Gangnam Style AI prompt (copy and paste) for the horse dance video
The copy-ready Gangnam Style AI prompt: the crossed-wrist trot and the empty-handed lasso on a sunset rooftop. The elevator version, with fixes.
7 lettura minima · Aggiornato
Diventa il protagonista di Gangnam Style con una foto
Crea il tuoThe Gangnam Style AI video prompt
Here’s the copy-ready prompt for the Gangnam Style trend, written the way we write ours for the rooftop: you stand head to toe on a city rooftop at sunset, do the crossed-wrist trot, then swing the lasso with one empty hand to the end, with the camera still. Attach your photo as the reference image and paste it in.
Person, sunset rooftop
A vertical 9:16 phone video, exactly 10 seconds, photoreal live action in full natural colour, one continuous take from a phone on a tripod at chest height, about three metres away. The dancer is seen head to toe the whole time, filling most of the frame height. The camera does not move. No cuts. No text, no subtitles, no labels, no logos, no watermarks. Setting: a city rooftop terrace at golden hour: warm low sun from one side, a soft orange and pink sky, glass towers and apartment blocks in the distance, a low concrete parapet and a few potted plants behind the dancer, a string of warm bulbs overhead, a pale concrete floor. Nobody else is there. Nobody resembles a famous person. WHO: one dancer, the person from @Image1, in exactly the clothes they wear in @Image1, alone, empty-handed, facing the camera the whole time. ACTION: the Gangnam Style horse dance, on the beat, dancing in place. [00:00-00.5] the video opens with them already standing in the middle of the frame, knees bent, wrists crossed low in front of the belly, hands empty, a confident grin. [00.5-04.25] the trot: the wrists stay crossed in front of the belly, hands empty, and they bounce from foot to foot on every beat, knees lifting, shoulders loose. [04.25-10.0] the lasso: one empty hand circles above the head, nothing in it, the other wrist stays at the belly, the feet keep bouncing, all the way to the last frame. IDENTITY: the dancer is exactly the person from @Image1 in every frame: the same face, age, skin tone, the same hairstyle, hair colour and hair length, which stay the same through every bounce, the same clothes and the EXACT body weight and build, never slimmer, heavier or younger. 真人实拍、五官清晰、面部稳定、不扭曲、不变形、人体结构正常、同一角色、服装一致、发型不变. Avoid: walking off, turning away from the camera, a second person, a horse, a rope, reins or anything held in the hands, sunglasses or a suit unless they are in @Image1, a moving camera, close-ups, face changes, the dancer talking or singing, captions, text, logos, watermarks, cuts, extra arms, extra fingers, distorted hands, morphing faces.
A prompt gives the tool a plan, not a promise; two or three runs is normal. Want it without a prompt, with the song in? The Gangnam Style page makes it from one photo.
How to use it
- Open an AI video tool’s image-to-video or reference-image mode: Kling, Higgsfield, Runway, Veo or CapCut’s AI video all have one.
- Attach one photo of you facing the camera, head to toe if you can.
- Paste the prompt. If your tool doesn’t use @Image1 for the first photo, swap it for “the person in my reference photo”.
- Set 9:16 and 10 seconds, or the nearest length. Turn off the tool’s own music if it adds some.
- Generate, watch it against the frames below, and use the fix-it table for anything off.
What each part of the prompt does
The first line
One continuous take from a phone on a tripod at chest height, about three metres away, the dancer head to toe and the camera still. The horse dance is a full-body dance, so a tool that zooms in or cuts away loses the feet, and the feet are the trot.
Setting
Golden-hour light, an orange and pink sky, towers in the distance, a low parapet, a few plants and a string of warm bulbs overhead: the rooftop described piece by piece so the tool builds it and doesn’t guess. “Nobody else is there” keeps the frame to you.
WHO
You, in the clothes from your photo, alone and empty-handed, facing the camera. “Empty-handed” is there for the lasso: without it, tools hand you a real rope.
ACTION
Three timed lines: the pose until 00.5, the trot until 04.25, then the lasso to the end. We render the video silent and lay the song over it from the start; its beat kicks in just under a second in, right as the trot gets going, so moves on time mean moves on the beat. “Dancing in place” stops the tool from walking you around.
No left or right
The prompt never says which hand points up or swings the lasso. Tools mix up left and right anyway, and either hand looks right, so leaving it open saves a retry.
IDENTITY
Ten seconds of bouncing is where faces, hair and builds drift. So this line says the same thing more than one way on purpose: the same face, hair and clothes in every frame, hair that stays the same through every bounce, and the exact build, never slimmer, heavier or younger. The line in Chinese repeats it for models that follow Chinese more closely.
Avoid
The things video tools tend to get wrong in this clip, named so they’re less likely: walking off, turning away, a real horse or rope, the music video’s sunglasses and suit when you’re not wearing them, a moving camera, close-ups, singing along, text and the usual hand problems.
There’s no sound line. We make the video silent and lay the song over the whole 10 seconds afterwards, so do the same: add the song when you post.
Shot by shot: what each line does
- The opening: “already standing in the middle of the frame” means the video opens on the pose, with no walk-in to waste the first second.
- The reins: “wrists crossed low in front of the belly” puts the hands low, where the trot needs them, not up at the chest.
- The trot: “knees lifting, shoulders loose” is the bounce; without it the trot turns into a shuffle.
- The lasso: “the other wrist stays at the belly” keeps it the horse dance, not a wave.
- To the end: “all the way to the last frame” keeps the lasso going, so the clip doesn’t stop dead before the song does.
Variants
The elevator
Replace the first line and the Setting line with the ones below, and put the doors at the start of the Avoid line, keeping the rest of it from “turning away from the camera” on. WHO, ACTION and IDENTITY stay exactly as they are. The camera stands in the lobby looking into the open car, so the doorway frames you.
Elevator lines
A vertical 9:16 phone video, exactly 10 seconds, photoreal live action in full natural colour, one continuous take from a phone on a tripod at chest height in an elevator lobby, about three metres from the elevator, looking straight into the car through its wide-open doors. The dancer is seen head to toe the whole time, filling most of the frame height. The camera does not move. No cuts. No text, no subtitles, no labels, no logos, no watermarks. Setting: a modern office elevator, its two brushed-steel doors slid fully open and staying open the whole time. Inside the car: brushed-steel walls, a handrail at hip height, a row of warm ceiling lights and a pale stone floor; the lobby floor in front of it is polished grey tile. The dancer stands in the middle of the car, framed by the open doorway. Nobody else is there. No floor numbers, buttons or signs are readable. Nobody resembles a famous person. Avoid (start of the line): the doors closing, the dancer stepping out of the car, turning away from the camera, …
“Staying open the whole time” and “the doors closing” in Avoid are there because tools like to close elevator doors. “No floor numbers, buttons or signs are readable” keeps made-up text off the walls.
Somewhere else
Want a dance studio, a backyard party or your own living room? Replace the Setting line with your place, described piece by piece, and keep “Nobody else is there”. Keep the camera, the dance and IDENTITY the same.
Shorter clip
If your tool only makes 5 seconds, keep the pose until 00.5, the trot until 02.5 and the lasso to the end. The trot and the lasso are the parts people watch for.
Fix it: symptom and change
| What went wrong | Change in the prompt |
|---|---|
| A real rope in the lasso hand | Write “one empty hand, nothing in it” for the lasso, never “as if swinging a rope”, and put “a rope” in Avoid; retry, since it still slips now and then |
| A horse shows up | Keep “a horse” in Avoid and never add the word “riding” |
| The camera zooms in or moves | Keep “The camera does not move” in the first line and “a moving camera, close-ups” in Avoid |
| Your feet are cut off | Keep “seen head to toe the whole time” and use a head-to-toe photo |
| You walk around or turn away | Keep “dancing in place” and “facing the camera the whole time”; keep “walking off, turning away from the camera” in Avoid |
| You look slimmer or heavier | Keep “the EXACT body weight and build, never slimmer, heavier or younger” in the IDENTITY line |
| Your hair changes mid-bounce | Keep “which stay the same through every bounce” in IDENTITY |
| Sunglasses or a suit appear | Keep “sunglasses or a suit unless they are in @Image1” in Avoid and use a photo without them |
| The elevator doors close | Keep “staying open the whole time” in the Setting line and “the doors closing” in Avoid |
| You sing or mouth the words | Keep “the dancer talking or singing” in Avoid |
Tips for free tools
- Most free tiers add a watermark. Check before you plan to post.
- Free tiers often make shorter clips. Use the shorter-clip variant above rather than cramming every move into 5 seconds.
- Budget for retries. Check the lasso first, for a rope, then your build and your feet.
- Don’t cut the camera or Avoid lines to save space; they’re the fix.
- Try the same prompt in two tools and keep the one that looks most like you.
The photo you attach
- Face toward the camera, sharp and well lit, nothing covering it.
- Head to toe is best: you’re seen head to toe for all 10 seconds, in the clothes from your photo.
- No sunglasses or hats that hide your face.
- Just you, no group photos, and only your own photo or an adult’s who has said yes. No celebrities and no one under 18.
Examples are in the photo guide.
What a prompt can’t do
- Add the song. You pick the Gangnam Style sound when you post.
- Promise an empty lasso hand every time. Tools still add a rope now and then, so check each run.
- Promise the trot right on the beat every time, or your exact build.
- Promise the same take twice. Every run is a new rooftop and a new dance.
The no-prompt route: add one photo on the Gangnam Style page and get the finished video with the song in, on the rooftop or in the elevator. If it fails, the credits come back automatically, and prices are on Pricing. Comparing options? See CapCut vs AI vs one photo.
NowStarring is an independent fan tool, not linked to PSY, his labels, the song’s rights holders or the creators who started the trend, and the song belongs to its rights holders.
Domande frequenti
What is the prompt for the Gangnam Style trend?
Describe a vertical 10-second take from a phone on a tripod: you stand head to toe on a sunset rooftop, trot with your wrists crossed on the reins, then swing a lasso over your head with one empty hand to the end. The full copy-ready prompt is at the top of this guide.
What does @Image1 mean?
It’s the photo you attach first. If your tool doesn’t name reference images that way, replace each @Image1 with “the person in my reference photo”.
Why is there a line in Chinese?
It repeats the identity rules: real footage, clear and steady features, no warping, the same person, the same clothes and the same hairstyle. Some video models follow it closely. If your tool ignores it, it does no harm, and you can delete it.
Why is there a rope in my video?
Tools read “lasso” and draw one. Keep “empty-handed” in the WHO line, “one empty hand, nothing in it” in the lasso line and “a rope, reins or anything held in the hands” in Avoid. Never write “as if swinging a rope”: tools take the rope literally. It still slips in now and then, so check each run.
Can I do the elevator version?
Yes. Swap the first line and the Setting line for the elevator ones in Variants, and add the doors to the start of Avoid. The dance lines stay the same.
Will the prompt add the song?
No. There’s no sound line in it: we make the video silent and lay the song over it afterwards, with the trot on the beat. Pick the Gangnam Style sound on TikTok or Reels when you post.
