Gangnam Style AI Prompt: Copy It for Seedance and Other Video Models
The exact prompt this site uses for the Gangnam Style AI trio video, plus the one-person and pet versions and fixes when the result is off.
Published 2026-10-22 · Updated 2026-10-22
The short answer
To make the Gangnam Style AI trio video yourself, give a reference-video model a trend clip (at most 15 seconds) as reference video 1, one photo per friend as reference images 1 to 3, and a prompt that says to re-shoot the clip shot for shot and which photo replaces which person. The prompt below is the one this page uses; paste it as is and add the song afterwards.
What's new on this page
- 2026-10-22 — Guide published.
Before you paste the prompt
Use a model that accepts a reference video and several reference images in the same request, such as Seedance 2.0. Upload the clip first and the photos in order, so "reference image 1" is the middle friend, 2 the right and 3 the left. Set the output to 15 seconds, 16:9. If your clip has different people or clothes, change the role descriptions to match it.
Gangnam Style AI trio prompt
Recreate reference video 1 as a new 16:9 clip: same night city street, lights, framing, handheld selfie-camera moves and timing. Use reference video 1 only for motion, framing and setting; never copy any face, hair or body from it. Replace its three people completely: the middle woman (first solo) becomes the person in reference image 1; the right woman in white (second solo) becomes the person in reference image 2; the left woman in black (third solo) becomes the person in reference image 3. Each face must match its reference image exactly: face shape, eyes, nose, mouth, skin tone, age, gender, hairstyle, and the same clothes. In every close-up and group shot the faces are the ones from the reference images; no face from reference video 1 appears anywhere. Keep the moves, gestures and lip-sync, same order. Realistic night phone footage. No text, captions, logos or watermark.
Variations
Change who replaces the three people.
- One person three times: "All three people are replaced by the same person, the one in reference image 1, appearing three times side by side like identical triplets."
- Pets and toys: "…replaced, each one entirely, by the animals or plush toys in the reference images, standing upright and moving like people."
If the result is off
Each render is new, so results vary. These fixes help.
- Two friends swap places: describe each original person by position and clothing, as in the prompt.
- A face is not right: use sharper, front-facing photos without sunglasses.
- The tool refuses the photos: some models reject real people; try another model or other photos.
- No sound: most models drop the reference audio; add Gangnam Style back in an editor.
Questions
Does the prompt work in Seedance 2.0 Mini?+
Yes, this page runs it on Seedance 2.0 Mini at 720p.
Does it work without a reference video?+
Not well. Without the clip, models invent their own shots and rarely keep the solo-by-solo timing.
Keep going
Gangnam Style AI
Gangnam Style AI Video Generator
The night-street trio trend, not the horse dance: three photos, or just your selfie, become each of you singing Oppa Gangnam Style to a selfie camera in turn. 15 s, 16:9, 720p, song under it. No prompt, no app, no editing.
Make my videoHow to Make the Gangnam Style AI Trio Video With Your Friends
Three ways to make the Gangnam Style AI trend video: three photos on this page, by hand in a reference-video model, or a face-swap template.
Updated 2026-10-22
What Is the Gangnam Style AI Trend? The Night-Street Trio, Explained
Where the October 2026 Gangnam Style AI trend comes from: a TikTok clip of three friends in Gangnam, the AI remakes that followed, and what people make.
Updated 2026-10-22