"Did You Think I Was Real?" The Exact Prompt Behind a Realistic AI Fashion Film

By Clayton H · Published 7 November 2026 · Tools: ChatGPT, Nano Banana, Seedance

The “Did you think I was real?” fashion film is one 10-second Seedance 2.5 generation with hard cuts between eight shots. It uses two reference images: one for the character and her outfit, one for the location. The prompt then sets the attitude, gives every shot a timecode, camera and action, and lists every spoken line, including the lines heard over the close-ups. Realism comes from asking for 35mm film texture and natural skin, and from banning the glossy, HDR look most AI video defaults to.

What you get

• Two image prompt templates: your character in the outfit, and the location.
• The full multi-shot Seedance prompt with timecodes, camera moves, attitude and dialogue.
• How to adapt it to your own character, outfit, setting and script.

What you need

• An image generator that accepts a reference photo (ChatGPT or Nano Banana).
• A character reference of your AI character, ideally a character sheet with a clear face.
• Seedance 2.5 (used via Dreamina for the original film).

How to make it in 3 steps

1. Make Image 1 (the look). Upload your character reference and generate your character in the outfit with the first template below.
2. Make Image 2 (the scene). Generate the location with the second template.
3. Generate the film. In Seedance 2.5, attach Image 1 and Image 2, paste the multi-shot prompt and generate at 9:16, 10 seconds.

Image prompt 1: the character and outfit

Use the attached character sheet as the identity reference only: keep her exact face, bone structure, eye colour and freckle placement. Restyle her as described below. She wears [OUTFIT: an oversized cobalt #1F4FBF wool coat with sculptural shoulders and a shawl collar, a cream rib knit and cream trousers], with [JEWELLERY: gold earrings, a gold cuff on her left wrist and a gold signet ring on her right hand]. Full-length, standing facing the camera, calm and faintly amused. Plain white studio background, soft even light. Natural skin with visible pores and freckles. Vertical 9:16.

Image prompt 2: the scene

No people. [LOCATION: a pale gravel path through a formal French parterre garden on a bright, slightly hazy morning]. [DETAILS: low clipped boxwood hedges in geometric patterns, tall clipped yew cones and a pale limestone palace façade softly out of focus in the background]. Soft, diffused morning light. Vertical 9:16, 50mm lens, shallow depth of field, authentic 35mm film texture, muted natural colours.

The multi-shot Seedance prompt

Attach Image 1 and Image 2, then paste:

Vertical 9:16, 10 seconds, one continuous generation with hard cuts between shots. Modern fashion-film realism on 35mm: fine organic grain, natural skin with pores and freckles, soft halation, no HDR, no glossy commercial finish.

References: Image 1 controls her identity (face, eye colour, freckles, red lip, black bob with micro-fringe), the cobalt wool coat, cream knit and trousers, the gold earrings, the gold cuff on her left wrist and the gold ring on her right hand. Image 2 controls the formal parterre garden, the gravel paths, the yew cones, the pale façade, the soft overcast daylight and the palette. Do not use either image as the first frame or copy its pose.

ATTITUDE: cool, unbothered, editorial. Chin slightly lifted, steady eye contact, slow deliberate blinks, a barely-there smirk rather than a smile. She speaks low and close, unhurried, almost bored, as if letting the viewer in on something obvious. Minimal head movement, no eyebrow raising, no big smiles, no eager energy.

0:00–1:40 — Medium shot from mid-thigh up, 50mm, slow push-in. She is already walking toward camera, stops, settles her weight and holds the lens with a level, faintly amused look. She says: “Did you think I was real?”

1:40–2:60 — Cut to medium close-up, 85mm, locked off. A slow blink, a tiny one-shoulder shrug, the corner of her mouth lifting. She says: “Sorry, but I’m not.”

2:60–3:30 — Cut to extreme macro of one eye and brow, slow sideways drift; a slow blink, then her gaze locks into the lens. Her voice continues over the shot: “My eyes,”

3:30–4:00 — Cut to extreme macro of her freckled cheek; a strand of black hair drifts across it in the breeze; gold earring at the frame edge. Her voice continues: “my skin,”

4:00–4:70 — Cut to close-up at chest height; her right hand with the gold ring smooths the shawl collar of the cobalt coat, showing the wool weave. Her voice continues: “my clothes,”

4:70–5:50 — Cut to top-down drone view rising straight up over the geometric boxwood parterre; she is a small cobalt figure walking the central gravel path. Her voice continues: “this scene.”

5:50–6:40 — Cut to medium shot from the waist up, locked off; she turns back to the lens, holds it for a beat, a small slow nod. She says: “All prompted.”

6:40–10:00 — Cut to wide full-length shot down the central gravel path; she walks toward camera with an easy, confident stride, coat swinging, ending knee-up, eyes on the lens. She says while walking: “Comment prompt, and we’ll send the exact prompt we used.”

EXACT DIALOGUE, NO ADDITIONS. Every line is spoken once, by her, in the same low, calm, slightly amused voice throughout, including the lines heard over the close-ups. Natural pace; never rushed.

SOUND: garden ambience, birdsong, light breeze, gravel footsteps, fabric movement. No music. No subtitles, no on-screen text, no other people, no wardrobe changes.

How to adapt it to your character

• Identity line: in the References paragraph, replace the face, hair and lip details with your character’s, and list every piece of the outfit and jewellery so they stay consistent across cuts.
• Scene line: describe your Image 2 location in the same order: place, key features, light, palette.
• Close-ups: keep the macro shots (eye, skin, hand on fabric). They are what sells the realism, so pick details your outfit and location can show off.
• Dialogue: keep it short. Every line must appear in the prompt, including voiceover lines over close-ups, or Seedance may skip them.

Things worth knowing

• Separate the references’ jobs. Say clearly what each image controls. Without that, the model may copy the reference image’s framing instead of creating new shots.
• Name the attitude. Without it, AI characters default to eager smiling. Describing the energy you do not want (no big smiles, no eyebrow raising) helps as much as describing what you do want.
• Short shots can blur together. Shots under a second sometimes merge or get skipped. If that happens, generate the final walking shot (6:40–10:00) on its own and join the two clips in your editor.
• Ban the AI look in words. “No HDR, no glossy commercial finish” plus film grain and visible pores are what push the result toward real footage.

FAQ

How do I keep my AI character consistent across shots?

Use one character reference image for every generation, and list the face, hair, outfit and jewellery details in the prompt so the model has them in writing as well as in the image.

Can Seedance make the character talk?

Yes. Write each line in the prompt with the shot it belongs to, and state that every line is spoken once with no additions.

Why one 10-second generation instead of separate clips?

One generation keeps the light, grade and character consistent across cuts and saves editing time. Split it only if shots blur together.

Does this work with other video models?

The prompt is written for Seedance 2.5. Other video models that accept reference images and timed shot lists can follow it, but test one run before you build a full piece around them.