Until now, one video was one video. A new face, a new product or a new location meant a new shoot, or rebuilding the whole thing from zero. Once you have a locked character, that changes: the face and the character sheet stay fixed, and only the scene changes.
That is the reason one good character is no longer one video. It becomes a format you can run forever: different setting, different product, different audience, same person. Below is the whole pipeline in the order it runs, done in Studio AI, with every prompt.
Part of our AI Video Generation Statistics 2026 series.
1 character
Becomes endless ads. Lock the face once, then put her in any scene you can describe.
Four steps, all in Studio AI: the face, the character sheet, the talking clip, then putting it all together.
Our take: Nothing here is theory. It is the same pipeline, step by step, with the prompts.
Free AI Video Hub01 / 6
The pipeline in one look
Four steps, each one feeding the next. The face decides how real everything looks, the sheet locks the identity, the clip adds the performance, and the final prompt puts her in any scene you can describe.
| Step | Model + tool | What it gives you |
|---|---|---|
| 1 · The face | Seedream 5.0 in Studio AI | Skin, light and the real-photo look. |
| 2 · The character sheet | GPT Image 2 in Studio AI | The same face from every angle. |
| 3 · The talking clip | Seedance 2.5 in Studio AI | Emotions, tone, expressions and movement. |
| 4 · Putting it all together | Seedance 2.5 in Studio AI | Any scene you can describe, with the same character. |
02 / 6
Step 1: The face
Everything is decided here. If the base image already looks like a real photo, the video stays real. If the image looks like AI, nothing downstream saves it.

A tight and hyper detailed close-up, straight-on selfie captures a young adult Caucasian woman, likely in her early 20s, reclining with her head propped on her left hand. She displays striking, symmetrical, model-like facial features with full lips, strong brows, and long dark eyelashes. Her brown hair is styled sleekly, framing her face and tumbling over her shoulder. The woman is wearing a black robe or jacket with a deep neckline, revealing smooth skin and a hint of cleavage, emphasizing the glossy, hydrated finish of her complexion. In a playful, flirty gesture, her tongue touches her upper lip as her gaze looks off to the side, outside the frame. There is no visible text, and no logos or brands are discernible. The background consists of a softly focused cream-colored wall with a dark fabric or curtain to the left. The lighting is warm and diffused, likely from an artificial indoor light source positioned above and to the side, giving the scene a gentle orange-yellow hue. The color palette is dominated by warm skin tones, dark hair, and black clothing with cream and slate accents in the background. This image exhibits typical smartphone characteristics: shallow depth of field blurring the background with digital sharpening and slight smoothing of skin texture. The dismissive, playful expression paired with the framing and soft-focus lends the photo a lighthearted, flirtatious, and intimate mood.
HEX VALUES: ["#2b2c3d", "#373746", "#201d30", "#de9d87", "#0e0f2a", "#564c56", "#e7b39f", "#c68c7d", "#4e393d", "#6b5f67", "#f7d899", "#b27765", "#735044", "#8d6152"]Opens with this prompt already filled in. No signup.
03 / 6
Step 2: The character sheet
One image gives you one angle. A character sheet gives the video model the whole head, so the face stays the same person when she turns, talks and moves.

Character sheet of the woman from the reference photo [img1] - the same person shown in the provided image. Preserve her exact facial identity, bone structure, hairline and hair shape from the reference - do not restyle or idealize her: long dark brown wavy hair with a soft center-to-side part and loose waves framing the face, warm lightly tanned olive skin, large deep brown eyes with long natural lashes, full dark well-groomed brows, high cheekbones, a straight nose, full lips, and a soft defined jaw. Keep her natural asymmetry and any faint freckles or skin variation from the reference. Vertical 9:16 frame divided into an even 2×2 grid of four views of the same person. Top left, top right and bottom left are close-up head-and-shoulders views cut at the upper chest, all three at identical head size, identical eye-line and identical camera distance: top left, front view facing the lens; top right, her left-side profile; bottom left, her right-side profile. The bottom right quadrant is different - a front-facing full-body shot showing her entire outfit and figure from shoulders down, framed so the top of the frame cuts across at the neck and the head is not included. Same neutral posture throughout - head level in the three portraits, shoulders relaxed and square, calm neutral expression, lips lightly closed, gaze straight ahead; standing relaxed and upright in the full-body view, arms resting naturally at her sides. Wardrobe identical across all four views, exactly as in the reference: black one-shoulder top with an asymmetric neckline and a structured black blazer draped over one shoulder, natural creasing across the fabric, shown in full length in the full-body quadrant. Seamless off-white studio backdrop, large soft frontal key with gentle fill from both sides, neutral white balance, soft shadow falloff under the jaw. 85mm lens at eye height for the portraits and a matching clean full-length framing for the body shot, no distortion, clean headroom in each quadrant, sharp throughout. Realistic skin with natural texture and tone variation, faint natural blemishes and asymmetry kept, hair reading as separate strands rather than a smooth mass, visible fabric weave and creasing. No retouching or smoothing, no beauty lighting, no glamour styling, no stylization. No text, labels, borders or watermarks. [img1]Opens with this prompt already filled in. No signup.
04 / 6
Step 3: The talking clip
This is where you write the things that make it feel human: emotions, tone, expressions, where the eyes go, when she breathes. Skip those and you get a face that talks. Write them and you get a person. Use Seedance 2.5 in Studio AI, upload the face image as @img1 and the character sheet as @img2.
REFERENCE MAP
@img1 = subject AND the exact first frame of the clip. Her face, hair, wardrobe, the warm wall, the dark curtain and the ring-light catchlights in her eyes all continue unchanged from this frame.
@img2 = her character sheet, for identity across angles. Match face and hair to it whenever her head turns.
FORMAT
One unbroken 8 second take, handheld front-camera selfie. No cuts, no transitions, no edits of any kind. The filming phone IS the camera and is never visible; no phone appears anywhere in frame, no mirror, no reflection of a phone. The phone hand stays welded to it off screen the whole time; every visible movement belongs to her one free hand.
Camera character: iPhone-style front camera, 24-28mm equivalent, arm extended roughly 35-55cm. Her face is never locked to centre. It drifts continuously, rides high or low, sometimes a third off frame, and her corrections are loose and late.
The ring light sits behind the phone and is NEVER visible in frame. Only its effect is seen: soft frontal light on her face, a faint halo rim on her hair, and a sharp circular catchlight in each eye that re-catches at a new angle on every head turn. Every head move carries her hair with real live physics.
Audio is arm's-length phone-mic capture, never close-mic, never studio-present: raw and unprocessed like a clip straight from a phone gallery, with natural volume fluctuation. Under it, unbroken indoor room tone and faint house hum, plus faint unintelligible voices from elsewhere in the house once. No music.
TIMELINE (8s, single take, no cuts)
0.0-1.0 SETTLE, NO SPEECH. Frame one is exactly @img1. The camera has just been raised and is still settling: over the first 0.3s the angle levels from slightly below toward eye height, her face swings into the upper frame then drifts off left.
Her free hand leaves her hair and travels back toward her in real time. Palm turns, forearm folds under her cheek, her head's weight visibly settles onto it. Beginning, middle and end of the movement all happen on camera, nothing is already done.
Her expression eases out of the frame-one pose into a relaxed pre-speech look, eyes on the lens the whole way. A near-inaudible creak as her weight shifts. Nothing is spoken in this first second.
1.0-4.3 LINE ONE. "I was made with Studio AI." Warm and easy, like a video message to a friend, not something filmed for an audience. Eyes locked into the lens, chin dipping slightly on the first word.
Delivered completely flat and unbothered, as if it were the least interesting fact about her. No smirk, no wink, no pause for effect. She says the whole line once. No word is ever repeated.
The phone mic auto-gain settles over the first syllables then breathes back up, a capture artifact not a mix choice.
4.3-4.9 THE SHIFT, SIGNATURE MOMENT. As she draws breath into the second line her weight shifts on her forearm and the phone tips a few degrees in her grip. Autofocus hunts: her face goes soft for about 0.3s, the lens searches past her toward the wall behind, then snaps back and locks hard on her eyes. Auto-exposure rides up a stop and clamps back down, and the white balance drifts a touch cooler then corrects.
Entirely silent and in-lens. Nothing in the room changes, no light source ever enters frame. She does not react to it, she is mid breath with head and eyes already moving into the next line. One soft low frequency mic brush as her fingers reset on the phone body.
4.9-7.0 LINE TWO. "If you want the same, comment Studio and a human will send you the prompt." Steadier now, a touch warmer, easing a few centimetres closer to the lens. The word "Studio" is said clearly and slightly set apart from the words around it so it reads as an instruction.
A small lift at the corner of her mouth arrives on "a human", the only tell that she knows the line is funny.
7.0-8.0 THE LAUGH. She breaks into a short genuine laugh at her own line, eyes crinkling, head dipping a few degrees toward her forearm, one small shoulder bounce. Brief and breathy, not a big laugh, cut off by the clip ending rather than by her stopping.
Her framing hand gives one last loose correction. Ends mid-laugh, no settle, no fade, no wrap-up gesture.
CONSTRAINTS
Two lines only, delivered once each. No extra words, no repeats, no voiceover, no narration.
The ring light, lamp, softbox, light stand or any glowing circle is NEVER visible in frame at any point. No lens flare from a light source. Only catchlights in her eyes.
The phone is never visible. No mirror, no selfie stick, no second person, no hand reaching in.
One laugh only at the very end. No laughing during the lines, no covering her mouth.
No speed manipulation, no transitions, no digital zoom, no stabilisation, no gimbal smoothness, no slow motion, no music, no subtitles or on-screen text.
Everything on screen is capture behaviour or performance. Nothing is a post effect.
Do not change her face, hair, wardrobe, the wall, the curtain or the lighting from @img1.
Lip sync must match the lines exactly.Opens with this prompt already filled in. No signup.
05 / 6
Step 4: Putting it all together
Once you have the face, the character sheet and a clip that holds her identity, you can create anything. The character is locked, so a new scene is just a new setting. Below is one more prompt built on the same @img1: a five second, three shot hotel room sequence with two hard cuts. Same face, completely different world.
REFERENCE MAP
@img1 = subject: her face, hair and likeness. Match exactly.
FORMAT
Cinematic narrative sequence, 5s, three shots, two hard cuts. 16:9, 24fps, full-frame cinema camera. Warm tungsten practicals against cool shadows, amber highlights, gentle halation, fine 35mm grain, visible skin texture and pores, no smoothing.
Classic hotel room, dark wood, beige walls, lived-in rather than boutique. Evening, curtains drawn, all light from practicals.
Wardrobe continuous: dark navy blazer, slate blue top, matching trousers, small gold hoops, long dark hair parted down the middle.
SHOT 1 (0.0-1.5) OVER THE SHOULDER, VANITY MIRROR
85mm, shallow focus, camera at her seated eye height just behind and right of her head. The frame looks past the dark out-of-focus silhouette of her right shoulder and jaw, which holds the left third and stays soft, directly into a vanity mirror with faint smudges on the glass.
A tungsten wall sconce with a white shade sits at the left edge, blooming softly and throwing warm directional light across the reflected side of her face.
In the reflection she holds a small dark-handled brush in her right hand and works eyeshadow into her left eyelid, which appears on the right of frame. Small precise strokes, wrist steady. Expression neutral and absorbed, one blink. Only the reflection is sharp.
SHOT 2 (1.5-3.0) PUNCH-IN CUT, SAME MIRROR
Hard cut, not a zoom. Same mirror, same angle, same light, tighter lens: 135mm on her reflected face from brow to chin, shoulder gone from frame.
She lowers the brush hand out of the bottom of frame and it stays there. She raises her left hand and blends the shadow with the pad of her ring finger, three small dabs. Her eyes close smoothly as she blends and open naturally after. Lips slightly parted, quiet focused intensity, no glance at camera.
SHOT 3 (3.0-5.0) HARD CUT, ROOM WIDE
Hard cut to a locked-off tripod, 35mm, deeper focus, chest height across the room. No camera movement at all.
She stands leaning slightly forward over the foot of the bed, same blazer now with trousers. The bedspread is heavily textured with broad horizontal stripes in navy, rust orange and cream. A large structured brown leather duffel sits open on it, silver hardware, worn corners.
She rests her right hand on the zipper pull, then draws it closed in one steady pull and puts it over her shoulder as she walks off the frame
Behind: dark wood nightstand, lit bedside lamp with a flared white shade, black landline telephone, beige wall with framed pictures. Warm pools of light, corners in shadow.
AUDIO
Close dry room tone, carpet and heavy curtains, faint HVAC hum. Minimal foley: brush friction on skin and blazer rustle, a faint tap as the brush hand lowers, then shoes on carpet and the long heavy metallic drag of the duffel zipper seating with a click. No dialogue, no music, no voiceover.
CONSTRAINTS
The camera, operator, crew, tripod or light stand is NEVER visible in the mirror. The reflection shows only her, the wall behind her and the sconce.
The reflection is geometrically correct and consistent: hair parting, brush side and sconce all flip the same way. No second face, no duplicate hand.
One hand per job. The brush stays in her right hand throughout, her left hand only blends. No hand swaps, no extra fingers, no brush in both hands.
The brush never touches her eyeball. No product floating or smearing.
She never looks into the lens, never speaks, no smile, no expression change beyond concentration.
Both cuts are instant hard cuts. No dissolve, no fade, no whip, no morph, no zoom.
Shot 3 is fully locked off. No handheld, no drift, no push-in, no rack focus.
Nobody else in the room. No hand entering frame, no second reflection, no background movement.
No logos, no readable text anywhere. No slow motion, no speed ramp, no lens flare, no on-screen text.
Identity, hair and wardrobe consistent with @img1 across all three shots.Opens with this prompt already filled in. No signup.
06 / 6
What you can do with this
Once the character is locked, every new video is just a new setting.
Ready to make your own?
Start with the face prompt, build the sheet, then put her in any scene in Studio AI.
Frequently asked questions
How do you make endless ads from one character?
Lock the character once: a realistic face, a character sheet from every angle and a talking clip that holds her identity. After that, every new video is just a new setting, so the same person can appear in any scene, with any product, for any audience.
Why does the character sheet crop the head out of the full-body shot?
If a sheet shows a small distant head and a close head at the same time, the video model does not know which face to use. When it picks the far one, the face comes out washed out and the skin goes smooth and plastic. Keep the head only in the close-ups.
How do you make an AI talking clip feel human?
Write the human details into the prompt: emotions, tone, expressions, where the eyes go and when she breathes. Without them you get a face that talks. With them you get a person.
Can I create any scene with the same character?
Yes. Copy the Step 4 prompt, paste it into Claude and ask it to write a new one that matches your request. As long as you are clear about the setting and the character reference, you will get a quality output.
