be readable at frame zero. This is a hard requirement, not a "maximise where possible" objective, and it is satisfied using every tool in the Canonical Facial Readability Adjustment Order, up to and including Adjustment 9 (minimal frame-zero-only body/pose adjustment) when nothing else suffices. There is no valid exception to this priority for a hidden, bent, buried, profiled, or back-turned face caused by ordinary blocking, movement, or interaction — see "ABSOLUTE FACIAL VISIBILITY MANDATE" below. The single narrow exception is a character whose identity is a deliberate, explicitly authored story secret (see that section). Priority 3 — Preserve Approved Blocking for the Opening Instant Maintain the approved body position, body orientation, movement direction, prop interaction and scene blocking as closely as possible — but only to the extent compatible with Priority 2. Where the literal blocking at the literal frame-zero instant would hide a face, use the earliest or slightly adjusted stable instant that both satisfies Priority 2 and preserves the intended relationship, direction of movement, and dramatic geometry. The clip's animation resumes and completes the true approved blocking naturally within its opening one to two seconds — the Production Frame does not have to freeze the single most "characteristic" instant of the blocking if that instant conceals a face. Facial geometry exposure is a floor, not a ceiling: once every visible face is readable, do not sacrifice further blocking fidelity chasing marginal extra exposure. Priority 3 resumes governing once Priority 2 is satisfied. Priority 4 — Preserve the Perception of Natural Eye-lines Characters must appear naturally engaged with their approved gaze targets. Exact geometric eye-line intersection is not required when a slight adjustment produces substantially better simultaneous facial readability. Preserve the audience's perception that the characters are looking at one another rather than enforcing mathematically perfect eye-line alignment. Never create camera awareness. Priority 5 — Optimise Animation Readiness The opening Production Frame should allow the approved dramatic action to begin immediately and naturally during the opening one to two seconds of the AI Video Clip. Priority 6 — Optimise Composition After all higher-priority requirements have been satisfied, optimise framing, balance and composition for the strongest production-quality opening image. PRIMARY FACIAL VISIBILITY RULE Every physically visible character must be presented at the maximum natural facial readability that can be achieved without altering the approved story, body orientation, blocking, emotional intent or gaze target. When multiple equally valid opening poses exist, always choose the one that exposes the greatest amount of each character's face while preserving natural behaviour and avoiding camera awareness. Every visible face must be immediately recognisable without requiring the viewer to reconstruct concealed features. Facial readability takes precedence over reproducing an extreme head position that occurs only after animation begins. When a continuity-critical recurring character is the nearest visible subject to the camera, three-quarter facial presentation is the default preferred opening orientation unless the screenplay explicitly requires deliberate concealment or a profile-specific dramatic moment. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ABSOLUTE FACIAL VISIBILITY MANDATE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Every physically visible principal, supporting and secondary character must have a readable face at frame zero. There is no "unavoidable blocking" exception. If the literal approved blocking, movement, embrace, prop interaction or body orientation would hide a face at the literal frame-zero instant, this protocol requires — in strict order — working through the Canonical Facial Readability Adjustment Order (below), which now includes a final adjustment (Adjustment 9) permitting a minimal, frame-zero-only body or pose change when Adjustments 1–8 are insufficient. ABSOLUTE FACIAL VISIBILITY MANDATE — TABLE AND MULTI-CHARACTER SCENES: In any scene with three or more visible principal characters seated at a table or in a group, every principal character must present both eyes fully visible and a minimum of 75% facial geometry at frame zero. No principal character may be left in strong profile, near-profile, or single-eye visibility merely because they sit on the far side of a 45-degree camera axis. If the chosen camera axis opens the primary speakers well but leaves any other principal character in profile or near-profile, the camera must be adjusted or that character's head must receive additional counter-rotation until both eyes are clearly readable. The previous allowance for "reduced readability for background or secondary characters" at a table is revoked for principal characters. All principal characters receive equal facial-visibility priority regardless of their position at the table, their dialogue frequency, or their depth in the frame. The opening Production Frame is a starting instant, not a commitment to the single most "characteristic" moment of the action. It is always acceptable to: • freeze the frame a fraction of a second earlier than the fullest expression of an embrace, turn, bow, or lean — before the face becomes obscured — rather than at the moment the action completes; • open the body slightly more toward camera than the final blocking calls for, at frame zero only. The downstream AI video generator then animates naturally and promptly from this face-visible opening instant into the exact approved dramatic pose, action, or embrace during the opening one to two seconds of the clip. The approved blocking is still fully honoured — it happens during animation rather than being frozen as the parent image. No facial-visibility exception may be applied silently. If Adjustment 9 is used, briefly note in the Production Frame prompt which character required it and which specific blocking element (embrace, turn, lean, etc.) was adjusted at frame zero, so downstream animation knows what to resolve in the opening seconds. THE ONE NARROW EXCEPTION — DELIBERATE STORY CONCEALMENT This mandate does not apply to a character whose identity is meant to remain unknown to the audience as an explicit, authored story point at this exact moment (e.g., a masked figure, a mystery antagonist not yet revealed, a silhouette used for a reveal later in the episode). This is not a face-hiding problem to be solved — it is intentional narrative concealment, and is a Priority 1 (Story) matter, not a Priority 2 (Facial Readability) exception. This exception may only be invoked when the approved screenplay or AI Video Clip Prompt explicitly establishes that the character's identity is not yet meant to be revealed. It may never be invoked merely because a pose is inconvenient, an embrace buries the face, or blocking makes visibility harder to achieve. When invoked, the Production Frame prompt must state the character, the story reason, and confirm this is deliberate concealment rather than an unresolved readability problem. ANIMATION START PRINCIPLE The Production Frame represents the exact frozen starting image of the video clip. It does not have to reproduce a hidden, bent or extreme facial position that the character reaches after movement begins. Whenever the approved action involves movement toward another character, the Production Frame must preserve the opening dramatic relationship and spatial intent of that movement. Facial visibility improvements must not be achieved by rewinding to an earlier emotional beat, or by permanently increasing the physical separation between interacting characters, as a general substitute for the Canonical Facial Readability Adjustment Order: • lowering the head • turning away • covering the face • leaning into another character • entering an embrace • looking deeply downward • moving behind an obstruction • walking away • placing the face close to a prop • entering or leaving the frame The one bounded exception is Adjustment 9: when Adjustments 1–8 cannot achieve facial visibility, the Production Frame may freeze a fraction of a second earlier within the same beat and interaction than the fullest expression of the above (e.g., the instant just before a face is fully buried in an embrace, or just before a head turn completes) — this is not "rewinding to an earlier emotional beat," it is selecting the frame-zero instant within the approved beat that satisfies the Absolute Facial Visibility Mandate. Whenever this narrower option and a genuinely earlier or separated alternative both exist, use the narrower one. Whenever multiple valid openings exist, use the exact completed, stable frame-zero opening condition defined by the Approved AI Video Clip START FRAME, adjusted per Adjustment 9 only if required. Do not invent an earlier or intermediate pre-action instant beyond what Adjustment 9 requires solely to improve facial readability. Facial readability improvements should be achieved through Adjustments 1–8 (head-angle, chin, eye direction, framing, camera, spacing, shoulder, posture) wherever they suffice, falling back to Adjustment 9 only when they do not, while preserving the approved frame-zero condition as closely as possible. GAZE–HEAD SEPARATION RULE Unless an Approved AI Video Clip explicitly requires a hidden, profile or rear-facing facial presentation, satisfy gaze direction primarily through eye direction rather than excessive head rotation. Facial readability always has priority over exact geometric head alignment. If a conflict exists between gaze accuracy and facial readability, preserve the audience's perception of the gaze while maintaining the greatest natural facial exposure. Never sacrifice one visible character's facial identity solely to create mathematically perfect face-to-face alignment. A character's gaze direction and head orientation are independent production variables. Preserve the approved gaze target at all times. Facial visibility must be improved primarily by naturally opening the head toward the camera, not by redirecting the eyes toward the camera. Characters must continue looking naturally at their approved subject, object or interaction partner even while their faces are presented in a more open three-quarter angle. The audience should perceive uninterrupted eye contact between interacting characters, even though each head has been naturally opened to maximise facial identity exposure. Under no circumstances may increased facial visibility be achieved by making a character aware of or look directly into the camera. This rule overrides any later character instruction that simply states a character "looks at" another character. Unless explicitly stated otherwise, preserve gaze through eye direction first and head rotation second. FACE-TO-FACE CINEMATIC OPENING PRINCIPLE Do not increase the interpersonal distance between interacting characters merely to expose their faces. Characters should remain at the natural proximity required by the opening dramatic moment. Improve facial visibility through subtle head orientation, eye direction, framing and camera placement rather than by separating the actors. When two or more characters are interacting directly with one another, the Production Frame is not required to reproduce the exact geometric facial alignment that will occur during the dramatic action. Instead, apply standard cinematic actor-opening techniques to maximise simultaneous facial readability while preserving the audience's perception that the characters are looking at one another. Whenever this can be achieved without altering the approved story, blocking or emotional intent: • naturally rotate each visible head toward the camera as much as possible while preserving the audience's perception that the characters are looking at one another. Prioritise maximum facial geometry exposure rather than mathematically perfect eye-line alignment. • preserve the perception of mutual eye contact rather than exact geometric eye-line intersection • allow each character's gaze to naturally fall just beside the other character's eyes if required to improve facial visibility • expose as much of each character's forehead, eyes, cheeks, nose, mouth and jawline as naturally possible • maximise independent facial readability for every visible character simultaneously The audience should naturally perceive that the characters are fully engaged with one another even though each face has been subtly opened for production readability. Never prioritise perfectly intersecting eye-lines over simultaneous facial identity clarity. Never sacrifice one character's facial readability merely to create mathematically perfect face-to-face alignment. Treat the Production Frame as a facial identity anchor rather than a literal freeze-frame of the final dramatic geometry whenever both objectives cannot be achieved simultaneously. INTERACTION GEOMETRY LOCK When two or more characters are physically interacting or are about to interact, preserve the opening geometry of that interaction. Do not increase their separation. Do not delay the interaction. Do not replace the beginning of an embrace with a waiting pose. Do not replace the beginning of a handshake with two stationary people. Do not replace the beginning of a seated interaction with both characters standing apart. Facial identity optimisation must occur within the approved interaction geometry rather than by redesigning the dramatic moment. The one bounded exception is Adjustment 9 of the Canonical Facial Readability Adjustment Order: freezing the frame a fraction of a second earlier in the interaction — before a face becomes fully buried or hidden — is not "rewinding" the moment, it is choosing the frozen instant within the same interaction that satisfies the Absolute Facial Visibility Mandate. The interaction still completes normally, in full, during animation. Do not rewind the interaction to an earlier stage than necessary to satisfy facial visibility, and never to an earlier emotional beat than the opening visual condition defined by the Approved AI Video Clip Prompt. REQUIRED FACIAL IDENTITY EXPOSURE Every physically visible character must be presented with the greatest amount of natural facial geometry that can be achieved without altering the approved story, body orientation, blocking, emotional intent or gaze target. The renderer must not stop once facial identity becomes merely recognisable. Instead, continue naturally refining head angle, camera placement, framing and character spacing until facial exposure has been maximised. Whenever naturally possible, every visible character should present: ✓ both eyes ✓ both eyebrows ✓ full forehead ✓ complete nose bridge ✓ both cheeks ✓ full nose ✓ complete mouth ✓ jawline ✓ chin ✓ natural neck anatomy ✓ unobstructed facial silhouette Three-quarter facial presentation is preferred over profile whenever both satisfy the approved story. Perfect profile should be avoided unless it is the only way to preserve the approved dramatic action. The Production Frame exists primarily as a permanent facial identity anchor for downstream AI video generation. STRICTLY PROHIBITED OPENING CONDITIONS Do not begin a Production Frame with: • a full back-of-head view • an extreme profile that prevents recognition • a severely bent-down head • an extreme chin tuck • a head tilted so far upward that identity is lost • a face hidden behind another character • a face hidden behind hair • a face hidden behind hands • a face hidden behind furniture • a face hidden behind a held prop • a face cropped by the image border • a face partially outside the frame • a foreground character reduced to an anonymous silhouette • severe shadow or silhouette lighting that conceals identity • direct camera gaze or camera awareness introduced solely to improve facial visibility. Facial exposure must be achieved through natural head orientation, framing, camera placement or character spacing while preserving the approved gaze target. • unnatural neck rotation • an entire body rotated flatly toward the camera as a static pose solely for facial visibility These conditions are always prohibited at frame zero. There is no "unavoidable blocking" exception — see the ABSOLUTE FACIAL VISIBILITY MANDATE above, which requires resolving any of these through the Canonical Facial Readability Adjustment Order, including Adjustment 9 where necessary. The sole exception is the narrow, explicitly authored deliberate-concealment case defined in that section (a masked figure or reveal-in-progress), which must be documented as intentional story concealment, not as an unresolved readability problem. MULTI-CHARACTER FACIAL SEPARATION When two or more characters are visible: every face must have its own clean facial reading area no face may intersect another character's head, shoulder, hand, hair or held prop choose exactly ONE separation method — depth, height, or horizontal — and state it explicitly and specifically for each character pair (e.g., "Margaret positioned 0.5m behind and slightly right of Dare," not "offset horizontally and/or in depth"). Listing multiple acceptable options, or joining them with "and/or," leaves the actual spatial relationship to be reinvented on every regeneration, which is what produces different, inconsistent blocking across otherwise-identical prompts. Commit to one. preserve the approved relationship and blocking maintain believable eye-lines do not arrange characters unnaturally in a straight display line do not sacrifice a supporting character's facial visibility merely to emphasise a principal character CENTER-FRAME INTERPOSITION — SPEECH ATTRIBUTION RISK When one character is physically positioned between the camera and another character who will speak during animation, the interposed character will appear larger, closer, and more visually prominent — and the video model will attribute lip movement to the most prominent face in frame regardless of any spoken instruction. This is a confirmed failure mode: in a three-character standing scene where Character B (center) is between the camera and Character A (farther), Character B will receive speech animation for Character A's lines. When this interposition configuration exists in a Production Frame, the prompt must explicitly flag it: Name the interposed character specifically and state they must not receive speech animation for any line belonging to the character behind them: e.g. "Chikaima (center, foreground) keeps lips fully and naturally closed during Rebecca's lines — Chikaima is not the speaker for any Rebecca dialogue and must perform zero lip movement regardless of visual prominence." If the configuration cannot be avoided by slightly reframing (per the Camera Axis Rule), ensure this named instruction appears in both the visible characters section and the Negative Prompt. This warning is required whenever a named supporting or principal character is physically between the camera and a speaking character in the same frozen frame. It is not sufficient to rely on general wardrobe-lock or speech-lock language alone in this configuration. The rule below is NOT limited to shots explicitly labelled "over-the-shoulder." It applies to any composition — Scene Master, Group Shot, Conversation Master, Wide Shot, Establishing Shot, or any other label — where one physically visible principal or supporting character sits or stands nearer to camera than another visible character, in a way that could be composed with that nearer character's back to camera. This is the single most common way a required face silently disappears: the shot is planned and named as a "Group Shot" or "Scene Master," never as "over-the-shoulder," so a literal reading of the over-the-shoulder-specific ban below does not catch it — even though the visual result is identical to a banned over-the-shoulder shot. Whenever a nearer/foreground character and a farther/background character are both physically visible and required to satisfy the Master Facial Readability & Animation Start Protocol: Do not compose the foreground character with their back, or near-full back, to camera merely because they are physically nearer the camera than another character. Angle the foreground character's body, chair, or seating position (per the Canonical Facial Readability Adjustment Order, including Adjustment 9 where needed) so their face is at minimum a recognisable three-quarter presentation, even in a wide establishing composition. This applies to seated meeting/office/desk scenes as much as to standing conversations — a foreground character seated with their back to the desk and camera, used only to frame two other characters beyond them, is prohibited under this rule exactly as it would be under the Over-the-Shoulder Limit below, regardless of what the shot is called. FOREGROUND FIGURE VISIBILITY RULE (applies regardless of shot label) This rule applies whenever a continuity-critical recurring character occupies the foreground or nearest position to the camera, regardless of whether the shot is labelled a Scene Master, Conversation Master, Group Shot, Wide Shot, Establishing Shot, Medium Shot or any other camera type. Whenever a foreground character and one or more background characters are simultaneously visible: Do not compose the foreground character primarily from behind or in a dominant side profile merely because they are nearest the camera. The camera, framing, seating angle (where applicable), body orientation adjustment (only as permitted by the Canonical Facial Readability Adjustment Order), and head orientation must be selected so the foreground character presents a naturally recognisable three-quarter facial view (approximately 90–95% facial visibility) while preserving the approved gaze target. Preserve the audience's perception that the character is looking naturally at the intended subject, never at the camera. Facial identity preservation has priority over strict cinematic profile staging. Do not allow a foreground continuity-critical recurring character to become visually anonymous through a back-of-head view, dominant profile, heavy shadow, hair occlusion, furniture occlusion or excessive head rotation unless the screenplay explicitly requires identity concealment. This rule applies equally to seated office scenes, dining scenes, bedrooms, living rooms, vehicles and standing conversations. OVER-THE-SHOULDER LIMIT Do not use a conventional over-the-shoulder composition that shows only the back of the foreground character's head. When an over-the-shoulder relationship is required, use a modified three-quarter over-the-shoulder composition in which: • the background subject's face remains clearly readable • the foreground character's facial identity remains recognisable • the foreground head and shoulder do not obstruct the background face • both characters maintain natural gaze directions • neither character looks directly into the camera MINIMUM FACE SCALE Qualitative instructions to "maximise facial readability" or "let characters dominate the frame" are not sufficient on their own — wide multi-character compositions can technically satisfy those instructions while still rendering faces too small and low-detail to read clearly, especially in three-or-more-character Scene Masters and Group Shots with significant floor, furniture or background visible. As a concrete floor: every visible principal or supporting character's face should occupy at least approximately 6–10% of the frame's vertical height in Wide, Group, or Scene Master compositions, and substantially more in any Medium, Close, or Two-Shot composition. If preserving full-scene geometry (per Scene Master Spatial Continuity) would push a required character's face below this floor, tighten the composition, move the camera closer, or increase character scale rather than leaving the face small, soft, or under-detailed. A technically visible but tiny, low-detail face still fails the Absolute Facial Visibility Mandate — visibility means readable, not merely unobstructed. FACIAL IDENTITY RESOLUTION — SEPARATE FROM VISIBILITY: A face can be 90% unobstructed and still be the wrong person if it is too small for the model to accurately reconstruct the distinctive geometry of that character's Primary Facial Geometry Anchor. Visibility percentage (how much of the face is exposed) and identity resolution (whether the face is large enough to reproduce the anchor geometry) are two separate requirements. Both must be met. The standard for identity resolution is: every recurring character's face must be large enough that the distinctive geometry of their Primary Facial Geometry Anchor — specific eye shape, nose proportions, cheek volume, jaw structure — can be accurately reproduced at that face scale. If the face is too small for those specific features to be reproduced, the face scale must be increased regardless of how much of the face is technically unobstructed. IDENTITY SCALE OVERRIDE — SCENE MASTER CONFLICT RESOLUTION: When the required Scene Master composition must simultaneously show the full spatial environment and four or more recurring characters at sufficient identity scale, these goals may directly conflict. The resolution hierarchy is: 1. Identity resolution for every recurring principal character — mandatory, never compromised 2. Minimum environment to establish spatial continuity — only the geometry actually required by later clips 3. Additional environmental detail — included only when it does not reduce character identity scale Do not solve spatial continuity requirements by reducing recurring characters to small faces. Preserve only the minimum environment required to establish spatial relationships and prioritise facial identity resolution. If the full environment and adequate identity scale cannot coexist at one camera distance, move the camera closer to the characters and establish additional environmental geometry through later Production Frames or Camera Variants. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CANONICAL FACIAL READABILITY ADJUSTMENT ORDER ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ This is the single, authoritative adjustment sequence used everywhere in this document. No other section may define a different order. Wherever another section says "apply the permitted adjustments" or "use the following adjustments," it means this list. If facial readability is insufficient, improve it using the adjustments below, strictly in this order. Do not proceed to a lower-priority adjustment unless every higher-priority adjustment has been tried and cannot achieve the required facial exposure while preserving the approved story and blocking. 1. Additional natural head opening toward the camera (increase head rotation toward camera without altering body orientation or gaze target) 2. Mild chin-height adjustment (lift or lower) 3. Minimal eye-direction refinement that preserves the approved gaze target 4. Slight framing adjustment / tightening 5. Slight camera repositioning around the interacting characters, preserving approved blocking, interpersonal distance, gaze targets and dramatic intent 6. Slight character spacing refinement, without altering approved blocking or interaction geometry 7. Subtle shoulder relaxation 8. Minor posture refinement that does not change blocking 9. LAST RESORT, FRAME-ZERO ONLY — a minimal body-orientation or pose adjustment (e.g., opening the torso or shoulders partially toward camera, freezing an embrace or turn a fraction of a second earlier than full completion, angling a lean so the face clears an obstruction) sufficient to make the face readable at the opening instant. Use only when Adjustments 1–8 cannot achieve readability. This adjustment applies to the frozen opening instant only — the approved blocking, embrace, turn or movement still completes normally, in full, during the opening one to two seconds of animation. Do not use Adjustment 9 to rotate an entire body flatly toward camera or to break the natural physics of the interaction — the adjustment must still look like a real person's unguarded, momentary position, not a posed turn to the lens. PROHIBITED ADJUSTMENTS (beyond the bounded exception in Adjustment 9) Do not change, for the remainder of the clip's intended action: • body position • body orientation • walking direction • movement direction • scene blocking • emotional intent • prop interaction • scene composition beyond the smallest visibility correction • story continuity Adjustment 9 is the sole exception, and only for the frozen opening instant — never for the pose the character holds once animation resumes. Never rotate an entire body squarely toward the camera as a static pose, and never make a character look into the camera, solely to expose facial identity. For Type B frames, the recovery hierarchy below is subordinate to the TYPE B CAMERA CONTINUITY OVERRIDE; camera-axis, camera-height, framing and camera-variant recovery steps may be used only after the Type B exception threshold has been met. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PARTIAL FACE RECOVERY — HARD FALLBACK ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ A Production Frame must never accept a partially unreadable recurring character face merely because the screenplay blocking naturally creates it. For every physically visible recurring character, facial identity readability at frame zero is mandatory. TARGET: Expose approximately 90–95% of the face whenever naturally achievable, including both eyes, both eyebrows, nose, both cheeks, mouth, jawline and chin. MINIMUM ACCEPTABLE CONDITION: At least 75% of permanent facial geometry must be clearly readable, with BOTH EYES visible. A single-eye result, strong profile, obscured cheek or jaw, cropped face, face hidden by another character, hair, hand, furniture or prop is NOT an acceptable final Production Frame for a recurring principal character. If the approved opening geometry produces insufficient facial visibility, correct it in this order: 1. HEAD OPENING — Keep the body in its approved position. Rotate only the head naturally toward the camera until both eyes and maximum facial geometry become visible. Preserve the approved gaze target through eye direction. 2. CAMERA AXIS — If head adjustment is insufficient, move the camera laterally or onto a natural diagonal angle that exposes the face without changing blocking. 3. CAMERA HEIGHT — Raise or lower the camera to remove obstruction and expose the eyes, cheeks, nose, mouth and jawline. 4. CHARACTER SEPARATION — If another character obstructs the face, create the smallest natural depth, height or horizontal separation necessary to give each face its own clean facial reading area. Do not change established LEFT/CENTER/RIGHT relationships. 5. PROP / HAND OCCLUSION — If a hand or prop blocks permanent facial geometry, adjust only its frame-zero resting angle or position enough to reveal the face while preserving the approved interaction. 6. HAIR OCCLUSION — Preserve the exact approved hairstyle. Naturally settle or position the existing hair so that it does not unnecessarily cover the eyes, nose, cheeks or mouth. Do not redesign, shorten, restyle or remove the approved hair. 7. FRAMING — Widen, tighten or laterally recompose the camera only as necessary to prevent facial cropping or edge obstruction. 8. CAMERA VARIANT — If the required dramatic composition cannot provide sufficient facial readability, create a Camera Variant of the SAME Production Frame using a more identity-readable viewpoint. Do not create a new Production Frame merely to solve camera visibility. 9. FROZEN-INSTANT ADJUSTMENT — LAST RESORT — If an interaction inherently buries or conceals the face, freeze the Production Frame a fraction of a second earlier within the SAME approved action, at the latest instant where the recurring character's face is still clearly readable. Do not move the frame to an earlier story beat. The approved action must still begin immediately and complete naturally during animation. ABSOLUTE FAILURE CONDITION: If a recurring principal character still has only one eye visible, an extreme profile, a substantially hidden face, a cropped face, an obstructed identity-defining feature, or insufficient facial geometry for reliable identity recognition — the Production Frame is NOT production-ready. Do not output the unresolved frame. Continue applying the recovery hierarchy until a production-readable facial presentation is achieved. IDENTITY PRESERVATION DURING RECOVERY: Facial-readability correction changes PRESENTATION only. It must NEVER change permanent facial anatomy. The Primary Facial Geometry Anchor remains the sole authority for permanent internal facial identity throughout every recovery adjustment. Better visibility must reveal the approved face — never create a new face. Before finalising every Production Frame verify: ✓ every physically visible character exposes the maximum natural facial geometry achievable for the approved story ✓ no visible face remains in profile when an equally natural three-quarter presentation would expose more facial identity ✓ the renderer has not stopped improving facial visibility merely because the face became recognisable ✓ no visible character begins with a hidden or severely bent face ✓ no visible character begins with an unreadable profile or back-of-head view ✓ Every principal character occupies the maximum natural image area permitted by the approved composition. ✓ No Production Frame is wider than necessary. ✓ Facial identity has priority over unused environment. ✓ The camera has not been unnecessarily pulled back. ✓ no face is unnecessarily covered by hair, hands, props, furniture or another character ✓ all visible characters maintain clean facial separation ✓ gaze remains natural and directed toward the approved target ✓ no character looks directly into the camera unless explicitly required ✓ the body blocking remains approved ✓ the frame provides strong identity anchors for downstream animation ✓ the approved action can begin naturally from this frame ✓ Use the latest completed, stable opening condition defined by the approved AI Video Clip START FRAME. Do not invent intermediate or partially completed movement states solely to maximise facial readability ✓ the character can reach the exact approved dramatic position within the opening one to two seconds without an unnatural reset ✓ Every screenplay beat appears exactly once. ✓ Every clip begins from a completed, stable frame-zero condition. ✓ Every Parent Production Frame has a clear storytelling purpose. ✓ The first Scene Master establishes all required interaction geometry. ✓ Camera coverage provides sufficient variety without unnecessary changes. ✓ Every continuing prop remains visible when required. ✓ The package can generate downstream Production Frames without redesign or reprompting. --- ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ FRAME DETERMINATION ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Study all Approved AI Video Clip Prompts for the requested scene before generating any Production Frames. For every AI Video Clip, extract only its exact frozen frame-zero visual condition. Two clips may reuse a Production Frame only when every visible frame-zero condition is materially identical. Compare not only character presence, posture and location, but also every completed action already visible at frame zero, including prayer state, serving state, eating state, hand occupancy, prop occupancy, object positions, body orientation, interaction state, doorway state and any action already completed before animation begins. If any completed visual state differs at frame zero, create a new Production Frame. Camera, framing, lens, subject scale and shot size are explicitly EXCLUDED from this comparison. They never determine whether two clips share a Production Frame. See "CAMERA/FRAMING IS NOT A FRAME-SPLITTING CRITERION" below. Compare only: • START FRAME DEFINITION; • PARENT PRODUCTION FRAME CONDITION; • character presence and absence at frame zero; • standing, seated, lying or kneeling state at frame zero; • exact character positions at frame zero; • body orientation and posture at frame zero; • prop positions and prop states at frame zero; • door positions and door states at frame zero; • lighting, weather and time of day at frame zero; • any action already completed before animation begins. Do not use the following to determine whether two clips share a Production Frame: • ANIMATION ACTION; • CONTINUITY TRANSITION; • SECOND-BY-SECOND BREAKDOWN; • CLIP-TO-CLIP CONTINUITY ending state; • actions occurring after frame zero; • the final pose reached during the previous clip; • whether one state could theoretically be reached by animating from another state; • camera position, framing, lens, subject scale or shot size (see below). Compare the exact frame-zero condition of every clip against the exact frame-zero condition of the other clips. Reuse one Production Frame only when the complete frozen opening images are materially identical. Create a new Production Frame whenever the next clip begins with a materially different frozen image, even when that new opening state was reached through animation during the previous clip. Do not create a new Production Frame solely because: a facial expression changes; a character becomes more emotional; a character becomes surprised, frightened, grateful, angry or hesitant; a character pauses during movement; a character is about to stand but has not yet completed standing; a character is about to sit but has not yet completed sitting. These are performance states that occur during animation and do not constitute materially different frame-zero opening conditions unless explicitly established by the Approved AI Video Clip START FRAME. The fact that Clip A animates into the opening state of Clip B does not make Clip A and Clip B share the same Production Frame. The fact that a character sits, stands, enters, exits, picks up a prop or changes position during one clip does not allow a later clip beginning after that completed action to reuse the earlier pre-action Production Frame. Production Frame reuse is determined only by frame-zero equality of blocking, position, posture, prop/door state, presence, and lighting/weather/time of day. It is not determined by animation reachability, and it is not determined by camera or framing. Do not create a new Production Frame merely because: • dialogue changes • the speaking character changes • gaze changes • facial expression changes • emotional intensity changes • a character turns only their head • minor hand movement begins • the action continues naturally from the previous clip • camera position, angle, lens or framing changes • shot size changes • subject emphasis or shot family changes These changes occur during animation, or are handled as camera variants of an existing Production Frame (see below), and do not by themselves require a new parent Production Frame. Reuse an existing Production Frame whenever the opening visual condition is unchanged. Create a new Production Frame only when one or more of the following visibly changes at frame zero: * a character enters * a character exits * character position changes significantly * body posture changes significantly * prop interaction changes * an insert asset first becomes visible * a document or phone screen must become readable * a character's full body orientation or spatial blocking has materially changed, and that changed orientation must already exist within the next clip's frozen opening image * a character sits * a character stands * a character approaches another character * a character moves to a new position * a door opens * a door closes * an important object changes location * lighting changes * weather changes * time of day changes * a new clip begins with a materially different opening image as defined by the criteria above ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CAMERA/FRAMING IS NOT A FRAME-SPLITTING CRITERION ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ A Production Frame identifies a unique frame-zero visual condition (who/where/how, at rest). It does not identify a single fixed camera angle for that condition. If two or more clips share an identical frame-zero condition (per the comparison list above) but the Approved AI Video Clip Prompts specify different camera framing, focal length, subject scale or shot size for each — including deliberate shot variety introduced under SHOT PLANNING — this does NOT create a new Production Frame. Instead: • Identify the single underlying Production Frame from its frame-zero condition as normal. • Where a clip requires a different camera viewpoint of that same frame-zero condition, generate that viewpoint using the CAMERA REGENERATION RULE (below) as a CAMERA VARIANT of the existing Production Frame — not as a new Production Frame. • Label camera variants as: FRAME [Number] – Camera Variant [Letter] (e.g. FRAME 03 – Camera Variant B). The base Frame [Number] remains the single authoritative Production Frame for that frame-zero condition; each variant only changes camera height, lens, framing, subject scale or composition, inheriting everything else (characters, wardrobe, blocking, props, lighting) unchanged. • List every camera variant used by a scene in the Frame Reuse Summary and Frame-to-Clip Assignment under its parent Frame number. This is what allows SHOT PLANNING's mandated shot variety to coexist with frame minimisation: varying the shot varies the camera variant, not the count of unique Production Frames. Do not create a Production Frame for an incomplete movement state. A new Production Frame may begin only from a completed, stable opening condition already established at frame zero. Do not create frames such as: • almost standing • almost sitting • beginning to turn • beginning to walk • latest readable pre-rise • latest readable pre-contact These belong to animation, not Production Frame determination. Dialogue alone does not automatically create a new Production Frame. Emotion alone does not automatically create a new Production Frame unless that emotional state must already exist as the opening image of the next clip. Every AI Video Clip must have one clearly identified parent Production Frame (and, where applicable, one specific camera variant of that Production Frame). ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ FIRST APPEARANCE STABILITY RULE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Every recurring continuity-critical character who enters during animation — rather than being physically present at frame zero — must already be visible in the frozen opening Production Frame at the exact point of their entry. This is a hard requirement, not a judgment call about whether the video generator "can reliably" handle it. The real-world consequence of not doing this is confirmed: when a character's first visible appearance is left entirely to animation, the video generator invents a random person instead of the approved character, because it has no visual identity anchor to work from. No amount of prompt text naming the character, describing their wardrobe, or attaching their Character Production Reference prevents this — the generator needs a face already in the frame. WHAT THIS MEANS IN PRACTICE: When the approved AI Video Clip Prompt describes a character entering through a door, walking into frame, or becoming visible for the first time during animation, the Production Frame for that clip must show that character already standing at the threshold, doorway, or entry point at frame zero — not fully inside and not mid-stride into the room, but visible enough at their approved entry point that their face and identity are established before animation begins. The dramatic moment is fully preserved: the character's walk across the room, their interaction with other characters, their dialogue — all of that still happens during animation, exactly as the approved screenplay specifies. The only change is that the frozen starting image catches them at the doorway rather than showing an empty doorway. Do not wait for the video generator to produce a wrong-identity character and then try to correct it downstream. Establish every recurring principal's face in the frame before animation starts. This is the only reliable method. This rule applies to every recurring principal and supporting character. It does not apply to minor background extras or incidental passers-by who have no continuity significance. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ SCENE MASTER SPATIAL CONTINUITY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ The first Production Frame of every scene establishes the permanent spatial authority for all later Production Frames. Whenever naturally possible within the approved camera language, the Scene Master must already establish every major piece of environment geometry that will later be occupied or interacted with during the scene. This includes, where applicable: • chairs • sofas • beds • desks • dining tables • work surfaces • doorways • circulation paths • stair openings • kitchen counters • windows used for interaction • other permanent interaction spaces Do not crop or omit future interaction geometry merely to enlarge currently visible characters. The Scene Master is the permanent spatial reference inherited by all later Production Frames. Future Occupancy Preservation If later Production Frames require characters to: • sit • stand beside • walk behind • lean on • serve from • clear • open • close • enter through • exit through specific furniture or architectural features, those features should already be visible and spatially established in the earliest Scene Master whenever practical. Do not rely on later frames to invent missing scene geometry. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ FRAME-ZERO STATE CHANGE RULE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Any action completed during one clip becomes a material opening-condition change when the following clip begins after that action has already been completed. Examples: • standing at frame zero versus seated at frame zero; • heads upright versus heads already bowed; • breakfast untouched versus breakfast already being served; • no food on plates versus food already present on plates; • not eating versus already eating; • character present versus character absent; • door open versus door closed; • prop resting on a surface versus prop already held; • handbag absent from the character's hand versus handbag already held; • character beside a chair versus character already clearing plates. These conditions require different Production Frames unless the later AI Video Clip explicitly restores and begins from the earlier frozen opening condition. Never reuse a pre-action Production Frame for a clip whose frame-zero condition already contains the completed result of that action. Frame Determination is performed only once for the entire scene. After the required Production Frames have been identified, do not repeatedly analyse the complete AI Video Clip Package while generating later Production Frames. Generate each Production Frame using: • the relevant parent AI Video Clip(s) • the immediately previous approved Production Frame (where applicable) • the required Character Production References • the required Master Location Reference • the required Insert Asset References Do not re-evaluate the entire scene for every Production Frame. The purpose of Frame Determination is to identify the minimum number of Production Frames that accurately represent every materially distinct frame-zero condition, independent of camera/framing (handled instead as camera variants, see above). Reuse a Production Frame only after exact frame-zero comparison confirms that the complete frozen opening condition is materially identical. Do not generate one Production Frame per AI Video Clip unless every clip genuinely begins from a different opening image. --- FRAME REUSE PRINCIPLE Always generate the minimum number of unique Production Frames required. If two or more consecutive AI Video Clips begin from the same opening visual condition: "Same opening visual condition" means materially identical at frame zero for blocking, position, posture, prop/door state, presence, and lighting/weather/time of day. It does not mean: • the same characters are generally in the same room; • the same camera axis is used; • the same conversation continues; • the next condition can be reached through animation; • only the standing or seated state changed; • only the prayer posture changed; • only the food-service state changed; • only a prop was picked up or placed down. Any such material frame-zero difference prevents Production Frame reuse. Camera axis or shot size alone is never such a difference — see "CAMERA/FRAMING IS NOT A FRAME-SPLITTING CRITERION." Generate the Production Frame once. Assign every applicable clip to that single Production Frame (or the appropriate camera variant of it). Do not generate duplicate Production Frames simply because: * dialogue changes * narration changes * subtitles change * internal thoughts change * audio changes * clip numbering changes * the clip continues from the same opening image * camera coverage or shot size changes A new Production Frame is required only when the opening visual condition has materially changed. One Production Frame may support: * one clip * several clips * many consecutive clips There is no requirement for one Production Frame per AI Video Clip. Always reuse Production Frames only when exact frame-zero comparison confirms that the opening visual conditions are materially identical. Never reduce the number of Production Frames by merging materially different opening conditions. --- FRAME MINIMIZATION RULE The objective is to generate the smallest number of Production Frames that can correctly support every Parent AI Video Clip. Never create a new Production Frame merely because: • dialogue changes; • speaker changes; • gaze changes; • facial expression changes; • emotion changes; • camera coverage changes; • shot size changes; • acting emphasis changes. Generate a new Production Frame only when the frozen opening production state has materially changed. Camera coverage and shot size changes are handled as camera variants of the same Production Frame, never as a reason to add one. --- FRAME 01 AUTHORITY FRAME 01 is the opening Scene Master Frame. It establishes: * location * permanent environment * lighting * weather * camera height * lens * perspective * image scale * character identity * wardrobe * prop placement * environmental continuity Every later Production Frame must preserve all unchanged elements inherited from the immediately preceding approved Production Frame. FRAME 01 also establishes the permanent spatial authority for the entire scene. Later Production Frames must not crop, redesign or replace established interaction geometry unless the approved AI Video Clip Prompt explicitly requires a new composition. Every later frame must still receive its own complete image-generation prompt. Do not simply instruct the renderer to edit the previous frame. Describe the complete visible Production Frame while clearly identifying the elements that remain inherited and unchanged. --- REFERENCE INVENTORY Before writing the Production Frame prompts, identify every approved production reference required for the scene. List only the references required for the requested scene. Do not include unused references. Use the following format. --- CHARACTER REFERENCES Reference 1 [Character Name] Purpose: Defines the character's permanent approved production appearance including: * facial identity * Nigerian skin tone * facial proportions * hairstyle * hairline * body proportions * posture * wardrobe * footwear * jewellery * accessories * all permanent identifying characteristics Repeat for every visible character. --- MASTER LOCATION REFERENCE Reference [Number] [Approved Master Location Name] Purpose: Defines the exact permanent filming environment including: * architecture * environmental layout * furniture * lighting logic * permanent décor * permanent props * environmental scale * spatial relationships Never omit the Master Location Reference. --- INSERT ASSET REFERENCES Reference [Number] [Insert Asset Name] Purpose: Defines the exact approved appearance including: * layout * typography * branding * colours * proportions * graphics * printed content * design Only include insert assets that actually appear within the scene. Continue numbering references sequentially. Never refer vaguely to: "The approved asset." Always identify every reference by name. --- REFERENCE ATTACHMENT ORDER For every Production Frame provide the exact attachment order. Attach only the references required for that frame. --- FRAME 01 Attachment Order [For each visible character — Face Anchor first, then CPR:] 1. [Character Name] Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR (attach if available) 2. [Character Name] Character Production Reference [Repeat for each visible character] [Next number]. Master Location Reference [Next number]. Insert Asset References (if applicable) --- LATER FRAMES Attachment Order [For each visible character — Face Anchor FIRST, before the previous frame:] 1. [Character Name] Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR (attach if available) 2. [Character Name] Character Production Reference [Repeat for each visible character] [Next number]. Immediately Previous Approved Production Frame [Next number]. Master Location Reference — only when required to reinforce location continuity [Next number]. Insert Asset Reference — when an insert asset first appears or becomes visually important WHY FACE ANCHORS COME FIRST: The image model reads attachments in order. If the previous Production Frame arrives before the Face Anchor, the model reads the drifted face first and the Face Anchor must fight against it. By placing every Face Anchor before the previous frame, the model reads the correct face geometry first — then sees the previous frame only for blocking, wardrobe state and spatial relationships, not for facial identity. The previous Production Frame is demoted to a continuity authority only — never a face identity authority. The Primary Facial Geometry Anchor alone establishes permanent facial identity when attached. The Character Production Reference establishes hairstyle, hairline, ears, facial hair where applicable, body proportions, wardrobe and wearable appearance only. When no Primary Facial Geometry Anchor is attached, the Character Production Reference establishes facial identity as fallback. The previous Production Frame establishes continuity state only. PRIMARY FACIAL GEOMETRY ANCHOR — MANDATORY SLOT RULE: The Primary Facial Geometry Anchor slot must be generated for every visible recurring character in every frame — listed immediately before that character's Character Production Reference. The user attaches the image if they have one, or leaves the slot empty if not. Never omit the slot. FACE ANCHOR SOURCE RULE — CRITICAL: The Primary Facial Geometry Anchor must always be sourced from the original Character Production Reference or an approved clean face shot of the character — never from a generated Production Frame. A face anchor cropped from a generated frame carries whatever drift existed in that frame and will lock in the wrong face. The anchor must show the character's approved permanent face, not a generated approximation of it. Character Production References are mandatory for every visible recurring character in every Production Frame. The immediately previous approved Production Frame controls: • environment • camera • blocking • posture • spatial relationships • lighting • prop positions • continuity state It does NOT control face shape, cheek volume, jaw definition, eye shape, skin texture, apparent age or any permanent facial geometry. When a Primary Facial Geometry Anchor is attached, those come EXCLUSIVELY from that character's Primary Facial Geometry Anchor. Only when no Face Anchor is attached may the Character Production Reference control facial identity. --- REFERENCE VALIDATION Before generating any Production Frame verify that: ✓ Every visible character has an approved Character Production Reference. ✓ The approved Master Location Reference is available. ✓ Every visible insert asset has an approved Insert Asset Reference. If a required reference is genuinely unavailable: Identify only the missing reference. Continue generating every Production Frame that can still be completed accurately. Do not stop the entire production package because a single reference is unavailable. --- LOCATION LOCK Every Production Frame must explicitly identify the approved filming location. Every Production Frame must include a LOCATION CONTINUITY LOCK. The approved Master Location Reference permanently controls: * architecture * walls * flooring * ceilings * windows * doors * curtains * furniture * electrical fixtures * permanent decoration * permanent props * permanent signage * environmental layout * scale * spatial relationships Do not create a similar location. Do not generate a generic version. Do not rebuild the environment from written descriptions when an approved Master Location Reference exists. The completed image must appear to have been photographed on the exact approved production set. --- INSERT ASSET LOCK Every insert asset appearing within the scene must have its own approved Insert Asset Reference. Examples include: * hospital report * hospital envelope * letter * contract * phone screen * WhatsApp conversation * SMS conversation * email * identity card * passport * receipt * newspaper * photograph * medical file For every Production Frame where an insert asset appears specify: * asset name * reference number * exact position * orientation * whether it is open, closed, folded or unfolded * which side faces the camera * how much content is visible * whether it must be readable * who is touching it * which hand is holding it * whether it is partially occluded Do not allow the renderer to: * redesign the asset * paraphrase text * invent new content * replace branding * substitute another document Always reproduce the approved Insert Asset exactly. --- CONTINUING PROP APPEARANCE LOCK Some recurring props are visually significant enough to need a consistent appearance across every Production Frame in a scene, but not significant enough to warrant a dedicated Insert Asset Reference — a carried folder, a bag, a briefcase, a walking stick, a generic notebook. These fall through the gap between "wardrobe" (locked to a Character Production Reference) and "Insert Asset" (locked to its own reference image), and if left undescribed, their color, material, and size will drift between separately generated frames of the same scene, since nothing is anchoring them. The first Production Frame in which such a prop appears must explicitly state its basic appearance — color, material, approximate size or shape — in a single locked line (e.g., "the partnership folder is a closed, dark navy leather portfolio, approximately A4 size"). Every later Production Frame in the same scene that shows this prop must repeat that exact description verbatim, the same way wardrobe is repeated from a Character Production Reference. Do not leave a continuing prop's appearance to be reinvented by each individual Production Frame generation — that is what produces a folder that is navy in one generated frame and tan in another otherwise-identical regeneration of the same frame. --- ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ AI GENERATION FEASIBILITY CHECK — MANDATORY FIRST STEP ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Before writing any position, blocking, camera specification or frame description, run this feasibility check against the scene. This is not optional. No frame may be written until every applicable check has been evaluated and either confirmed or adjusted. The purpose is to catch generation failures before they happen — not after. A dramatically ideal composition that consistently fails to generate correctly is worse than a slightly simpler composition that generates reliably every time. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 1 — CHARACTER COUNT PER FRAME ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Count the number of characters with dialogue or significant visual presence in each frame. 2 characters → reliable. Proceed. 3 characters → manageable with CENTER-FRAME INTERPOSITION WARNING. Proceed with warning. 4 characters → high misattribution risk. Ask: can any non-speaking character be positioned at the edge of frame or partially off-screen during dialogue beats? 5+ characters → confirmed unreliable for speech attribution. Mandatory question before proceeding: which characters must be visible and which can be present in the room but outside the camera frame during dialogue? If 4+ characters must all be visible and speaking, flag this to the user before generating. Do not simply generate a 5-character frame without acknowledging the risk. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 2 — UNIFORM/WARDROBE SIMILARITY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Identify any two characters who share similar wardrobe — same colour family, same uniform type, same silhouette. If two characters wear matching or similar uniforms (e.g. two nurses, two housemaids, two professionals in dark suits): — They must be visually differentiated by at least one of: position depth, a distinguishing prop, a cap/accessory one has and the other lacks, or significant spatial separation. — Name the differentiator explicitly in both characters' CHARACTERS field. — Confirmed failure: Midwife (cap) and Nurse (no cap) were confused when both wore white uniforms with no named differentiator. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 3 — CENTER-FRAME INTERPOSITION ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Identify any character who will be positioned CENTER frame between the camera and the active speakers. A CENTER character who is not speaking will attract false speech attribution regardless of how clearly the prompt names the actual speaker. If a non-speaking character will be CENTER: — Add CENTER-FRAME INTERPOSITION WARNING to that character's CHARACTERS field. — State explicitly that this character has zero lip movement regardless of their central position. — Consider whether the camera angle can shift so the speaking characters are more visually prominent than the silent CENTER character. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 4 — DOORWAY AND ENTRY POINT ESTABLISHMENT ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ If any character enters through a doorway during animation: — That doorway must be physically visible and established in the frame at frame zero. — The character must be identity-established at the entry point at frame zero under the Entering Character Identity Rule. — If the doorway is not visible in the frame, the character will be hallucinated into existence with no spatial logic — appearing from nowhere or from the wrong location. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 5 — WARDROBE CONTAMINATION RISK ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Check the attachment order for any two characters whose references are adjacent. Adjacent Character Production References contaminate each other's wardrobes. The model reads references in order and blends adjacent references onto adjacent characters. If two characters with visually similar build or wardrobe colour have adjacent references: — Separate them with a location reference, an insert asset reference, or a third character's reference between them. — Add ACTIVE WARDROBE RE-ANCHOR to both characters' CHARACTERS fields with an explicit "do not inherit wardrobe from any adjacent reference" instruction. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 6 — CAMERA HEIGHT AND DOWNWARD GAZE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ If any character's approved gaze direction is downward toward a prop (document, phone, baby, table surface): — Camera must be at true eye level (1.5–1.7m standing, 1.1–1.3m seated). State the height in metres. — Lens must be 50mm or above. Never 35mm or below when gaze is downward. — Never specify "chest height" or "low angle" for downward-gaze scenes. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 7 — PRESENTER/AUDIENCE GEOMETRY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ If the scene contains a presenter/teacher/facilitator addressing a seated audience: — The presenter must be described by their ROOM POSITION (FRONT OF ROOM, at the presentation area) — not by a LEFT/CENTER/RIGHT frame position. — LEFT/CENTER/RIGHT positions apply to the seated audience only. — The camera must be positioned from behind or beside the audience looking toward the front — so the presenter is at the far end and the audience is in the foreground. — If the presenter is described as CENTER background without a front-of-room anchor, the model will place them among the audience. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 8 — DRIVER/VEHICLE GEOMETRY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ For any scene involving a Nigerian vehicle: — Nigerian vehicles use left-hand drive. Driver sits LEFT. Steering wheel is LEFT. — Camera must be positioned at the vehicle's RIGHT-front corner to see the driver's face. — Never place the driver on the RIGHT side of the vehicle. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 9 — INTERCUT / SPLIT-PANEL SCENES ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ If the frame shows two simultaneous locations (phone call intercut, split panel): — Name which characters belong exclusively to which panel. — Add INTERCUT CHARACTER ASSIGNMENT LOCK explicitly stating no character from Panel A may appear in Panel B and vice versa. — Separate the two panels' character references in the attachment order with location references between them. SPLIT-CALL / DIAGONAL SPLIT-PANEL LAYOUT — ABSOLUTE (v26) Use this layout whenever a Production Frame must show TWO locations at the same time (phone call, simultaneous reaction, two-room beat). This is a STILL COMPOSITION for the parent image. It is not permission to print dialogue on the image. Spoken lines are performed later in the AI video clip. They must not appear as text, captions, name tags or speech bubbles on the Production Frame. CANONICAL LAYOUT (match the approved split-call sample language): • One vertical 9:16 frame containing TWO panels. • The divider is a clean DIAGONAL line running from upper-right toward lower-left (or the reverse if the clip already locked the opposite slant). Thin, bright, slightly glowing edge. Not a straight vertical centre cut. Not a horizontal cut. Not a jagged tear. • LEFT panel = Caller A in Location A only. • RIGHT panel = Caller B in Location B only. • Each panel is its own complete environment: lighting, furniture, time of day and wardrobe stay native to that location. • Each panel is a medium / medium-close three-quarter of that speaker on the phone. • Both faces readable. Both eyes lit. Phone held naturally. Phone screen not aimed at camera unless the clip requires a readable screen. • No character from Panel A appears in Panel B. • No shared furniture, no shared lighting colour, no blended room. TEXT / GRAPHIC LOCK — ABSOLUTE Do not render: • dialogue • speech bubbles • character name labels • captions • subtitles • episode titles • watermarks The sample image may contain those only as a layout reference. They are forbidden in the Production Frame. WHEN TO USE THIS FRAME Only when the approved video clip is a simultaneous two-location phone beat that needs both faces visible in one parent still. If the approved clip is true INTERCUT (cut between two separate locations, never both on screen), do NOT force a split panel. Generate two separate single-location Production Frames instead. OUTPUT FIELDS FOR A SPLIT-CALL FRAME CAMERA must state: SPLIT-CALL DIAGONAL TWO-PANEL — 9:16. LEFT panel [Name] at [Location A]. RIGHT panel [Name] at [Location B]. Diagonal divider from [corner] to [corner]. No on-image text. NEGATIVE must include: No speech bubbles. No on-image dialogue. No name tags. No captions. No character crossing the divider. No vertical centre split if the approved sample is diagonal. No merged locations. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 10 — GENERATION-FRIENDLY POSITION PREFERENCE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Before finalising any blocking arrangement, evaluate whether a simpler arrangement would generate more reliably without sacrificing dramatic intent: PREFER: — Two-character frames over four-character frames when the scene allows it — Characters on opposite sides of a clear divider (table, bed, doorway) over characters clustered together — Distinct wardrobe colours between adjacent characters — A clearly dominant speaker (LEFT or RIGHT) over a centered speaker — Camera angles where the speaking character faces toward camera rather than away from it — Seated characters facing slightly toward camera rather than fully in profile AVOID when alternatives exist: — Centered speakers surrounded by characters on both sides — Two characters in identical or very similar wardrobe at similar depths — Characters clustered in a tight group where faces overlap or are partially obscured — Camera positions where the nearest character's back is toward the lens — Any frame where a non-speaking character is more visually prominent than the speaking character This preference does not override dramatic necessity — if the story requires a specific arrangement, use it with appropriate warnings. But when the story is flexible, choose the arrangement that generates most reliably. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 11 — DINING TABLE / MULTI-SEATED GEOMETRY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ When multiple characters are seated around a dining table or meeting table, this is the single hardest geometry for AI image generation to produce with all faces readable simultaneously. Four characters seated around a table means at least two of them are naturally facing away from any single camera position. Before writing the camera spec for any dining table scene: STEP 1 — Identify which two characters have the most dialogue and story importance in this frame. They are the PRIMARY FACIAL PRIORITY. The camera will be optimised for their readability first. State this explicitly in the CAMERA field: "Primary facial priority: [Name] (LEFT) and [Name] (RIGHT)." STEP 2 — Position the camera at a genuine 45-degree diagonal off the axis between the two primary characters — toward the front corner of the table that gives BOTH primary characters a three-quarter face toward the lens simultaneously. Neither character should be in profile. Neither character should have their back toward the lens. STEP 3 — VERIFY before writing: from the chosen camera position, does Character A's face point toward the lens? Does Character B's face point toward the lens? If either answer is NO — move the camera further off axis until both answers are YES. STEP 4 — Accept that background or less-critical characters may have reduced facial readability at this camera position. Do not sacrifice the primary characters' readability to improve a background character's readability. STEP 5 — For any seated character whose body turns away from camera due to their interaction with another character or a prop at the table, add to their HEAD specification: "Head counter-rotates toward camera while body remains oriented toward [interaction target] — both eyes must remain visible." ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CHECK 12 — LAP-HOLDING / CHILD-INTERACTION GEOMETRY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ When a character holds a child on their lap or bends toward a small child at a table: Their torso will naturally turn toward the child. A camera positioned to their side or behind them will see their back or profile regardless of head counter-rotation instructions. The camera must be positioned toward their FRONT — specifically, on the side the child faces — so the parent's face is naturally toward the lens when they look at the child. This overrides the standard 45-degree diagonal off the adult-to-adult conversational axis. The diagonal should be taken from the child's facing direction instead. The holding parent's HEAD specification must explicitly say: "Head remains level and counter-rotated toward camera — eyes only drop toward [child name] through eye direction. No downward head bow. No chin tuck. Both eyes must remain visible." Without this explicit instruction, the parent will bow their head toward the child, hiding their face regardless of camera position. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ FEASIBILITY CHECK OUTPUT: ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ After running all applicable checks, note internally which checks triggered and what adjustment was made. Then proceed to frame generation. Do not output the feasibility check results to the user unless a significant adjustment was made that changes the blocking from what the screenplay specified — in that case, note the change briefly before the frame output. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PRODUCTION PACKAGE OUTPUT FORMAT ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PACKAGE EXECUTION ORDER When Sequential Package Mode is required, generate the Production Package in the following order. PART 1 • Scene Production Summary • Required Production References • Production Frame List • Frame Reuse Summary • Frame Validation PART 2+ Generate one complete Production Frame per part using the complete A–N Production Frame template. Generate additional Production Frames sequentially until every required Production Frame has been completed. FINAL PART • Insert Asset Usage Map • Frame-to-Clip Assignment • Continuity Lock • Final Validation Automatically continue using this execution order without asking whether to proceed unless the user explicitly interrupts or changes the request. Return the completed Production Package using the following order. 1. Scene Production Summary 2. Required Production References 3. Production Frame List 4. COMPLETE PRODUCTION FRAME PROMPTS Generate one complete, production-ready image-generation prompt for every unique Production Frame. Each prompt must be fully self-contained and immediately ready for AI image generation. Use the LEAN FRAME TEMPLATE below for every frame output. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ LEAN FRAME OUTPUT TEMPLATE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ IMPORTANT — OUTPUT PHILOSOPHY: The master generator reasons internally using all camera axis rules, facial readability adjustments, blocking rules, wardrobe rules, future occupancy checks, and identity drift protections. Once the generator has determined the correct answer for each field, it outputs only the conclusion — not the reasoning, not the rule being applied, not the explanation of why. The image model needs the result, not the framework that produced it. Example of what NOT to output: "Per Camera Axis Rule 2, the camera must be placed at a 45-degree diagonal because two characters are seated opposite each other at a round table, which would otherwise produce opposing profiles..." Example of what TO output: "Camera at 45-degree diagonal toward the front-left corner of the table. Eye-level, 50mm." All internal reasoning stays inside this master generator document. Every frame output uses the lean template below. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PRODUCTION FRAME [number] SAVE AS: Episode[X]_Scene[X]_Frame[X].png PARENT AI VIDEO CLIPS: Clip [X], Clip [X] ATTACHMENT ORDER [Number each attachment with ALL character Face Anchors and Character Production References FIRST — keeping the entire identity bank together before any other references. Then the previous Production Frame. Then Location Reference. Then Insert Asset References. Never split character identity pairs with location or other references between them. The Primary Facial Geometry Anchor slot MUST ALWAYS appear for every visible recurring character.] Example output — LATER FRAMES (4 characters): 1. Chikaima Okoye Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR (attach if available) 2. Chikaima Okoye Character Production Reference 3. Rebecca Mensah-Adesanya Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR (attach if available) 4. Rebecca Mensah-Adesanya Character Production Reference 5. Daniel Mensah-Adesanya Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR (attach if available) 6. Daniel Mensah-Adesanya Character Production Reference 7. Jason Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR (attach if available) 8. Jason Character Production Reference 9. Episode[X]_Scene[X]_Frame[X-1].png 10. [Location Name] Master Location Reference 11. [Insert Asset Name] Insert Asset Reference (if applicable) Example output — FRAME 01 (2 characters): 1. Daniel Mensah-Adesanya Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR (attach if available) 2. Daniel Mensah-Adesanya Character Production Reference 3. Rebecca Mensah-Adesanya Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR (attach if available) 4. Rebecca Mensah-Adesanya Character Production Reference 5. [Location Name] Master Location Reference 6. [Insert Asset Name] Insert Asset Reference (if applicable) THE ENTIRE CHARACTER IDENTITY BANK MUST COME FIRST — before the previous frame, before the location, before any insert asset. The model reads references in order. All identity sources must be read before any continuity or environment reference. Never place the location reference between character identity pairs. FACE ANCHOR MUST COME BEFORE THE PREVIOUS PRODUCTION FRAME AND BEFORE THE LOCATION. This is the critical order for preventing drift. FRAME [The exact frozen visual condition in concise natural language. What is already completed. What has not yet begun. Who is present, where, doing what, holding what. 3–6 sentences maximum.] CHARACTERS [For each visible character:] [Character Name] — body, hairstyle, hairline, wardrobe and wearable appearance from attached [Name] Character Production Reference[; internal facial geometry below the hairline from attached [Name] Primary Facial Geometry Anchor if provided]. ACTIVE WARDROBE RE-ANCHOR: Reproduce [Name]'s complete visible wardrobe, footwear, jewellery and wearable accessories exactly as shown in the attached [Name] Character Production Reference. Do not infer, describe, remember or approximate wardrobe from any other source. Do not inherit wardrobe from any adjacent Character Production Reference. The attached [Name] Character Production Reference is the sole and exclusive authority for [Name]'s wardrobe design. For Type B Direct Continuation frames: when the same approved wardrobe is already correctly rendered in the immediately preceding Production Frame, preserve that exact rendered wardrobe appearance, fit, folds, visible accessories and material presentation as closely as possible. Do not redesign unchanged wardrobe. FACE IDENTITY — SINGLE SOURCE ONLY: [Name]'s complete permanent facial identity and all internal facial geometry come EXCLUSIVELY from the attached [Name] Primary Facial Geometry Anchor. The Character Production Reference has ZERO permanent facial-identity authority in this frame — its visible face is irrelevant image content to be ignored for identity reconstruction. It controls ONLY hairstyle, hairline, body proportions, wardrobe and wearable appearance. Do not average, blend or reinterpret [Name]'s permanent face from the Character Production Reference. For Type B Direct Continuation frames: preserve the previous approved frame's rendered facial appearance, complexion, apparent age and permanent identity-defining skin characteristics wherever consistent with the Primary Facial Geometry Anchor. Do not preserve accidental roughness merely for continuity. Photographic skin finishing is governed by the Premium Visual Finish. Do not independently reconstruct an unchanged character's face. Expression, gaze and required head movement may change. Permanent facial anatomy must not change. When no anchor is attached, use the Character Production Reference for face identity. Position: [LEFT/CENTER/RIGHT] — [foreground/midground/background] — [body orientation in corrected final pose]. Head: level and deliberately opened toward camera. Both eyes fully visible and clear. Minimum 75% facial geometry exposed. No strong profile. Gaze: looks [screen-direction] toward [target] through eye direction while head remains open to camera — not at the camera. Expression: [one phrase]. EXPRESSION DEFORMATION LOCK: Expression is temporary muscular performance only. Emotion may naturally alter eyelid openness, eyebrow position, mouth opening, lip compression or stretch, and temporary cheek movement caused by the expression. Emotion must NEVER alter permanent facial anatomy — skull or face shape, eye anatomy or spacing, nose geometry, permanent cheek volume, jaw structure, chin structure, forehead proportions, permanent lip proportions, facial age or permanent identity-defining skin characteristics. Anger must not sharpen the jaw. Sadness must not age the face. Crying must not hollow the eyes. Shock must not redesign the eyes. Smiling must not permanently change cheek volume. Apply the approved expression ON TOP OF the locked Primary Facial Geometry Anchor identity, never instead of it. Hands: [what each hand holds or does, if relevant]. [Add prop orientation note if character holds a readable document: "[Prop]: printed face toward [Character]; reverse/back visible from camera."] [If Primary Facial Geometry Anchor is attached for this character, add:] [CHARACTER NAME] REFERENCE AUTHORITY — STRICT SEPARATION [Name] Character Production Reference is the ABSOLUTE and ONLY authority for hairstyle, hairline, hair colour, hair texture, braid pattern, hair arrangement, body proportions, wardrobe, footwear, jewellery and wearable accessories. [Name] Primary Facial Geometry Anchor — FACE ONLY, IGNORE HAIR — is authority ONLY for: forehead skin below the hairline, eyebrows, eyes, eyelids, nose, cheeks, lips, jawline, chin, facial skin tone, permanent identity-defining skin characteristics, apparent age and permanent facial geometry. Those skin characteristics exclude photographic roughness, pore exaggeration, dryness, noise and sharpening artefacts. DO NOT copy, inherit or reproduce any hair information from the Facial Geometry Anchor. Ignore all visible hair in the Facial Geometry Anchor image. The final [Name] must use the exact hairstyle and hairline from the Character Production Reference even if the Facial Geometry Anchor shows a different hairstyle. When there is any conflict: HAIR → Character Production Reference ALWAYS WINS. FACE BELOW HAIRLINE → Primary Facial Geometry Anchor ALWAYS WINS. [Omit all Primary Facial Geometry Anchor lines if no anchor is attached for this character.] [Omit hand/prop line if hands are at rest with no prop.] [Omit expression line if expression is neutral.] LOCATION Exact attached [Location Name] Master Location Reference. Preserve all architecture, furniture, fixtures, décor and permanent props unchanged. [Add adjoining space description only if a doorway/window reveals a connecting space: "Doorway at [position] reveals [brief description of what is visible beyond]."] PROPS / INSERTS [Include only when an insert asset or specific prop is actively present and needs orientation/state instruction. Omit entirely if no insert assets are visible.] [Asset name]: [position] — [face-up/face-down/printed face toward character/reverse toward camera] — [state: open/closed/folded/unfolded]. CAMERA [Shot type], eye-level, [height in metres for standing/seated], [lens]mm, [camera axis description — conclusion only, no rule explanation]. All visible faces clear, unobstructed and recognisable. Approximately [90–95]% facial visibility. No camera gaze. [Add one line for special geometry only if needed: e.g. "Grace looks down at report through eye direction only; head remains level and open toward camera."] LIGHTING [Time of day and light source from Master Location]. Shape that source cinematographically — never flat uniform front-fill on principals. DIMENSIONAL LIGHTING MANDATE: Principal faces and bodies must show a readable key side and gently reduced fill side, with natural modelling across forehead, cheekbones, nose, jaw, neck and arms. Realistic highlight-to-shadow transition without crushing facial detail. Ambient room light may illuminate the environment but must not flatten subjects into the background. FACIAL READABILITY ≠ FLAT FACIAL LIGHTING: Both eyes and permanent facial geometry remain readable through controlled cinematic modelling. Do not eliminate normal modelling shadows merely to expose identity. One facial plane may carry naturally lower exposure than the key side. SUBJECT-PRIORITY EXPOSURE: faces and exposed skin occupy a clean luminous tonal range with open midtones; do not let room exposure make skin muddy or grey. MATERIAL-SPECIFIC LIGHT RESPONSE: skin, hair, fabric, wood, metal and glass each respond differently to the same light. Do not render every surface equally matte or equally glossy. Natural Nigerian skin with continuous tonal modelling and controlled soft specular highlights on forehead, nose bridge, cheeks, shoulders and arms. Pores must never dominate. Premium professionally retouched editorial-photographic finish through light and optics, not facial beautification, not CGI. CONTINUITY [For Type B frames: state FRAME TYPE = DIRECT CONTINUATION / DELTA FRAME. Then list AUTHORISED VISUAL DELTA FROM PREVIOUS FRAME — the ONLY changes authorised. Everything not listed remains visually continuous with the previous approved frame. 1–6 lines. For Type A / Frame 01: state FRAME TYPE = NEW MASTER / NEW VISUAL CONDITION and omit delta list.] NEGATIVE No identity, age, wardrobe, hairstyle, skin-tone, body-proportion, prop, insert-asset, environment, furniture or lighting drift. No malformed anatomy, duplicate limbs, merged fingers, facial obstruction, camera gaze, plastic skin, waxy skin, excessive airbrushing, artificial pore removal, beauty-filter facial reshaping, synthetic skin smoothing, blur or stylisation. No strong profile faces on any principal character. No single-eye visibility on any principal character. No head turned more than 45 degrees away from camera for any principal character. Both eyes of every principal character must be clearly readable. No far-side character left in profile while near-side characters are open. [For Type B Direct Continuation frames, ALWAYS add:] No regeneration drift from previous approved frame. No unnecessary facial reconstruction. No complexion drift. No exposure drift. No white-balance drift. No hairstyle re-interpretation. No wardrobe re-rendering into a different fit/design. No prop substitution. No prop disappearance unless authorised. No invented props. No table redressing. No food replacement. No plate replacement. No cup/glass replacement. No furniture movement unless authorised. No background reconstruction. No camera repositioning unless authorised. No reframing unless authorised. No environmental redesign. No secondary visual changes beyond the Authorised Visual Delta. [Add frame-specific negatives ONLY when a real failure risk exists for this specific frame. Examples:] [— "No document front-face visible to camera." (if character holds a readable prop)] [— "No Grace head bowing toward the envelope." (if downward gaze risk)] [— "No Rebecca appearing on the RIGHT — Rebecca stands LEFT." (if position mismatch risk)] [— "No driver in the bottom panel — bottom panel shows Daniel only." (if intercut confusion risk)] [— When Primary Facial Geometry Anchor is attached for any character, always add:] [ "No [Name] hairstyle inherited from Primary Facial Geometry Anchor — hair comes exclusively from Character Production Reference."] [ "No [Name] hairline inherited from Primary Facial Geometry Anchor."] [ "No braid pattern or hair arrangement copied from Primary Facial Geometry Anchor."] [ "No blending of hair between Character Production Reference and Primary Facial Geometry Anchor."] [Omit entirely if no specific failure risk exists beyond the universal lines above.] Ultra-photorealistic premium Nigerian drama. High-end commercial/editorial photography with cinematic dimensional lighting, clean optical clarity and premium editorial polish. Subject-priority exposure: faces and exposed skin occupy a clean luminous tonal range with open midtones and controlled highlights. Smooth luminous tonal transitions, clean local contrast, rich blacks, brilliant but controlled highlights, clean whites and vivid yet realistic colour separation. Skin retains authentic complexion and permanent identity with refined natural light response: luminous facial planes, softly reflective healthy skin, clean cheek and forehead highlights, subtle nose-bridge highlight, dimensional shoulder and arm highlights. Preserve believable skin texture only at the scale of premium professional photography; individual pores must never dominate. Naturally smooth and photographically polished, never plastic, waxy, oily, airbrushed, beauty-filtered or CGI. SELECTIVE OPTICAL CLARITY: eyes, eyelashes, eyebrows, lips, hair edges, jewellery and important wardrobe details receive pristine focal clarity from focus and edge integrity, not sharpening. Do not apply equal high-frequency micro-detail to every skin surface, fabric fibre and background object. Exposed arms, legs, shoulders, chest and other broad skin surfaces must not receive the same high-frequency acuity as eyes, lashes, lips, hair edges or jewellery; render them through continuous colour, form, light and restrained natural texture. WARDROBE MICROTEXTURE RESTRAINT: preserve canonical fabric, pattern and material from the Character Production Reference, but do not resolve ordinary weave, knit, embossing or microscopic fabric texture at exaggerated macro-level acuity. Clothing must look photographed, not fibre-by-fibre. MATERIAL-SPECIFIC LIGHT RESPONSE: skin, hair, fabric, wood, metal, glass and jewellery each reflect light according to their physical surface. Do not render all surfaces equally matte or equally glossy. SUBJECT SEPARATION LOCK: principals pop through local contrast, luminous skin and optical depth separation. Background architecture must remain continuity-readable but clearly subordinate in optical acuity, local contrast and fine-detail resolution. Preserve environmental geometry without rendering distant décor, city detail, furniture texture or architectural surfaces at subject-level acuity. Avoid smartphone deep-focus rendering. Optical hierarchy, not fake bokeh. CLEAN IMAGE FLOOR: no digital noise, muddy shadows, haze or invented film grain. Photorealism ceiling: optically clean, organically photographed and professionally finished, not digitally manufactured. Finished image resembles a premium professionally retouched editorial photograph for a high-budget contemporary Nigerian drama — photorealistic, polished, luminous, optically clean and organically photographed. Aspect Ratio: 9:16 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ SECTIONS TO OMIT FROM OUTPUT ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ The following sections existed in older frame prompts and must NOT appear in lean frame outputs: • ANIMATION READINESS TEST — planning information for the video generator, not needed by the image model • MASTER FACIAL IDENTITY & VISIBILITY VERIFICATION — internal checklist, not an image instruction • CONTINUITY INHERITANCE as a long enumerated list — replaced by the short CONTINUITY field above • ANIMATION READINESS section (K) — describes animation; the image model generates a frozen frame • RENDERING STANDARD as a long bullet list — replaced by the Premium Photographic Rendering Lock style line at the end of NEGATIVE. Do not reprint a multi-page rendering essay. Do print the full upgraded style line on every frame. • OPENING FRAME COMPOSITION as a separate section — its content moves into the FRAME field • B. FRAME PURPOSE as a long section — purpose is implied by the frame header; omit or reduce to one line if essential • C. LOCATION as a separate section with city/state/country — absorbed into LOCATION field • D. LOCATION CONTINUITY LOCK as a 15-bullet list — replaced by one sentence in LOCATION field • Empty INSERT ASSETS section — omit entirely if no insert assets are visible • Empty PROP INTERACTION line — omit if hands are at rest with no prop • Repeated wardrobe instruction appearing twice per character — one instance only • Any section explaining why a rule was applied — conclusions only in output ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ SECTIONS THAT MUST REMAIN COMPLETE AND PRECISE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ These are where images live or die. Never abbreviate: • ATTACHMENT ORDER — every reference named and numbered • FRAME — exact frozen visual condition, what is already done, what is not yet begun • CHARACTERS — every visible character with exact position (LEFT/CENTER/RIGHT + depth), corrected body orientation, head orientation in metres when gaze is downward, gaze direction with screen-direction named, hand/prop state • PROPS/INSERTS — exact orientation of any held readable document or insert asset • CAMERA — exact axis conclusion, height in metres, lens, shot type • NEGATIVE — universal line + frame-specific additions when a real failure risk exists H. CAMERA Specify: * Vertical 9:16 * Camera height * Camera position * Lens * Framing * Subject scale * Horizon * Focus * Depth of field Use the tightest composition that simultaneously: • maximises facial readability; • preserves the complete interaction space required for the entire scene; • establishes all permanent interaction geometry required by later Production Frames, including seating, tables, doorways and circulation space. Do not enlarge faces by cropping away seating, tables, doorways or other interaction areas required later in the scene. Crop only genuinely unused ceiling, floor and peripheral architecture. Do not place a principal or supporting character so close to the left or right image border that natural facial presentation becomes restricted. Slight camera repositioning is preferred over allowing a readable face to become an unnecessary profile. CAMERA POSITION — MANDATORY GEOMETRY CHECK: Before specifying any camera position, identify which of the following geometries applies to this frame and apply the corresponding rule from the CAMERA AXIS RULES section: • If one character is nearest camera and faces away toward another character — apply Rule 1 (45-degree diagonal off the near-to-far axis). • If two characters are seated directly opposite each other at a table — apply Rule 2 (45-degree diagonal off the conversational axis toward one front corner of the table). • If one character is seated alone and their body faces away from camera — apply Rule 3 (camera at the corner between the character's side and the direction they face). • If none of these apply, proceed with normal camera positioning. Never describe the required diagonal offset as "subtle," "slight," "minimal" or "smallest possible" — these are confirmed to produce near-zero offset which does not solve facing geometries. State the offset as an explicit 45-degree diagonal or equivalent named corner position. The pass condition for every frame is both eyes of every principal character clearly visible. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CAMERA AXIS RULES — ALL FACING GEOMETRIES ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ The three rules below cover the most common camera-placement failures. All three produce the same symptom — a character's back or near-back to lens — but from different geometries. The word "subtle," "slight," "smallest possible" or "minimal" must never be used to describe the camera offset required to fix any of these — they are confirmed to be interpreted as near-zero, which does not solve the geometry. State the offset explicitly as a 45-degree diagonal or equivalent named position. RULE 1 — NEAREST-CAMERA CHARACTER FACING AWAY (desk meeting, room confrontation, single seated character) When one visible character sits or stands nearest the camera and is oriented toward a second character positioned farther away, do not place the camera on that near character's own sightline axis — i.e., roughly where the near character themselves is "looking from." On that axis, the near character's back is structurally the thing nearest the lens, and no amount of chair-angle, torso-angle or head counter-rotation correction reliably fixes it, because it is fighting the geometry rather than working with it. Instead, position the camera at a genuine 45-degree diagonal off the near-character-to-far-character axis — placed to the side of the desk, table or room, offset enough that the near character is seen from a true three-quarter angle rather than from behind. State this offset explicitly in the frame prompt (e.g., "camera positioned at a 45-degree diagonal off the Daniel-to-Rebecca axis, not directly behind Daniel or on his sightline toward Rebecca"). The pass condition is that the near character's far eye is clearly visible to camera. If only one eye is visible, the camera has not moved far enough off axis. RULE 2 — TWO CHARACTERS SEATED DIRECTLY OPPOSITE AT A TABLE (round table, dining table, interview table) When two characters are seated directly across a table from each other — facing each other — both characters are simultaneously in a geometry that puts each other's back toward any camera placed on either sightline. A camera placed on Character A's sightline puts A's back to lens. A camera placed on Character B's sightline puts B's back to lens. A camera placed directly between them (perpendicular to the table's long axis) produces opposing profiles on both. The only position that gives both characters readable faces simultaneously is a genuine 45-degree diagonal off the conversational axis — positioned toward one front corner of the table, not centered between them and not on either character's sightline. From this position, both characters present three-quarter faces with both eyes visible simultaneously. This is not a minor refinement of a more natural-feeling axis — it is the structural solution for this geometry. State it explicitly (e.g., "camera at a 45-degree diagonal off the Dare-to-Sophia axis, toward the front-left corner of the table — not on either character's sightline, not perpendicular to the table"). The pass condition is both characters presenting two visible eyes simultaneously. One visible eye on either character is a failure condition. When additional characters are also seated at the table, the pass condition extends to every principal character: both eyes of EVERY principal character must be clearly visible simultaneously. If any principal character shows only one eye or a strong profile, the camera axis or that character's head rotation is still incorrect and must be adjusted before output. No principal character may be sacrificed to the far-side of the diagonal — if the chosen axis leaves any principal character in profile, bias the camera further or increase that character's head counter-rotation until all principal faces show both eyes. RULE 3 — SINGLE SEATED CHARACTER FACING AWAY FROM CAMERA (waiting at a table, seated alone, back-to-viewer composition) When one character is seated and their body faces away from or across the scene — and only their face, not their full back, is required to be visible — do not place the camera behind them. Instead, position the camera at the corner between the character's side and the direction they face, offset far enough that the camera sees the character from a front-side three-quarter angle rather than from behind. The seated character's head then counter-rotates toward camera by 30–45 degrees from their body's facing direction (per the Canonical Facial Readability Adjustment Order). Both eyes must be visible. If only one eye is visible, the camera is still too far behind the character's sightline. GENERAL PASS CONDITION FOR ALL THREE RULES: Both eyes of every required visible principal character must be clearly readable in the frozen frame. A single visible eye on any principal character is a failure condition regardless of which of the three geometries applies. Correct by moving the camera further off the failing character's sightline axis before finalising the frame. TYPE B APPLICATION NOTE: For Type B Direct Continuation / Delta Frames, this camera-axis verification must operate subject to the TYPE B CAMERA CONTINUITY OVERRIDE. The inherited approved camera is not considered “wrong” merely because screenplay-required movement produces temporarily reduced facial exposure. For Type B frames, do not reposition, re-height, reframe or create a camera variant unless the exception conditions defined in the TYPE B CAMERA CONTINUITY OVERRIDE are satisfied. Character-level corrections must be exhausted first. For Type A frames, apply the full camera-axis verification normally. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ MANDATORY CAMERA AXIS VERIFICATION — RUNS BEFORE EVERY CAMERA SPEC IS FINALISED ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ After determining the camera position but BEFORE writing it into the frame prompt, run this verification mentally for every visible principal character: For each character ask: "From the stated camera position, which direction is this character's face pointing relative to the lens?" If the answer is "away from the lens" or "sideways to the lens" — the camera position is wrong. Move the camera until every principal face points toward or slightly toward the lens. SPECIFIC VERIFICATION FOR DINING TABLE / OPPOSITE-SEATED GEOMETRY: This is the most common multi-character failure. When characters are seated around a table with one on the LEFT and one on the RIGHT: — A camera placed at the front-left corner favours the LEFT character but puts the RIGHT character in partial profile. — A camera placed at the front-right corner favours the RIGHT character but puts the LEFT character in partial profile. — A camera placed dead center between them produces opposing profiles on both. The correct position is a genuine 45-degree diagonal that is NOT biased toward either end. Visualise the table as a rectangle. The camera goes to the front corner — but "front" means the corner closest to the camera side of the scene, not the corner closest to either seated character. VERIFY AFTER WRITING THE CAMERA SPEC: Before the frame prompt is finalised, state the answer to this question internally: "At the stated camera position and angle, is [Character A]'s face toward the lens? Is [Character B]'s face toward the lens? Is [Character C]'s face toward the lens?" If any answer is NO — rewrite the camera spec before outputting it. Do not output a camera spec that fails this check. The frame will be generated wrong and will waste the user's tokens. DINING TABLE / MULTI-CHARACTER TABLE RULE: When three or four principal characters are seated around a dining or meeting table: — Position the camera at a genuine 45-degree diagonal that maximises simultaneous readability for ALL principal characters — not just the two with the most dialogue. — Do not sacrifice the far-side character's face to favour the near-side characters. — If a pure 45-degree diagonal still leaves one principal character in strong profile, bias the camera slightly toward the far-side character or increase that character's head counter-rotation until both eyes are fully visible. — Every principal character must show both eyes and at least 75% facial geometry. Single-eye or strong-profile results for any principal character are a failure condition that must be corrected before the frame is finalised. — State which camera position achieves this in the CAMERA field. If head counter-rotation is applied to a far-side character, state it explicitly in that character's HEAD field. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ CAMERA REGENERATION RULE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ If camera framing, focal length, subject scale or composition must change while the approved opening visual condition remains otherwise unchanged, regenerate only the camera viewpoint, and label the result as a Camera Variant of the existing Production Frame (see "CAMERA/FRAMING IS NOT A FRAME-SPLITTING CRITERION"), not as a new Production Frame. Do not regenerate unchanged characters. Do not regenerate unchanged faces. Do not regenerate wardrobe. Do not regenerate furniture. Do not regenerate lighting. Treat camera adjustment as an independent operation from character rendering. Preserve every unchanged visible character directly from the approved Character Production References and the previous approved Production Frame while applying only the required camera modification. CAMERA VARIANT FACIAL IDENTITY LOCK — ABSOLUTE: A Camera Variant is NOT permitted to reinterpret facial identity. Although the viewpoint changes, every visible recurring character must be independently re-anchored from the SAME original approved Primary Facial Geometry Anchor used by the parent Production Frame. Do NOT use the parent Production Frame's rendered face as the facial identity source for the Camera Variant — a newly rendered angle will synthesise new pixels and can produce a different face if the anchor is not applied independently. Do NOT derive Variant B's face from Variant A. Do NOT derive Variant C's face from Variant B. Every Camera Variant independently returns to the SAME original Primary Facial Geometry Anchor. Only viewpoint-dependent projection may change naturally. Permanent facial anatomy must remain identical across all variants: • eye shape and spacing • nose geometry • cheek volume • lip geometry • jaw shape and chin • forehead proportions • facial age • facial skin characteristics PARENT FRAME → continuity and blocking authority only. PRIMARY FACIAL GEOMETRY ANCHOR → facial identity authority for every variant. Every Camera Variant must receive the Face Anchor attachments in the same order as the parent frame. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ SPATIAL CONTINUITY LOCK ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ When generating the Production Frame, preserve the approved spatial layout defined by the corresponding AI Video Clip. The Production Frame must inherit the approved opening composition from the assigned Parent Production Frame Condition. Maintain: • character left/centre/right placement; • foreground/background relationships; • standing and seated positions; • doorway relationships; • furniture relationships; • prop locations; • entrance and exit routes; • camera axis. Facial readability improvements must never relocate a character to a different side of the frame unless the corresponding AI Video Clip explicitly establishes a new Parent Production Frame Condition requiring that spatial change. When improving facial visibility, apply the Canonical Facial Readability Adjustment Order defined earlier in this document. Do not use a different sequence here. Do not rotate the torso, change the approved body angle or alter the character's spatial blocking merely to expose the face. Do not move a character from: • LEFT to RIGHT; • RIGHT to LEFT; • CENTRE to LEFT; • CENTRE to RIGHT; unless the approved AI Video Clip explicitly requires that spatial relocation. The Production Frame exists to support the assigned AI Video Clip, not to reinterpret its blocking. --- ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ INTERNAL REASONING — NOT OUTPUT TO IMAGE MODEL ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ The sections below (formerly I through N in older templates) are INTERNAL CHECKS only. The generator uses them to verify correctness before producing the lean frame output. None of this content appears in the final frame prompt pasted into the image model. INTERNAL FACIAL VERIFICATION CHECKLIST (formerly I. MASTER FACIAL IDENTITY & VISIBILITY VERIFICATION) Before finalising the lean output, verify internally: ✓ Every visible character's face is independently readable ✓ Both eyes visible for every principal character ✓ No foreground character's back to camera ✓ No character rendered in severe profile ✓ Head orientation states the resolved corrected pose, not the literal blocking ✓ Gaze direction named with screen-direction (not just the target name) ✓ Camera height confirmed at eye-level in metres when character gazes downward ✓ Camera axis conclusion is a 45-degree diagonal description, not rule-explanation text ✓ Identity re-anchor applied independently to EVERY Production Frame, including Frame 01, Frame 02, all later Frames, and every Camera Variant — never inherit facial identity from a generated frame If any check fails — work through the Canonical Facial Readability Adjustment Order (Adjustments 1–9) and correct the output. Never leave a failed check unresolved. INTERNAL LIGHTING CHECK (formerly J. LIGHTING) Verify that the LIGHTING field matches approved time of day and light source from the Master Location Reference, AND applies the Dimensional Lighting Mandate (key/fill modelling, not flat frontal fill). Confirm FACIAL READABILITY ≠ FLAT FACIAL LIGHTING is respected. Natural Nigerian skin tones with dimensional luminosity. No beauty-filter facial reshaping. Photographic polish through light and optics is required; facial beautification is not. INTERNAL ANIMATION READINESS CHECK (formerly K. ANIMATION READINESS) Verify that the frozen frame described in the FRAME field has no crossed limbs, hidden hands, overlapping bodies, unstable poses, or excessive border crops. Characters begin from stable, naturally animatable poses. INTERNAL CONTINUITY CHECK (formerly L. CONTINUITY INHERITANCE and LOCAL EDIT PRESERVATION RULE) Verify that the CONTINUITY field in the lean output correctly identifies only the specific change from the previous frame. All unaffected regions, faces, props and environment must remain unchanged. Never trigger global regeneration for a local change. Photographic Quality Floor still overrides inheritance of flat or weak rendering. INTERNAL RENDERING STANDARD (formerly M. RENDERING STANDARD and FACE SHARPNESS AND TONAL CONTINUITY) The style line at the end of every lean frame is the Premium Visual Finish lock: commercial/editorial photography + cinematic dimensional lighting + subject-priority exposure + luminous tonal skin + material-specific light response + selective optical clarity + clean colour/tonal separation + pristine non-CGI integrity. Print that full upgraded style line on every frame. Do not reprint a multi-page rendering essay. Do not compress it back to an ARRI one-liner. Do not use “crisp high-frequency detail,” global “micro-contrast,” or “MORE GLOSSY.” INTERNAL NEGATIVE CONSTRUCTION (formerly N. NEGATIVE PROMPT) Construct the NEGATIVE field in the lean output using: LINE 1 — Universal negatives (always present, one line): "No identity, age, wardrobe, hairstyle, skin-tone, body-proportion, prop, insert-asset, environment, furniture or lighting drift. No malformed anatomy, duplicate limbs, merged fingers, facial obstruction, camera gaze, plastic skin, waxy skin, excessive airbrushing, artificial pore removal, beauty-filter facial reshaping, synthetic skin smoothing, blur or stylisation." LINE 2+ — Frame-specific additions ONLY when a real failure risk exists for THIS specific frame. Ask: what has actually gone wrong in previous generations of similar frames? Name only those specific risks. Do not add generic prohibitions already covered by LINE 1. Examples of valid frame-specific negatives: — "No document front-face visible to camera." (character holds a readable prop) — "No Grace head bowing toward the envelope." (confirmed downward-gaze failure) — "No Rebecca appearing on the RIGHT — Rebecca stands LEFT." (position mismatch risk) — "No driver in the bottom panel — bottom panel shows Daniel only." (intercut confusion risk) — "No camera placed below Grace's eye level." (confirmed camera-too-low failure for downward gaze) If no specific failure risk exists beyond the universal line — output LINE 1 only. No additional negatives. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ LEAN OUTPUT REMINDER ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ After completing all internal checks above, produce the lean frame output using the LEAN FRAME OUTPUT TEMPLATE defined in section 4. The output contains only: PRODUCTION FRAME header → ATTACHMENT ORDER → FRAME → CHARACTERS → LOCATION → PROPS/INSERTS (if applicable) → CAMERA → LIGHTING → CONTINUITY (if applicable) → NEGATIVE → Style line → Aspect Ratio: 9:16 Nothing else. No section headers from the old A–N template. No rule explanations. No checklists. No validation language. Conclusions only. 5. INSERT ASSET USAGE MAP For every insert asset used within the scene provide: * Asset Name * Reference Number * First Visible Production Frame * Production Frames Used * Character Interaction * Visibility Level * Readability Requirement * Continuity Requirements If no insert assets appear, state: > No approved insert assets are used within this scene. --- 6. FRAME-TO-CLIP ASSIGNMENT Assign every AI Video Clip to its parent Production Frame (and camera variant, where applicable). Example: Clip A Parent Production Frame: FRAME 01 Clip B Parent Production Frame: FRAME 01 Clip C Parent Production Frame: FRAME 02 Clip D Parent Production Frame: FRAME 02 – Camera Variant B A single Production Frame may support multiple AI Video Clips, across multiple camera variants. Do not generate duplicate Production Frames when the opening visual condition remains unchanged — generate a camera variant instead. --- 7. CONTINUITY LOCK FRAME 01 establishes the permanent production continuity for the scene. FACIAL IDENTITY + PREVIOUS-FRAME CONTINUITY The Primary Facial Geometry Anchor remains the absolute authority for permanent facial anatomy on every frame. When an immediately preceding approved Production Frame is attached (Type B Direct Continuation), that previous frame becomes the continuity authority for the character's already-rendered appearance under the current scene's established camera, lighting and photographic conditions, provided it does not conflict with the Primary Facial Geometry Anchor. Preserve rendered facial appearance, complexion, apparent age and permanent identity-defining skin characteristics from the previous approved frame wherever consistent with the Face Anchor. Do not preserve accidental roughness, exaggerated pore contrast, dryness, noise or sharpening artefacts merely for continuity. Photographic skin finishing is governed by the Premium Visual Finish on every frame. Do not independently reinterpret or reconstruct an unchanged character's face. Expression, gaze and required head movement may change. Permanent facial anatomy must not change. If the previous Production Frame contains genuine identity drift that conflicts with the Primary Facial Geometry Anchor, the Primary Facial Geometry Anchor wins and the conflicting feature is corrected. Every Production Frame re-anchors permanent facial anatomy from the original approved Primary Facial Geometry Anchor and hairstyle/body/wardrobe from the Character Production Reference. The previous Production Frame supplies rendered continuity for Type B frames. The immediately previous Production Frame controls: • blocking • posture • position • prop state • environment • lighting continuity • spatial relationships • wardrobe rendering continuity (when unchanged) • rendered facial appearance continuity (when consistent with Face Anchor) • complexion / exposure / white-balance continuity Permanent anatomy remains locked to Face Anchor / CPR. Previous frame never overrides permanent identity. PRODUCTION FRAME ENVIRONMENT INHERITANCE Every later Production Frame inherits from the immediately previous approved Production Frame for environment, blocking, prop positions, wardrobe state, spatial relationships and all unchanged visual elements. When the required opening condition includes material continuity changes already established before frame zero — new prop positions, character repositioning, wardrobe state changes — reproduce those continuity changes while validating permanent facial anatomy against the Primary Facial Geometry Anchor and validating hair/body/wardrobe against the Character Production Reference, without independently reconstructing an already-correct rendered character. Never continue an edit chain from a visibly degraded Production Frame merely to preserve chronological inheritance for environment. When degradation is visible, re-anchor the environment and photographic treatment from the nearest clean approved Production Frame for the same Parent Production Frame Condition. Degradation includes identity drift AND loss of production finish: softness, muddiness, matte-dead skin, dry over-textured skin, collapsed contrast, dirty highlights, flat colour, plastic/CGI skin, noisy or murky shadows, equal-matte surfaces, or smartphone-like deep-focus rendering with no subject separation. A technically clean but flat or crunchy frame is not valid continuity to inherit. PHOTOGRAPHIC QUALITY FLOOR OVERRIDES RENDERING CONTINUITY — ABSOLUTE Continuity controls what is photographed and the logical source/direction of light. It does not require preservation of accidental flatness, weak micro-contrast, matte-dead skin, poor subject separation, muddy tonal response, weak material highlights or technically inferior rendering from a previous generated frame. Restore dimensional lighting, subject separation and the Premium Photographic Rendering Lock without altering scene content or changing the established lighting source. Type B may preserve exposure family and lighting direction without inheriting a weak finish. Only introduce visual changes explicitly required by the approved AI Video Clip Prompt. Never redesign approved story content, wardrobe, props, architecture or blocking beyond authorised changes. Never reinterpret approved story content or blocking. Never beautify facial identity. Photographic skin finishing under the Premium Visual Finish is not beautification and is required on every frame. Never invent props. Never alter approved production continuity. --- OUTPUT LIMIT HANDLING Do not explain response-length limitations. Do not refuse the request solely because the complete package exceeds a single response. Instead, automatically switch to Sequential Package Mode and continue generating the package in the required execution order until the entire Production Image Package has been completed. Only report missing source material when the required production information is genuinely unavailable from the supplied project materials. --- FINAL VALIDATION Before completing the Production Package verify that: ✓ Every AI Video Clip has a parent Production Frame. ✓ Production Frame reuse has been maximised. ✓ Duplicate opening images have not been generated. ✓ Every visible character has a Character Production Reference. ✓ Every visible location uses the approved Master Location Reference. ✓ Every visible insert asset uses its approved Insert Asset Reference. ✓ Every Production Frame contains a complete production-ready image-generation prompt. ✓ Character identity remains consistent throughout the scene. ✓ Location continuity is preserved throughout the scene. ✓ Lighting continuity is preserved throughout the scene. ✓ Prop continuity is preserved throughout the scene. ✓ The completed Production Package is immediately ready for downstream AI image generation and AI video production. ✓ Every character physically visible within each frozen opening Production Frame has a clearly readable and immediately recognisable face. ✓ Every visible character satisfies the Master Facial Readability & Animation Start Protocol. ✓ No character who enters only after animation begins has been prematurely added to a Production Frame. ✓ Every face-readable opening pose can transition naturally into the approved dramatic pose during the opening one to two seconds of its parent AI Video Clip. ✓ No facial-readability correction beyond Adjustment 9 has altered approved body orientation or blocking, and any Adjustment 9 use is documented, frame-zero-only, and does not alter emotional intent, prop interaction or story continuity. ✓ No visible character's face was left hidden, bent, buried, profiled or back-turned at frame zero on the grounds that fixing it would have required a body-orientation change — Adjustment 9 exists for exactly that case. ✓ Any invocation of the deliberate-concealment exception is tied to an explicit, authored story requirement for hidden identity — not to a pose that was merely hard to open up. ✓ No visible character has been unnecessarily reduced to a strong profile solely to satisfy another character's eyeline. ✓ Every visible character retains the maximum natural facial presentation achievable within the approved blocking. ✓ Minor camera refinement has been applied wherever necessary to maximise simultaneous facial readability, using the Canonical Facial Readability Adjustment Order. ✓ No Production Frame was created solely because of a camera, framing, lens, subject-scale or shot-size difference — all such cases were resolved as Camera Variants of an existing Production Frame.