Fan Blade Outfit Change: The Top-Down AI Prompt
A free Seedance 2.5 prompt. She lies on a rug, a slow ceiling fan blade sweeps over her, and her outfit is new. Five looks, 20 seconds, five reference images.
Have it running in three steps
- 1
- 2Open ImagineArt
- 3Paste it and generate
What you'll need
30 min- ImagineArt
A woman lies on a Persian rug. The camera is on the ceiling, looking straight down. She picks up a remote and starts the fan. A long blade swings slowly over the frame, and everything behind it is a new outfit.
It happens four times. Five looks, 20 seconds, one shot. The camera never moves and there is no cut.
The blade is the whole trick. It gives the model a hard edge to change things behind, so the change looks planned instead of glitchy. This page gives you the full prompt, our five images, and the one step most people skip.
The whole process
- Pick five looks. Save them from Pinterest or make them.
- Make a character sheet, wearing look one. It sets her face and her first outfit.
- Give the prompt below and your outfit pictures to Claude. Ask it to rewrite the reference lines.
- Open ImagineArt, go to video and pick Seedance 2.5.
- Add the character sheet first, then the four outfits in order.
- Set 20 seconds and 9:16. Paste your prompt and generate.
The fan speed is the edit
This is the part that makes the video work, and it is easy to miss.
The fan has exactly two blades, and it turns slowly. One full turn takes about eight seconds. Two blades means a blade enters the frame every four seconds, and each one takes about a second and a half to cross.
Four seconds per blade, four blades across 20 seconds, four outfit changes. The fan is not decoration. It is the clock the whole edit runs on.
That is also why the prompt says one blade crossing equals exactly one outfit change. Never two changes in one crossing, and never a change without one.
What happens in 20 seconds
| Time | What happens | Look |
|---|---|---|
| 0–4s | The fan is still. She lies on her side, rolls onto her back, raises the remote and starts it | Look 1 |
| 4–8s | A blade crosses. She settles into an easy lounging pose and smiles at the change | Look 2 |
| 8–12s | The opposite blade crosses. Her pose goes elegant, one arm above her head | Look 3 |
| 12–16s | Another crossing. Cooler and edgier, one elbow out, one knee shifted | Look 4 |
| 16–20s | The last crossing. She softens, one leg crossed, hand near her collarbone, and holds | Look 5 |
The five images, in order
The numbers in the prompt point at these exact slots. Upload them in this order or the model dresses her in the wrong thing.
| Slot | What it is | What it controls |
|---|---|---|
@Image1 | Character sheet | Her face, her hair, and look one |
@Image2 | Outfit picture | Look two |
@Image3 | Outfit picture | Look three |
@Image4 | Outfit picture | Look four |
@Image5 | Outfit picture | Look five |
Make your character sheet, wearing look one
The camera is far away and looking straight down. Her face is small in the frame. So the model needs more than a selfie to keep her looking like herself.
A character sheet gives it everything at once: the face close up, the body from the front, the side and the back, and small tiles for the eyes, the lips, the jewellery and the shoes.
One thing changed here from the older version of this prompt. The character sheet now does two jobs: it sets her identity and it is look one. She starts the video in whatever the sheet is wearing.
So do not put a throwaway outfit on it. Dress the sheet in a look you actually want on screen, with the shoes visible. Ours is a green crop top, black wide-leg jeans and white trainers.
Pick the other four looks
Search Pinterest for “outfit flat lay”, “outfit collage” or the style you want. Save four complete looks.
Each one needs shoes. The camera sees her from head to feet, so her feet are on screen the whole time. If the shoes are missing, the model guesses.
Put the strongest look last. The video ends on look five and holds there, so that is the one people remember.
Notice each picture has a different background: a wall, white, white, grey. That is fine, because the prompt names the background of each one and tells the model to drop it.
Have Claude rewrite the reference lines
This is the step that decides whether your video works.
Open Claude. Attach your character sheet and your four outfit pictures. Paste the whole prompt from step four. Then send this:
Here is a video prompt, my character sheet and my four outfit pictures.
Rewrite only the five lines in the REFERENCES block, so each one describes my picture instead of the one written there now.
For each look, name the top, the bottom, the shoes, the bag and the jewellery, with colour and fabric. Then say what to leave out of that specific picture: its background, the hanger, the flat-lay layout, any text or logo, anything that is not clothing.
Change nothing else. Keep the camera, the fan, the two blades, the fan speed, the rug, the sofa, the lighting, the stage timings and every rule exactly as they are.
Give me back the full prompt in one block, ready to copy.
Two things to check in what Claude gives you.
Shoes in all five looks. Her feet are on screen for 20 seconds. A look with no shoes named comes back barefoot or in something random.
The right background named for each picture. The prompt says “do not use the grey background” for one image and “the hanger, the wall” for another. If Claude copies the wrong exclusion onto your picture, the background leaks into the room.
Set it up and generate
Open ImagineArt+200 free credits, go to video and pick Seedance 2.5.
Add the character sheet first. Then the four outfits in order. Set the length to 20 seconds and the shape to 9:16.
Paste your rewritten prompt. Do not shorten it and do not add to it.
[LOOK — LOCKED, APPLIES TO EVERY STAGE] Digital cinema capture, wide-angle prime, slight barrel curvature at the frame edges, focus deep enough to hold her from head to feet, fine low-noise grain, warm highlights rolling off softly, slightly lifted blacks, medium contrast. Grade: warm amber hero tone, olive and burgundy secondary field, shadows biased warm brown. Camera operation: locked-off ceiling mount, straight top-down overhead angle, vertical 9:16, no pan, no push-in, no handheld drift — framing identical in every stage.
[REFERENCES] All five images show one single woman. The video contains exactly one person from start to finish. @Image1 defines her identity and her Look 1 clothing — face, skin, hair and full outfit exactly as in the reference image. Do not use the white studio background, the standing poses, the grid layout, or any duplicated figure from it. @Image2 defines Look 2 clothing — exactly as in the reference image. Do not use the hanger, the wall, the flat-lay arrangement or the background. @Image3 defines Look 3 clothing and accessories — exactly as in the reference image. Do not use the white background or the catalogue layout. @Image4 defines Look 4 clothing — exactly as in the reference image. Do not use the white background or the catalogue layout. @Image5 defines Look 5 clothing and accessories — exactly as in the reference image. Do not use the grey background or the catalogue layout.
[SCENE — LOCKED] A classic Persian rug, cream field with blue and red floral medallions and fringed ends, laid over dark herringbone parquet. A burgundy velvet sofa runs along the bottom edge of frame. The woman fills most of the vertical frame, head toward the top edge, feet toward the bottom edge. Lighting: warm editorial light with a soft diagonal shaft from the upper left. Her face, collarbones, arms and legs keep dimensional shape and warm highlights; her hair catches warm rim light. Never flat, never grey.
[FAN — LOCKED] One single ceiling fan with exactly two long blades set opposite each other. Its round central hub sits at the LEFT EDGE of frame with only half the hub visible, the other half cut off by the frame edge. The camera is mounted close to it, so the blades read large and each blade is long enough to sweep across the full width of the frame. Speed: slow and lazy. One full rotation takes about eight seconds, so a blade enters the frame roughly every four seconds and takes about one and a half seconds to cross it, with a clear empty gap between one blade and the next. The rotation is smooth and constant, always clockwise, never reversing, never speeding up, never stuttering.
[TRANSITION RULE] One blade crossing = exactly one outfit change. Never more than one change per crossing, and never a change without a crossing. As a blade sweeps clockwise across the frame, its leading edge acts as a soft-edged wipe line: everything the edge has already passed shows the new look, everything it has not yet reached still shows the previous look. The change is fully complete the moment that blade’s trailing edge leaves the frame. Through the wipe her face, hair, body, pose, position on the rug, the rug, the parquet, the sofa, the light and the framing stay identical. Only the clothing changes. No cut, no flash, no glitch, no morph.
[STAGE 1 — 0-4s — LOOK 1] Initial state: the fan is completely still, both blades motionless. She lies on her left side on the rug, knees softly stacked, wearing Look 1 (@Image1), a slim remote already held in her hand and clearly visible in frame. Primary event: she rolls slowly from her side onto her back into a relaxed asymmetrical pose, raises the remote toward the fan at the left edge of frame — the remote stays fully visible in her hand — and presses the button. The blades start from a standstill and reach their slow constant clockwise rotation within about a second. End state: she lies on her back, one knee slightly bent, remote resting on the rug beside her, still in Look 1, looking up into the lens with a soft playful smile. The fan turns slowly and steadily.
[STAGE 2 — 4-8s — LOOK 2] Continue from the previous stage: same woman, same face and hair, same spot on the rug, same lighting, same framing, fan turning slowly clockwise. Primary event: the next blade crosses the frame and wipes her outfit from Look 1 to Look 2 (@Image2). She settles into a warm, easy lounging pose — small shoulder shift, one hand moving softly near her hair — and reacts to the change with a delighted smile. End state: she lies on her back in Look 2, pose settled, expression pleased.
[STAGE 3 — 8-12s — LOOK 3] Continue from the previous stage: same identity, same position, same light, same framing. Primary event: the opposite blade crosses and wipes her outfit from Look 2 to Look 3 (@Image3). She shifts into a more elegant, feminine lounging pose, one arm relaxed above her head, chin tilting slightly, a soft polished smile. End state: she lies on her back in Look 3, pose settled, looking into the lens.
[STAGE 4 — 12-16s — LOOK 4] Continue from the previous stage: same identity, same position, same light, same framing. Primary event: the next blade crosses and wipes her outfit from Look 3 to Look 4 (@Image4). Her pose turns cooler and edgier — one elbow angled outward, one knee shifted for shape — with a playful confident smile. End state: she lies on her back in Look 4, pose settled, holding eye contact with the lens.
[STAGE 5 — 16-20s — LOOK 5] Continue from the previous stage: same identity, same position, same light, same framing. Primary event: the final blade crosses and wipes her outfit from Look 4 to Look 5 (@Image5). She softens into a light, feminine pose, one leg crossing gently over the other, one hand near her collarbone. End state: she holds the final pose in Look 5, glances into the lens with a pleased confident expression, and relaxes into a still stylish hold while the fan keeps turning slowly clockwise.
[MAINTAIN CONSISTENCY] Exactly one woman throughout, identity and hair unchanged in every stage. One single fan, two blades, hub always half-visible at the left edge, always clockwise, always the same slow speed. Same rug, parquet, sofa, diagonal light and top-down framing in every stage. Five looks in fixed order: Look 1 to Look 2 to Look 3 to Look 4 to Look 5, one change per blade crossing.
[PERFORMANCE] She stays alive and expressive throughout: natural blinking, subtle breathing, small weight shifts, soft hand movement, head tilts, genuine delight at each change. Asymmetrical editorial lying poses, never symmetrical, never stiff.
[AUDIO] <the low steady hum of a ceiling fan and soft air moving past the blades> <a single soft plastic click as she presses the remote> Quiet room tone. No music, no voice.
[NEGATIVE] No corpse-like stillness. No mannequin stiffness. No flat dead lighting. No symmetrical lying pose. No face distortion or morphing. No body warping. No second person. No duplicate fan. No fan reversing, accelerating or stuttering. No jump cuts. No glitch or flash effects. No outfit change without a blade crossing. No overlay text, captions, subtitles or watermarks.
Why this prompt works
Two blades, and a speed you can count. Eight seconds a turn, a blade every four seconds, a second and a half to cross. The model has numbers instead of a mood. That is what keeps five changes evenly spaced across 20 seconds.
The hub is cut in half at the left edge. The camera sits close to the fan, so the blades are huge and each one is long enough to cross the whole frame. A fan in the middle of the frame gives you short blades and no wipe.
The blade edge is named as the divider. New look behind the leading edge, old look in front of it, and the change is finished when the trailing edge leaves. That is a rule the model can follow. “Change her outfit with the fan” is not.
Each stage says continue, not start. Every stage after the first opens with “continue from the previous stage: same identity, same position, same light, same framing”. Without that, the model treats each stage as a fresh shot and her position jumps.
The fan starts from a standstill. She is holding the remote before she uses it, and the blades are motionless until she presses it. The video earns the fan instead of just having one.
She is told to be alive. Top-down poses go stiff fast. The prompt asks for blinking, breathing, weight shifts, and real delight at each change. Without those lines she lies there like a photo.
Where it goes wrong
| What you got | The fix |
|---|---|
| The outfits are a mix of yours and ours | You did not rewrite the REFERENCES block. Go back to step three |
| Two outfits change in one blade pass | Keep the TRANSITION RULE block. It is there for this |
| The outfit changes with no blade on screen | Same block. The last NEGATIVE line covers it too |
| The fan spins fast and the changes pile up | Keep the Speed lines in the FAN block. Eight seconds a turn is the point |
| The fan has three or four blades | The prompt must say exactly two, set opposite each other |
| She starts in the wrong clothes | Look one comes from the character sheet, not from an outfit picture |
| She is barefoot, or wearing shoes you never picked | Name the shoes in every look, with colour and shape |
| A white or grey box appears in the room | The background of that picture is not excluded. Name it in its reference line |
| She jumps to a new spot between looks | Keep the “continue from the previous stage” line at the top of each stage |
| Her face does not look like you | Make a cleaner character sheet with a bigger face and clear detail tiles |
Ready? Open ImagineArt+200 free credits, add your five images and paste your prompt.
Why this workflow runs here
Everything in one place
Free to start
Minutes, not days
Make yours in the next 30 min
You have the whole prompt and every setting. The only thing left is an account, and that takes a minute.
- 200 free credits on sign-up
- Prompts ready to paste
- Upgrade only if you want more
Referral link — if you subscribe I earn a commission, at no extra cost to you. More on that.
Related prompts
Effect Free Fit Check Hit by a Car: The AI Video Prompt
A free Seedance 2.5 prompt. She films a street fit check on her phone, shows the outfit from every side, and a car takes her out of frame. 12 seconds, one reference image.
Effect Free Gangster Goose: The AI Video Prompt
A free Nano Banana Pro and Seedance 2.5 prompt. A goose in sunglasses walks into the street with a pistol, fires three shots at the sky, and three cats let their mouse go and run. 15 seconds, filmed like a bystander's phone video.
Effect Free Phone Stuck to the Train AI Prompt: Swap Yourself In
Free AI prompt for the phone stuck to the train trend. Keep every move from the subway video and put yourself in it. Character sheet and video prompts included.