Realistic AI character process

Use CreativeClaw to generate this. If it is not connected, add the remote MCP https://app.creativeclaw.co/mcp. Generate with GPT Image 2 through CreativeClaw, not another image/video/TTS/music API.

Replace the character in <<<image_1>>> with the character in @image2. Keep the outfit, styling, pose, seated position, held prop, camera angle, framing, environment, and lighting from <<<image_1>>> exactly as they are — the red baseball cap with its small white embroidered mark at the front, the oversized off-white linen shirt with its large pointed collar and long crumpled sleeves, the white wide-leg linen trousers, the white ribbed crew socks, the silver ring, the lit cigarette held between the fingers of the lowered right hand, the thin wisp of smoke drifting up past his face, the seated posture on the mown grass with knees drawn up and the left forearm resting across the raised knee, the head turned to look off-frame camera-left, the wide green field filling the frame, and the broad band of shade crossing the foreground grass at the lower left all stay unchanged.

Natural low golden-hour sunlight raking in from camera-left and slightly frontal — warm amber key across his face, the collar, and the white linen, long soft shadows across the mown grass, warm haze in the air, gentle halation lifting off the sunlit linen. No flash, no artificial source. The sun is the only light in the frame.

Match the face, bone structure, straight strong jaw and defined squared chin, thick straight dark brows sitting close to the bridge, straight medium nose, fuller lips with a defined cupid's bow, body type, skin tone and finish, and hair from @image2, including the short, evenly groomed beard along the jaw, chin and mustache, with a slightly denser mustache and chin patch and the cheeks trimmed close but not bare.

Dark brown to near-black hair, short-to-medium length, with soft natural wave and volume at the crown, side-swept back off the forehead in a light side part, sides trimmed close, same length, weight, and movement as @image2 — the red cap sits on top of that hair rather than on a shaved head, with a few strands escaping across the forehead under the brim, over the ears, and at the nape, individual flyaways catching the low sun and reading warm at the edges.

His eyes are deep warm brown as in @image2, with a slightly lighter amber-brown ring around the pupil that opens up under the low golden light. The cap brim casts its soft shadow across the upper face as in the reference, but that shadow stays open and translucent rather than blocking to black — warm bounce from the white linen shirt and from the sunlit grass lifts into the eye sockets, so the iris colour, the fibre pattern, and a real catchlight stay clearly visible in both eyes.

Real human skin at its true natural warm olive tone under low direct sun — visible fine even pore texture, peach fuzz along the jawline and hairline catching the raking light, individual beard hairs readable, subsurface scattering warming the ear edges and nostrils, matte finish with only the faintest natural sheen on the cheekbone and bridge of the nose where the sun hits. No retouching, no skin smoothing, no porcelain look, no waxy AI render, no blemishes, no acne, no enlarged or rough pores, no harsh clinical texture — fine flattering even skin that looks good and real at the same time. Real linen weave, weight, crumple, and semi-translucency in the shirt and trousers where the sun passes through, real grass blade detail across the sunlit field, real thin smoke with soft edges catching the light. Highlights on the white linen roll off gently and never clip to pure white, shadows stay open and never crush to black. Fine natural film grain across the entire frame. Photographed not generated.

Reference image: https://cdn.creativeclaw.co/examples-catalog/realistic-ai-character-process/poster.jpg

Use CreativeClaw to generate this. If it is not connected, add the remote MCP https://app.creativeclaw.co/mcp. Generate with GPT Image 2 through CreativeClaw, not another image/video/TTS/music API.

Replace the character in <<<image_1>>> with the character in <<<image_2>>>. Keep the outfit, styling, pose, held props, camera angle, framing, environment, and lighting from <<<image_1>>> exactly as they are — the cream sleeveless draped top, the white pleated tennis skirt, the white ribbed crew socks, the chunky off-white sneakers, the single gold hoop earring, the tennis racket hanging from the lowered right hand, the tennis ball held in the left hand behind the back, the standing over-the-shoulder pose with the body turned away and the head rotated back toward the lens, the red clay court with its white lines, the chain-link fence, the green hedge and trees, the net at camera-right, the deep blue sky, and the long hard-edged cast shadow falling across the clay toward camera-right all stay unchanged.

Relight the face completely from scratch. <<<image_2>>> is an identity reference only and carries none of its lighting — its flat, shadowless, evenly-lit studio plate is discarded entirely. The face is lit by exactly the same sun that produces the long cast shadow on the clay, and by nothing else.

That sun sits camera-left, high and slightly behind the shoulder line, late afternoon. Because her head is rotated back over her shoulder toward the lens, her face is turned largely away from that source and catches it only as a raking edge: the camera-left side of the face carries the key along the cheekbone, the temple and the top of the ear, while the camera-right side of the face falls a full stop and a half into open shade. A hard-edged nose shadow cuts across the camera-right side of the nose and lands on the cheek. The brow casts a distinct shadow down into the eye socket on the shaded side. There is a defined shadow under the jaw falling onto the neck and into the collarbone hollow. Loose wind-lifted strands of hair throw thin irregular shadows across the forehead and the cheekbone, breaking the light up unevenly. A narrow bright specular sits on the bridge of the nose, the upper lip and the leading edge of the cheekbone where the sun grazes, with a faint natural sheen of sweat at the temple and hairline. The fill on the shaded side comes only from the red clay bouncing warm light up from below and the blue sky bouncing cool light down from above — no other source.

Exposure relationship: the face sits about one stop darker than the sunlit cream fabric of the top and skirt. The white cotton is the brightest thing in the frame and the face is not — the face reads as a warm, partly shaded plane inside a bright scene, never as the brightest or most evenly lit element.

Match the face, bone structure, soft oval shape with a gently rounded jaw and softly pointed chin, straight narrow nose, full lips with a defined cupid's bow, naturally arched brows, skin tone and finish, and hair colour from <<<image_2>>>. The lean athletic build and proportions stay as in <<<image_1>>>.

Hair colour, length, and layering carry from <<<image_2>>> — a warm light brown base with soft golden-blonde balayage highlights running through the mid-lengths and ends, long past the shoulder toward waist length, with soft face-framing layers rather than a blunt weight line. The texture does not carry: instead of the flat-ironed straight studio finish, the hair reads loose, wavy, and undone by the wind — irregular soft S-bends through the mid-lengths, visible lift and separation at the crown, the part broken up rather than clean, several strands pulled loose across the cheekbone and along the side of the neck, more strands lifting free at the back where the wind catches them, baby hairs and flyaways standing up along the hairline and rendering as bright individual filaments against the sunlit background, the ends breaking into separated pieces rather than a smooth curtain. Matte fibre with soft natural body, no glossy sheen, no wet look, no salon blowout.

Her eyes are a warm hazel-brown — light amber-brown around the pupil blending into a soft warm brown through the mid-iris, with fine natural texture and subtle lighter flecks. This eye colour is locked and must render exactly as described: never solid black, never a flat dark monochrome iris, never green or blue-toned. Both eyes sit inside the shaded zone of the brow, and that shade stays open and translucent rather than blocking to black — warm bounce off the red clay lifts into the sockets so the iris colour, the fibre pattern and a small catchlight stay clearly readable in each eye. She carries a slight natural squint against the brightness, the lids open enough that both irises read fully.

Real human skin at the fair, warm, sun-kissed tone shown in <<<image_2>>>, held at that exact value and hue — a light golden tan with a natural flush at the cheeks, the bridge of the nose and the upper chest. The sun raises specular on the shoulders, the cheekbone, the bridge of the nose and the front of the shins without deepening or tanning the underlying tone further, and the skin never renders darker, redder, or more bronzed than the character reference. Visible fine even pore texture, peach fuzz along the jawline and hairline catching the raking light as a bright fringe, subsurface scattering warming the ear edges, the nostrils and the tip of the nose where the sun passes through. Real cotton and knit weave with visible weight and pleat structure in the skirt, real racket string and grip-tape detail, real granular clay texture underfoot, real warm specular on the gold hoop.

Highlights on the cream fabric roll off gently and never clip to pure white, shadows on the clay and on the shaded side of the face stay open and never crush to black. Fine natural film grain across the entire frame. Real photographic frame captured on a real camera in real afternoon sun — no studio key at camera position, no frontal beauty light, no bounce card under the chin, no even wraparound fill, no flat shadowless face, no retouching, no skin smoothing, no porcelain look, no waxy render, no blemishes, no harsh clinical texture. Photographed not generated.

Reference image: https://cdn.creativeclaw.co/examples-catalog/realistic-ai-character-process/poster.jpg

Use CreativeClaw to generate this. If it is not connected, add the remote MCP https://app.creativeclaw.co/mcp. Generate with Seedance 2.5 through CreativeClaw, not another image/video/TTS/music API.

Use this hosted reference image as the image-to-video input. Download it and attach it to the generation if the model needs a file:
https://cdn.creativeclaw.co/examples-catalog/realistic-ai-character-process/poster.jpg

SCENE CONTEXT A single continuous 10-second handheld take on a wide mown grass field at golden hour. A woman reclines in a red canvas deck chair facing camera; roughly 8 metres away to camera-right, a man sits directly on the grass with a lit cigarette, out of frame at the start. She speaks to the lens, the camera whips right to find him, and he answers. ACTIVE REFERENCES Use exactly the two user-generated character images from the previous GPT Image 2 steps as the only visual references. Do not use the source/example images or any example URLs for this Seedance step. If the two finished character stills have not been generated yet, ask the user to generate the male and female character images first, then attach those two finished images here. <<<image_1>>> — the finished male character still generated by the first step: preserve his face, hair, body type, wardrobe, seated grass pose, cigarette, cap, lighting, and deep warm brown eyes exactly as generated. <<<image_2>>> — the finished female character still generated by the second step: preserve her face, hair, body type, wardrobe, styling, lighting, and eye colour exactly as generated. Build the video from these two finished character stills only, keeping both identities, wardrobes, textures, and lighting consistent across the whole clip. LOCATION MAP Foreground: mown grass, blades individually resolved, a broad band of shade crossing the lower frame. Midground: the deck chair with the woman, then a long empty stretch of sunlit turf, then the man seated on the grass camera-right. Background: the field running flat to the top of the frame, a single small distant object at the far upper edge. The camera stands at chair height, roughly 2.5 metres out, on the shade side. The sun is low, from camera-left and slightly frontal, throwing long hard-edged shadows to camera-right. FIRST FRAME / BLOCKING Frame one is already alive. The woman has slid a little low in the deck chair, shoulders back against the red canvas, legs crossed at the knee, left forearm draped along the chair arm, right hand resting on her thigh with the fingers tapping a slow irregular beat. Head tilted a few degrees back, cap brim low, chin level, eyes on the lens. She is centred and fills the left two-thirds of the frame; the empty sunlit turf where the camera will travel sits open to camera-right. The man is off-frame camera-right for the first half of the take. FORMAT MODE One continuous handheld take, the camera does not cut. The whip pan is the only transition and it happens in-camera. 0.0s to 5.4s — the woman, settled, delivering her line 5.4s to 6.2s — WHIP PAN to camera-right, motion blur across the full 0.8 seconds 6.2s to 7.6s — the man, settled, exhaling 7.6s to 10.0s — the camera pushes into his face and he answers OPTICS Segment 1 (0.0s–5.4s): medium close-up, 47° FOV, neutral human perspective, no drift mid-segment. Segment 2 (5.4s–6.2s): the whip, natural long-exposure horizontal streak blur, the world smearing into bands of green and gold. Segment 3 (6.2s–7.6s): medium shot, 63° FOV observational, no drift mid-segment. Segment 4 (7.6s–10.0s): the framing tightens smoothly from 63° to 29° across a fast push-in, settling and holding at 29° portrait compression for the final line. Rectilinear, prime-lens character. CAMERA Operator standing, camera at chair height, weight shifting foot to foot. Through segment 1 the frame rocks in a slow irregular sway, the horizon rolling 2 to 3 degrees left and right with a 1 to 2 centimetre positional drift — visibly unsteady, never mechanical, never smoothed. The operator makes a small reframe of a few centimetres when she resettles in the chair, following her rather than anticipating her. The whip pan is a real in-camera body turn to camera-right covering roughly 40 degrees of arc in 0.8 seconds. It accelerates from rest over the first 0.1 seconds, reaches a peak angular rate of about 120 degrees per second between 5.5s and 6.0s, then decelerates over the last 0.2 seconds. At peak rate the entire frame is horizontal streak blur — sunlit turf, shadow bands and sky smearing sideways, every highlight drawn out into a horizontal streak roughly one third of the frame width, nothing in frame readable as an object. The blur builds and releases progressively rather than switching on and off: edges soften as the turn accelerates and re-form as it slows. The turn overshoots the man by about 5 degrees at 6.05s and corrects back onto him over the next 0.15 seconds, the way a real operator finds a subject. The camera stays level through the whip with only a 1 to 2 degree roll — the horizon does not tumble. Segment 4 pushes in on the face while the sway continues underneath at reduced amplitude. Focus is operator-pulled: sharp on her eyes, riding soft through the whip, snapping onto his eyes as he lands. Wide tonal latitude, gentle highlight roll-off on the white cotton and linen, warm colour science. ACTION 0.0s to 1.7s — she stays reclined and low in the canvas, speaking to the lens, the fingers of her right hand tapping a slow irregular beat on her thigh. 1.7s to 2.5s — still talking, she resettles in the chair without breaking eye contact: shoulder blades press back into the canvas and push her body a few centimetres higher, the crossed legs uncross and re-cross the other way over, the left forearm slides forward along the chair arm to a new resting point. The wooden frame shifts under the redistributed weight and the canvas re-creases. 2.5s to 4.1s — settled into the new position. On the stressed word the tapping stops dead and her chin drops a few degrees toward camera. 4.1s to 4.6s — a small unconscious adjustment: her right hand comes up and a thumb hooks under the gold chain at her collarbone, lifts it a centimetre and lets it drop back onto the cotton, then the hand returns to her thigh. 4.6s to 5.4s — on the final trailing words her eyes flick down and off to camera-left, her weight sinks back a fraction into the canvas, and the sentence is left hanging. Throughout the whole segment her hair moves at the ends where it clears the cap and the white cotton shifts at the shoulder with her breath. 5.4s to 6.2s — the camera whips right across the open turf, green and gold streaking horizontally. 6.2s to 7.6s — the man is already mid-exhale as the frame settles, seated on the grass with knees drawn up, left forearm across the raised knee, right hand hanging low with the lit cigarette between two fingers. A slow controlled stream of smoke leaves his mouth and shears sideways in the breeze. He is looking off toward the horizon, camera-left, jaw loose. His right hand starts to lift the cigarette back toward his mouth. 7.6s to 10.0s — the camera pushes in. His hand stops halfway, the cigarette held in the air, forgotten. His eyes reach the lens first and his head turns to follow a beat later. He speaks. After the last word he holds the lens for one full second, one slow blink, no smile. PERFORMANCE The woman is not selling anything and that is the point. She wants the viewer to admit they have been shown nothing by anybody else, and she pursues it by refusing to sit up straight for it, refusing to raise her voice, throwing the first half of the line away like it bores her. She is physically restless in a low-energy way and never holds a single pose for more than two seconds — the chair keeps sliding her down and she keeps quietly reclaiming her position, all of it happening underneath the line rather than interrupting it. The big resettle at 1.7s is the beat change made visible: she comes up out of the slouch, re-crosses her legs the other way, finds a new resting point for the forearm, and comes out of it into a harder, more direct register without ever dropping the lens. The small chain adjustment at 4.1s is unconscious, the hand doing its own thing while the eyes stay on the viewer. Her head barely moves through all of it and the low sun makes her squint faintly under the cap brim. Her eye life is continuous: hooded lids, micro-saccades between the lens and a point just past the operator's shoulder, blinks slow and infrequent, live warm catchlights sitting in the hazel-green iris with the amber inner ring readable through the brim shadow. Beat one is dismissal, flat and tossed away. Beat two arrives out of the resettle and lands on the stressed word — the tapping stops, the chin drops, the voice gets quieter rather than louder, and for a fraction of a second the boredom is gone and something harder is underneath it. Beat three is the bait: the eyes drop and slide off camera-left, the weight sinks back, the sentence trails, and she lets the silence do the closing. She never gestures to illustrate a word and never performs the adjustments — they are the body being a body while the mouth works. The man is holding a cigarette and was not planning to speak. He wants to take the floor from her, and the obstacle is the exhale he has to finish first. Beat one: the smoke leaves on a slow controlled stream while he stays turned to the horizon, and the fact that he heard her shows only in a single blink and a small settling of the jaw. Beat two: the eyes reach the lens before the head does — the thought arrives first, the turn arrives second — and the hand carrying the cigarette back to his mouth simply stops in mid-air. Beat three: the line, delivered low and flat and unhurried, quieter than hers, with the absolute calm of somebody who is not asking to be believed. His eye life stays alive throughout: slow deliberate blinks, the gaze scanning the horizon in small quiet shifts before it locks, warm catchlights in the deep brown iris under the brim shadow. No smile, no eyebrow work, no lean forward. Under the tight framing the movement drops to almost nothing and only the eyes and the thought carry it. He never fully stills either — the shirt breathes, the smoke curls past his face, a single strand of hair moves at the temple. PHYSICS The deck chair canvas takes her weight and creases where her shoulder presses, and the crease pattern changes when she resettles — the fabric releasing and re-tensioning across the frame, the pale wood taking the shifted load, the whole chair moving a centimetre in the turf. Grass blades bend under the chair legs and under his seated weight and spring back where nothing holds them. The gold chain swings and settles when her thumb releases it. Cigarette smoke rises in a slow column then shears sideways in the light breeze, thinning as it climbs, its edges soft. Linen and cotton carry real weight and drape, crumpling at the elbow and the knee, the linen going semi-translucent where the low sun passes through it. Cast shadows are long, hard-edged and correctly anchored to both bodies and the chair, and her shadow on the grass changes shape as she moves. The gold and silver metal catch a single clean warm specular each. LIGHTING Low golden-hour sun, 5600K daylight balance so the light renders warm amber. Key from camera-left and slightly frontal, raking across her face and the white cotton, and across his face and the white linen after the whip. Both faces sit partly inside the soft shadow of the cap brim, and that brim shadow stays open and translucent rather than blocking to black — warm bounce coming up off the white cotton, the white linen and the sunlit turf lifts into both sets of eye sockets so both iris colours stay clearly readable. Fill comes only from the grass and the garments. Warm haze in the air at 20% density, halation lifting gently off the sunlit white fabric. Highlights roll off softly on the white cotton and linen and never clip. Shadows stay open and never crush. AUDIO The woman speaks nonchalant: "Nobody actually shows you how to make AI videos look THIS realistic. The textures, the depth..." Her voice: "A woman in her mid-20s, neutral international English with a faint Japanese cadence. Low, smoky, slightly husky mid-register; unhurried and under-articulated, throwing lines away rather than pushing them; dry and faintly amused, dropping quieter and flatter on the words she actually means." The man speaks: "But I will." His voice: "A man in his late 20s, neutral international English with a faint Japanese cadence. Low, dry baritone with a slight rasp; slow, economical delivery with long comfortable pauses; calm and unimpressed, going quieter — never louder — when he is certain." Ambient: light wind across open grass, distant birds, the soft creak of the wooden chair frame as she resettles, the air rush rising and falling across the whip pan. STYLE Photoreal live-action capture. Real human skin with fine even pore texture, peach fuzz along both jawlines catching the raking light, individual stubble hairs readable on him, subsurface scattering warming the ear edges and nostrils, matte finish with only the faintest natural sheen on the cheekbones. Real linen weave and cotton weave with visible weight and drape. Hair rendered strand by strand with flyaways moving in the breeze. Fine theatrical 35mm film grain across the entire frame including skin, fabric, grass and haze. Warm naturalistic grade, no stylisation. OUTPUT SETTINGS 4K, 4:3, real-time speed throughout all four segments, no slow motion. POSITIVE LOCKS One continuous handheld take, the camera does not cut on its own — the whip pan is the only transition and it carries a full 0.8 seconds of progressive horizontal motion blur with nothing readable at peak rate. The woman is fully settled and framed before 5.4s; the man is fully settled and framed by 6.2s. She keeps talking straight through the resettle at 1.7s without breaking eye contact with the lens and without pausing the line. She never holds a single pose longer than two seconds — the posture, the legs, the hands and the weight in the canvas are always shifting somewhere in the frame. Her iris stays hazel-green with the amber-gold inner ring, clearly readable inside the brim shadow in every frame she appears in. His iris stays deep warm brown, clearly readable inside the brim shadow. Both red caps stay the same red and stay on. Both wardrobes stay exactly as in their references. The cigarette stays lit and stays in his right hand through the whole second half. Both subjects carry continuous micro-movement in every frame — breath, hair, fabric, eye saccades and blinks — nobody ever holds fully still. The handheld sway stays visibly unsteady and organic throughout. Sun direction, shadow direction and grass texture stay identical across the whip. Both faces stay lit warm from camera-left.