Recurring AI characters for a series: one portrait, one 3-view sheet, three new heads

GPT Image 2.5, Seedream 5.0 Pro and Nano Banana 2 Lite prompt on Twistavo, an AI video studio for character swap: replace a person in a real video from 1 to 4 photos; plans from $3.99/month.

Updated

A series needs characters people recognize in the first second of every post. We built ours on September 30, 2026 in three passes. First, one waist-up phone-style portrait per character, text only, on GPT Image 2.5. Second, a 3-view sheet (front, back, close portrait) made from that portrait on Seedream 5.0 Pro, which becomes the reference for every later image and video. Third, new heads on the same suited body with Nano Banana 2 Lite, to try other cast members without writing a body again. The two characters kept for the series are called CLAUDE and BRIGITTE; the prompts still carry their working names. All seven results are below with the exact prompt, the settings and the cost in Twistavo today.

Make your own with Twistavo — 1 free image

Setup

Model
GPT Image 2.5, Seedream 5.0 Pro and Nano Banana 2 Lite
Settings
Order matters. The portrait fixes the face, the hair and the outfit; nothing else in the chain is text-only. The sheet prompt is a long reusable template (about 4,000 characters) that says to copy the person from the attached image and adds a single line at the end to complete the outfit below the waist, since a waist-up portrait never shows trousers or shoes. Only that last line changes between CLAUDE and BRIGITTE. The head prompts are a short edit template: keep the body, the suit, the pose and the light, replace only the head and hair. In Twistavo: open the AI Studio, generate the portrait from text, then pick the sheet model, upload the portrait as the reference image and paste the sheet prompt. For new heads, upload the sheet and paste the head prompt. Each character costs one portrait and one sheet; each head test costs one image.
Models used
GPT Image 2, Seedream 5.0, Nano Banana

1. CLAUDE is born: a phone-style portrait outside a bakery (GPT Image 2.5)

AI portrait made with GPT Image 2.5: a lanky freckled man with two tall ginger hair columns, small round gold glasses and an olive velvet three-piece suit stands outside a Paris bakery
  • Made September 30, 2026
  • Text to image, no reference
  • Aspect ratio 9:16
  • Quality high
  • Resolution 2k
  • Variant sunburst

20 credits in Twistavo

Prompt

Create ONE original fictional recurring social-media character, photographed on a normal city pavement outside a mundane neighborhood bakery. Waist-up vertical casual phone photo, eye level, looking directly at camera. Character: Lucien, a lanky French man aged 45 with a long narrow face, prominent ears, pale freckled skin and tiny round gold spectacles worn low on his nose. His signature is a spectacular authentic ginger hairstyle shaped into two tall vertical symmetrical rounded columns above each temple, smooth shaved central part, and a thin straight pencil moustache. Hair is visibly made of real hair, meticulously groomed, physically plausible but outrageously unfashionable. He wears a fitted olive-green velvet three-piece suit, a cream shirt with enormous pointed 1970s collar, burgundy necktie. Perfectly serious lips, mild self-important eye contact. Hands relaxed, no held props. The funny unsettling effect must come entirely from his unforgettable silhouette, very specific face and completely sincere formal bearing in ordinary life. Human anatomy and real textured skin, believable wrinkles and velvet nap, natural cloudy daylight, slightly imperfect amateur phone framing. Feels like an unexpected real person found on the feed, not a fashion campaign, not a horror character or cartoon. No typography, no logos, no fantasy anomalies. Do not use a blond pageboy haircut.

Tip: The prompt describes one silhouette that reads at thumbnail size (two hair columns, tiny glasses, velvet suit) and asks for an ordinary phone photo, not a fashion shot. The hair columns, the pointed collar and the velvet came out as written. Two details did not: the pencil moustache came out thick, and the glasses are not worn low on the nose. The last line bans a blond pageboy cut: when you build a cast, write down the looks already taken so two characters never share a silhouette.

2. BRIGITTE is born: the same recipe in a cafe (GPT Image 2.5)

AI portrait made with GPT Image 2.5: a tall red-haired woman with one large round hair coil, cobalt round glasses, a scalloped cream collar and a cobalt jacket sits in a Paris cafe
  • Made September 30, 2026
  • Text to image, no reference
  • Aspect ratio 9:16
  • Quality high
  • Resolution 2k
  • Variant sunburst

20 credits in Twistavo

Prompt

ONE photoreal portrait of an entirely original fictional eccentric social-media character, waist-up vertical casual phone snapshot in an ordinary French cafe. Mireille, a very tall slender woman aged 48 with a long narrow face, prominent cheekbones, wide-set hazel eyes, sparse straight brows, long nose and tightly pursed lipstick-red lips. Her sculptural copper-red hair is rolled into one enormous perfectly round smooth vertical coil centered high above her head, like an oversized polished cinnamon-roll silhouette made only of real hair, with tiny tightly curled side locks at her ears. Small circular cobalt-blue spectacles. High-neck cream blouse with a large scalloped collar, cobalt-blue fitted jacket and matching blue fabric-covered buttons. Her fixed solemn gaze and formal upright posture make her seem completely unaware of how extraordinary she looks. Empty hands resting naturally at waist. The character herself is the hook: elegant, strange, immediately recognizable, human and slightly awkward rather than glamorous. Mundane cafe furniture behind her, authentic skin texture and flyaway hairs, flat window daylight, ordinary imperfect phone-photo realism. No cinematic grading, no horror, no supernatural reflection, no caption, no logo, no watermark. No blond pageboy hairstyle, no copying existing people.

Tip: Same structure as CLAUDE: name, age, one sculpted hairstyle, glasses, a period outfit, a serious face, a mundane place. The coil, the glasses and the collar came out as written. What the prompt did not control: the model added a chalkboard menu behind her, and the menu came out readable, although the prompt bans captions. Check the background before you post it.

3. From portrait to 3-view sheet (Seedream 5.0 Pro)

AI character sheet made with Seedream 5.0 Pro: the ginger man in the olive velvet suit shown full length from the front and the back and in a three-quarter close portrait on white
  • Made September 30, 2026
  • Image to image, 1 reference image (the portrait above)
  • Aspect ratio 16:9
  • Resolution 2k

22 credits in Twistavo

Prompt

A soft, low-contrast film character sheet with quiet, delicate texture: the entire image lives in a broad, gentle middle register — highlights roll off early into soft cream while keeping full detail, shadows sit lifted, open and airy, micro-contrast stays low and tender, skin and fabric resolve with smooth, quiet, organic detail, edges stay rounded and creamy under a gentle veil of optical diffusion, as through an old uncoated vintage cinema prime, with a soft halation bloom breathing from the brightest points — the whole sheet reading like a delicately printed film photograph with a modern neutral colour grade.

Within that look, and using the attached image strictly as the identity reference for the person — their face, age, build, hair and costume are canon and must match the reference 100% — generate the character sheet: three stills of this same person arranged side by side on one sheet: FULL-BODY FRONT view on the left, FULL-BODY BACK view in the middle, and a THREE-QUARTER portrait on the right. The same person, identical outfit and proportions across all three panels, standing on a seamless PURE WHITE studio background that keeps its own even, luminous brightness, reading as clean paper; full figure in frame head to shoes with clear space above the head and below the feet, a soft contact shadow grounding them. Carry over only the person themselves: they stand alone in clean, empty white studio space, lifted out of the reference's scene and simply relit.

The character, anchored to the reference: exactly the person in the attached image, at their exact age and build, with their exact face carried over one-to-one as photographed — the precise hairline and haircut, the natural asymmetry between the left and right halves of the face, the true set of the eyes, nose and mouth, every crease, mark, shadow and skin texture kept as in the reference, real skin softly resolved — the same lived-in cinematic face, simply relit. They wear their reference costume exactly: every garment, its colours, closures, layers, accessories, folds and wear marks carried over consistent with the reference. Below the reference's frame edge the costume is completed faithfully in the same style, period and quality as the visible garments, down to period-true footwear. They stand NEUTRALLY with empty, relaxed hands: weight even on both feet, arms at their sides, shoulders level, head straight; expression neutral and at rest, mouth closed, face calm, eyes open and readable with soft catchlights.

FRONT panel: facing camera square, the costume's front reading clearly — its closures, collar, accessories and colours exactly as in the reference. BACK panel: turned fully away — the haircut from behind, the costume's back completed faithfully in the same period and construction, the sleeves at their sides, the trouser or hem line, the heels of the footwear. THREE-QUARTER portrait panel: a CHEST-LEVEL close portrait — the frame runs from just above the top of the head down to the upper chest, the bottom edge cutting across the chest, so the head and face fill most of the panel; face turned three-quarters to camera, calm and still, with the reference's exact asymmetry and every facial feature carried over unchanged — hairline, creases, skin and all.

The light is neutral and even across all three panels — a soft ambient wash, as if the whole space sits under a broad overcast sky: directionless, calm, arriving frontal and even, wrapping the figure gently, open transparent shadows, every fabric reading true, the exposure holding the low-contrast filmic register declared above. The figure's outline meets the pure white background softly along every edge. The colour story comes from the reference itself: the costume's dominant colour as the leading mass against the pure white field, its secondary tones as the supporting masses, and the smallest bright detail of the reference costume as the single accent.

Wardrobe completion: complete Claude Frangeçois's outfit with full-length straight olive-green velvet tailored trousers matching the visible jacket, and polished dark-brown leather Oxford shoes. Keep these trousers and shoes identical in the front and back panels.

Tip: The face, the hair columns and the glasses carried over from a single waist-up photo, and the model invented the lower half from the last line of the prompt: velvet trousers and brown Oxford shoes, identical in the front and back views. This sheet is the file we reuse; the portrait is not uploaded again.

4. Same template, second character (Seedream 5.0 Pro)

AI character sheet made with Seedream 5.0 Pro: the red-haired woman with the round hair coil in a cobalt suit with flared trousers and black ankle boots, front, back and close portrait on white
  • Made September 30, 2026
  • Image to image, 1 reference image (the portrait above)
  • Aspect ratio 16:9
  • Resolution 2k

22 credits in Twistavo

Prompt

A soft, low-contrast film character sheet with quiet, delicate texture: the entire image lives in a broad, gentle middle register — highlights roll off early into soft cream while keeping full detail, shadows sit lifted, open and airy, micro-contrast stays low and tender, skin and fabric resolve with smooth, quiet, organic detail, edges stay rounded and creamy under a gentle veil of optical diffusion, as through an old uncoated vintage cinema prime, with a soft halation bloom breathing from the brightest points — the whole sheet reading like a delicately printed film photograph with a modern neutral colour grade.

Within that look, and using the attached image strictly as the identity reference for the person — their face, age, build, hair and costume are canon and must match the reference 100% — generate the character sheet: three stills of this same person arranged side by side on one sheet: FULL-BODY FRONT view on the left, FULL-BODY BACK view in the middle, and a THREE-QUARTER portrait on the right. The same person, identical outfit and proportions across all three panels, standing on a seamless PURE WHITE studio background that keeps its own even, luminous brightness, reading as clean paper; full figure in frame head to shoes with clear space above the head and below the feet, a soft contact shadow grounding them. Carry over only the person themselves: they stand alone in clean, empty white studio space, lifted out of the reference's scene and simply relit.

The character, anchored to the reference: exactly the person in the attached image, at their exact age and build, with their exact face carried over one-to-one as photographed — the precise hairline and haircut, the natural asymmetry between the left and right halves of the face, the true set of the eyes, nose and mouth, every crease, mark, shadow and skin texture kept as in the reference, real skin softly resolved — the same lived-in cinematic face, simply relit. They wear their reference costume exactly: every garment, its colours, closures, layers, accessories, folds and wear marks carried over consistent with the reference. Below the reference's frame edge the costume is completed faithfully in the same style, period and quality as the visible garments, down to period-true footwear. They stand NEUTRALLY with empty, relaxed hands: weight even on both feet, arms at their sides, shoulders level, head straight; expression neutral and at rest, mouth closed, face calm, eyes open and readable with soft catchlights.

FRONT panel: facing camera square, the costume's front reading clearly — its closures, collar, accessories and colours exactly as in the reference. BACK panel: turned fully away — the haircut from behind, the costume's back completed faithfully in the same period and construction, the sleeves at their sides, the trouser or hem line, the heels of the footwear. THREE-QUARTER portrait panel: a CHEST-LEVEL close portrait — the frame runs from just above the top of the head down to the upper chest, the bottom edge cutting across the chest, so the head and face fill most of the panel; face turned three-quarters to camera, calm and still, with the reference's exact asymmetry and every facial feature carried over unchanged — hairline, creases, skin and all.

The light is neutral and even across all three panels — a soft ambient wash, as if the whole space sits under a broad overcast sky: directionless, calm, arriving frontal and even, wrapping the figure gently, open transparent shadows, every fabric reading true, the exposure holding the low-contrast filmic register declared above. The figure's outline meets the pure white background softly along every edge. The colour story comes from the reference itself: the costume's dominant colour as the leading mass against the pure white field, its secondary tones as the supporting masses, and the smallest bright detail of the reference costume as the single accent.

Wardrobe completion: complete Mireille Bigoudi's outfit with high-waisted cobalt-blue tailored flared trousers matching the visible jacket, and black leather ankle boots with a low sculpted heel. Keep these trousers and boots identical in the front and back panels.

Tip: Only the wardrobe line differs from the CLAUDE prompt: flared cobalt trousers and black ankle boots. Both came out as written, and the coil keeps its shape from the back. One template for every member of the cast keeps the sheets consistent with each other: same white floor, same light, same three framings.

5. New head on the same body: blond bob and walrus moustache (Nano Banana 2 Lite)

AI character sheet edited with Nano Banana 2 Lite: the olive velvet suit and pose are unchanged, the head is now a square-jawed man with a blond bob and a thick dark walrus moustache
  • Made September 30, 2026
  • Image to image, 1 reference image (the CLAUDE sheet above)
  • Aspect ratio auto (16:9 output)
  • Thinking high
  • Resolution 1k

9 credits in Twistavo

Prompt

Use image_0.png as identity and composition reference for a faithful image-to-image edit. Keep the body, the brown checked tweed suit, white shirt, tie, pose, framing, background and lighting exactly the same. Replace ONLY the head and hair with this new design: a strong square face with a prominent chin, thick dark-brown walrus mustache in sharp contrast with the neat blunt blond bob, bushy expressive eyebrows, small sharp hazel eyes with a confident squint, weathered tanned skin, faint smirk of absolute self-satisfaction. A tiny braggart champion energy. 100% a real believable human. Ultra-photorealistic skin with visible pores, mustache hairs and natural asymmetry, looks like a real photograph, not CGI, not a caricature drawing. Same neutral lighting and photo style. No text, no labels, no watermark.

Tip: The head changed in all three views and the velvet suit stayed. Note the prompt: it says to keep a brown checked tweed suit, because the same text was written for another sheet. Here the model followed the picture and ignored that line. The next example shows the opposite.

6. Failure: the head changed, and so did the suit (Nano Banana 2 Lite)

AI character sheet edited with Nano Banana 2 Lite: a gaunt man with a strawberry-blond fringe and dark circles, now wearing a brown tweed suit instead of the olive velvet one
  • Made September 30, 2026
  • Image to image, 1 reference image (the CLAUDE sheet above)
  • Aspect ratio auto (16:9 output)
  • Thinking high
  • Resolution 1k

9 credits in Twistavo

Prompt

Use image_0.png as identity and composition reference for a faithful image-to-image edit. Keep the body, the brown checked tweed suit, white shirt, tie, pose, framing, background and lighting exactly the same. Replace ONLY the head and hair with this new design: a long narrow aristocratic face — sharp cheekbones, pronounced brow, large straight nose, thin disapproving mouth, strong jaw, deep-set pale grey eyes with heavy dark circles and a tired contemptuous stare, light freckles, short messy strawberry-blond hair with a blunt fringe. A gaunt exhausted nobleman who has seen everything. 100% a real believable human. Ultra-photorealistic skin with visible pores, subtle veins and natural asymmetry, looks like a real photograph, not CGI, not a caricature drawing. Same neutral lighting and photo style. No text, no labels, no watermark.

Tip: Same input sheet, same template, and this time the model followed the words: the olive velvet became a brown checked tweed suit. The head did not fully change either: the face stayed close to CLAUDE's, with a new fringe. Lesson: an edit prompt must describe the body that is really in the reference image, or the result depends on which of the two the model trusts on that run.

7. New head: an elderly woman's face on the suited body (Nano Banana 2 Lite)

AI character sheet edited with Nano Banana 2 Lite: the olive velvet suit is unchanged and the head is a smiling elderly woman with silver permed hair and small gold glasses
  • Made September 30, 2026
  • Image to image, 1 reference image (the CLAUDE sheet above)
  • Aspect ratio auto (16:9 output)
  • Thinking high
  • Resolution 1k

9 credits in Twistavo

Prompt

Use image_0.png as identity and composition reference for a faithful image-to-image edit. Keep the body, brown checked tweed suit, white shirt, tie, pose, framing, background and lighting exactly the same. Replace ONLY the head and hair with this design: the kind, warm face of a plump elderly grandmother on this suited man's body — soft round wrinkled cheeks, gentle laughing crow's feet, small kind eyes behind tiny oval gold-rimmed glasses, silver-grey permed hair set neatly, small pearl earrings, a faint benign smile. The face reads clearly as a sweet grandma in her seventies while the body is the adult man's tailored tweed suit. Uncanny, tender and absurd. 100% photorealistic: believable aged skin, visible pores and wrinkles, natural asymmetry, looks like a real photograph, not CGI, not a mask, not horror. Same neutral lighting and photo style. No text, no labels, no watermark.

Tip: The most extreme test of the series, and the edit held: the silver perm and the small glasses follow the prompt in the front and close views, and the body did not move. Three heads on one body make the casting choice concrete: put the sheets side by side and keep the one you recognize first at thumbnail size.

Make it in Twistavo

  1. Write one portrait prompt per character: name, age, one hairstyle that reads at thumbnail size, glasses or another accessory, a period outfit and an ordinary place. Generate it on GPT Image 2.5, 9:16.
  2. Open the sheet model in the AI Studio, upload the portrait as the reference image and paste the sheet template with one wardrobe line for the legs and shoes.
  3. Check the sheet: same face in the front and close views, outfit complete in the back view. Keep this file as the character's reference.
  4. To test another cast member, upload the sheet to Nano Banana 2 Lite and describe only the new head. Describe the suit that is really in the image.
  5. Use the sheet as the reference image of every later image or video of that character.

Variations to try

  • A cast of three: give each character one hair shape and one accent colour that no other member uses.
  • Keep the head and change the outfit: same sheet, an edit prompt that replaces only the clothes.
  • Age the character up or down for a flashback episode, with the same head-swap template.
  • Add a background character the same way: a text-only portrait first, then the same sheet template, so it can come back in later episodes.

Notes

Every person on this page is fictional and was generated with AI from the prompts shown; the chain starts from text, with no photo of a real person. Credit costs are Twistavo's quotes for the same model and settings on November 18, 2026. Results change from run to run.

Make your own with Twistavo — 1 free image

More AI video prompts

This prompt, its model, settings and result are part of the open dataset of every tested prompt (CSV and JSON).