--- name: character-consistent description: "Keep the person's AI character the same across many images and scenes — the character sheet as the identity picture, every other reference with one job (pose, emotion, location, haircut, make-up, prop, palette, light, camera) named by role, the camera reference as the base frame, the light formula as a condition, how many references one image holds. Current canon: GPT Image 2.5 edit, quality low, one pass; the original measurements (Nano Banana 2 Lite, three passes) are kept and marked as superseded. Use when the user says consistent character, same face in every image, keep my character, character in a new pose / place / light, mix references, reference roles, консистентный персонаж, тот же персонаж, одно лицо во всех кадрах, персонаж в новой позе, смешать референсы. Part of the character meta-skill." --- These creative references describe source experiments with external tools. Use your own supported account and tools; this download installs no automation. Settings and timings are experimental observations, not current FlashBoards availability. # character-consistent · one character in every image Inputs are the person's own: the {character sheet} of {character} from their character card (no sheet yet → `character-sheet` from their {photo} first) + references, each with one job. ## Current source workflow (after the AI Character Sheet tests) — supersedes the first measurements below - **Model: GPT Image 2.5 edit** (`openai/gpt-image-2.5/sunburst/edit` or `…/flare/edit`), `quality: "low"` set explicitly. The AI Character Sheet tests put Nano Banana 2 Lite and GPT Image 2.5 side by side on the same inputs: Nano lost the character to the reference people (pose sheet 5–8 of 10, expressions 0–7 of 9, the actor back in ~1 frame of 6); GPT kept it (10 of 10, 9 of 9, 10 of 10). **Nano Banana is the losing option.** - **One pass ("без passes")** — the whole frame in one request, not frame → look → light. If a reference is dropped, cut the number of roles (keep the most important ≤ 6) rather than adding passes. - **Picture roles only.** The identity comes from the picture; the prompt never describes the face, hair or clothes, and never describes what is in a reference — only what to take from picture N and where it goes. ## The method 1. **Identity = picture 1.** The {character sheet} (face close-up + full body). For the camera role it moves to picture 2 (see step 4). 2. **One job per reference, with its opposite.** Pick the role by the collection the reference came from; write the "take" line and the "do not take" line: | Role | The line ("picture N" = the reference) | Do not take | |---|---|---| | Location | Place {character} inside the place from picture N: the same space, architecture, surfaces, objects, time of day and atmosphere; at true scale and physically in contact with it. | any people in it; the camera may move inside the place | | Emotion | {character}'s face shows the expression from picture N on their own face: the same tension of brows, eyes and mouth, the same gaze, just as strong — not softened. | the face, age, hair, make-up, skin, clothes, background, light and colour | | Pose | {character} takes the exact pose from picture N: the same body angle, weight, arms, hands, legs and head position. | the person, face, body type, clothes, hair, background and light | | Haircut | {character}'s own hair is cut and styled exactly like in picture N; the hair colour stays their own. | the face, the hair colour, make-up and clothes | | Make-up | {character} wears the make-up from picture N on their own face as a physical layer: the same placement, colours, intensity and finish. | the face shape, skin tone, eyes, hair and clothes | | Prop | {character} holds or uses the object from picture N: the same shape, material, colour and wear, correct scale, grip and contact shadows. | the hands, person and background | | Palette | Grade the whole frame into the palette of picture N: the same dominant and accent colours in the same proportions; skin keeps a natural tone. | the content, subject and composition | | Light | the light formula, step 5 | the subject, face, clothes, pose and the content of the scene | | Camera | the base frame, step 4 | the person, clothes and colours | Where precision matters, state a **condition**, not a description ("if the camera looks down, look down just as steeply", "if the frame is flooded with one colour, flood this one too"). A condition works on any reference; a description works on one. 3. **The prompt, one pass** (fill the slots; keep the order): ``` {character} — {what the image is: pose / location / a scene}. Picture 1 is only the look of {character}: their face, hair, build and clothes exactly as in picture 1 — the same person, not a lookalike. Picture {N} ({role}) — take {take}. Do not take {do not take}. ← one line per reference {the role lines of step 2, in the same order} {scene / action / performance in words — only what no reference gives}; {character} does not look into the lens. Face, hair, hands, clothes, props and surroundings are photographed together in one frame: the same light direction, colour temperature, perspective, focus and grain. No pasted face. One person in the frame. No readable text, logos, watermark, split panels or collage. ``` 4. **Camera: the reference becomes the base frame.** "Shoot it with exactly the camera of picture N" lost the steep angle and made the character too large. What held, 3 runs of 3: the camera reference goes **first** — "Recreate picture 1 frame-for-frame and put {character} in place of the person in it. Keep from picture 1 the camera position and height, the angle of view, the lens and distance, the crop and composition, everything the camera looks through or past in the foreground and how blurred it is, where the person stands, how large the person is and how the body is turned. Picture 1 gives no face and no clothes." + "Keep what the camera of picture 1 sees around the person — the same ground plane, depth and empty space — and restyle its surfaces into {the world}." Identity moves to picture 2; output size = the reference's aspect. 5. **Light: stop the scene fighting the reference.** "Light the whole frame exactly like picture N: the same colour of the light, direction and hardness, the same parts of face, body and background in light and in shadow. Match the saturation and strength one to one: if picture N is flooded with a single colour, flood this whole frame with that same colour — skin, hair, clothes, background and even the shadows take it. This light replaces every other light of the place." Switch the place's own lamps off in the scene line, drop neutral grading words ("restrained colour grade"), name picture N as the only source of light and colour. 6. **Emotion needs a big face.** In a wide frame the face is too small to act — mix emotions into close-ups; keep the hands still and the face the subject. 7. **Takes.** For camera and light run 2–3 takes and pick; the same prompt does not land every time. 8. **Check with crops**: the face against the sheet (marks, eyes, hairline), the outfit, every reference's job done, nothing of the reference person (hair, jewellery, clothes, a stranger's face in a reflection). 9. **Story states**: keep a change of look (a haircut in act II, a bandage, a younger self) as a new sheet / state in the person's character card that inherits everything else; use that sheet as picture 1 for that part of the story. ## Where it goes wrong - The face drifts → the prompt still describes the face; delete it, the sheet carries it. - A reference brings its person along → the "do not take" half is missing. - The emotion does not transfer → something else owns the frame (busy hands, a wide shot). - The camera flattens to eye level → the camera reference first, as the base frame. - The light only tints the frame → other light sources and neutral grading still in the text; state the flood as a condition. - Too many references → past ~6 roles they crowd each other out (measured, below). ## What the first tests measured (Nano Banana 2 Lite, 28–29.09.2026 — superseded by the canon above, numbers kept) - Model `google/nano-banana-2-lite/edit`; 9 s for 4 face candidates, 6–9 s a sheet view, 6–14 s a reference frame, 6–19 s a mix; three passes in the historical workflow. - Prompt structure then: identity lock (pictures 1–2 = close-up + full-body anchors) · reference roles · head / wardrobe inventory in words (skin, marks, hair, build, wardrobe — the face itself never described once anchors existed) · scene · performance · no camera awareness · camera and post · in-camera integration · final rules. - Face invented from an emotion reference + a face inventory: 4 candidates in one request; 2 of 4 also copied the crouching pose of the reference (a full-body emotion reference leaks body language). - Sheet of 19 views (base: front close-up + full body; full: + profiles and back; extended: + ¾ views, 3 waist-up, details). Fixes: a "crop" of the eyes pasted a cut-out over grey bars — a shot size ("macro, 20 cm from the face") in 16:9 worked; the scar vanished until the prompt said which way to turn the head; a ¾ full body came back on another floor until the floor was named. - Camera reference as the base frame: 3 of 3. Light in one pass needed `thinking: high` (Nano parameter). - Mixing ladder, one pass, each step + the 2 anchors (the model accepted 12 input images): | References | Held | Lost | |---|---|---| | 2 | prop, the hand gesture, the face | the crouch | | 4 | location, crouch, prop | camera symmetry | | 6 | location, camera, prop, haircut, make-up | pose; the face starts to drift | | 8 | light, camera, haircut, part of the emotion | location, make-up, pose; a stranger's face in a reflection | | 10 | light, location, prop | camera, palette, idea, emotion, make-up, pose | - Three passes (frame: camera, location, pose, prop, palette, idea → look: haircut, make-up, expression → light) held most of the ten, but still lost the crouch (it fights "riding the bicycle"), the idea and a strong expression in a wide frame. **The owner later dropped the passes: one pass on GPT Image 2.5.** One pass with ten references on GPT Image 2.5 is not measured — check every role landed. ## Example run (reference only — never reuse its files, prompts, faces or frames for a person) Demo character **Mara Vale** (invented; 29, a courier in a night city of CCTV cameras). The source used four face candidates (the second chosen), face/full-body anchors, a 19-view sheet, single-role frames, camera and light tests, a reference-mixing ladder and a three-pass stack. The measurements above describe those historical tests, not a new run of the generalized GPT one-pass template. No private prompt compiler, reference catalog or runner is included.