Gangnam Style AI Prompt: Template, Tips and Fixes for the Trio Trend

Do you even need a Gangnam Style AI prompt?

Not on this site. The Gangnam Style AI generator here already knows the trend: you upload three photos, one per friend, and it makes the night street, the selfie camera, the turns and the group finish in one 19-second landscape clip with sound. There is no prompt box and nothing to tune.

A Gangnam Style AI prompt matters when you use a general video model, like Google Flow, Gemini, Higgsfield, Kling or a Seedance-based app. Those tools can make almost anything, so you have to tell them exactly which moment you want and who goes where. The rest of this page is for that case.

What the Gangnam Style AI trend looks like

The Gangnam Style AI trend started on TikTok at the end of September 2026 and took off in October: three friends (sometimes two or four) put themselves into a scene inspired by PSY’s 2012 hit "Gangnam Style". In the version most people copy, the three start squeezed into one selfie on a city street at night. Then each one gets a solo close-up, singing straight into the front camera, with a fast whip pan from one friend to the next, and the clip ends with all three back in one frame. The clips are landscape and about 15 to 20 seconds long.

You will also see it called the "Open Gangnam Style" trend, mostly in India and Pakistan. "Open" is how many people hear "Oppa", the first word of the famous line, so "open gangnam style AI" and "Gangnam Style AI" mean the same trend. Other names you will meet are the Gangnam Style trio and the Gangnam Style AI dance.

The five things your Gangnam Style AI prompt must describe

Group scenes are where AI video models struggle most, so a vague prompt like "my friends doing Gangnam Style" rarely works. Spell these out in order:

• The cast: three people, one from each reference photo, and where each stands in the group shot (middle, left, right). Say the middle one holds the phone.
• The place: a busy city street at night with office towers, street lights and car lights behind, a different spot on the same street for each solo shot.
• The camera: a front-facing phone selfie at arm’s length, a little handheld shake, a whip pan between shots, landscape 16:9.
• The solos: who sings first, second and third, and when. The song runs at 132 BPM (about 1.8 seconds a bar): group selfie until about 1.6 s, solos from about 1.6 (middle), 4.5 (left), 7.7 (right) and 11.7 s (middle again), and the trend’s sound lines up when you add it.
• The finish: from about 15.6 s the other two squeeze back into the frame and all three shout, laugh and wave on every beat, still moving when the clip ends.

A Gangnam Style AI prompt template to copy

Use it in reference-to-video or "ingredients" mode with your three photos as the references, in the order below. Replace the parts in brackets:

"Landscape 16:9 smartphone selfie video, about 19 seconds, at night on a busy city square with lit office towers and street lights behind. Three friends: [Person 1] from image 1, [Person 2] from image 2, [Person 3] from image 3. Keep each person’s exact face, hair, skin tone and [outfit] from their own photo; never swap or blend faces. Steady 132 BPM beat from the first frame, every new shot on the beat. 0–1.6 s: all three squeeze into one selfie, Person 1 in the middle holding the phone, Person 2 on the left, Person 3 on the right, waving and laughing. 1.6–4.5 s: Person 1 alone in a close-up sings to the front camera, then a fast whip pan. 4.5–7.7 s: Person 2 alone, a little farther back, hand on chest, sings with confidence, then a whip pan. 7.7–11.7 s: Person 3 walks away from the camera, turns around and sings close to the lens, then a whip pan. 11.7–15.6 s: Person 1 alone again, pointing at the camera. 15.6–19 s: the other two squeeze back in and all three shout and wave together. Handheld, slight shake, realistic, faces always clear. Original upbeat electronic beat at 132 BPM with wordless chants like "hey" and "oh-oh", no real song, no lyrics, no text on screen."

If your tool only makes shorter clips, render the group selfie, the three solos and the finish as separate clips with the same setting sentence and join them in an editor; the whip pans hide the joins. Keep the photos in the same order every time so nobody swaps places.

What makes a Gangnam Style AI video prompt work

• Tie every person to an image number and a position. "Image 1 in the middle" works far better than "three friends".
• Say "never swap or blend faces" once near the start and once near the end; with three people, faces drift first.
• Name the camera: "front-facing phone selfie, handheld". Without it, models make a smooth film shot that does not look like the trend.
• Give each shot its own sentence with rough seconds. If you only write "they take turns singing", models often keep everyone in frame and let them sing at once.
• Ask for an original beat or no music, and add the sound when you post. Models that sing tend to invent lyrics and mumble.

What to leave out of the prompt

• PSY, the original music video or "make it look like the 2012 video". Ask for your own scene instead; copying the video is a rights problem, and many tools refuse it.
• Celebrities, politicians or anyone who did not agree. Most tools block famous faces, and this site does not allow them.
• The song lyrics. They do not help the motion, and models may try to sing them badly.
• The original TikTok clips as reference videos. Those belong to their creators.

Fixes for common Gangnam Style AI problems

• Two faces look the same: use photos with different light and angles, keep them one person each, and repeat "each person keeps the face from their own image".
• Someone is missing from the group shots: say "exactly three people in the first and last shot" and give each one a position.
• Everyone sings at once: write each solo as its own shot with seconds (1.6–4.5 s, 4.5–7.7 s, 7.7–11.7 s, 11.7–15.6 s) and say only that person is in frame.
• The sound does not line up with the trend’s audio: ask for a steady 132 BPM beat from the first frame, and start the song on the first beat when you add it.
• Wrong clothes: name each outfit in brackets, or say "keep the outfits from the photos". Models like to dress everyone in a blue tuxedo.
• Too cinematic: add "amateur phone selfie, slight shake, filmed by one of the friends".
• Wrong shape: set 16:9 in the tool’s settings, not only in the text.

Or skip the prompt

If you would rather not tune a Gangnam Style AI prompt, the generator above makes the whole trio clip from three photos: the group selfie, a solo for each friend with whip pans in between, and the group finish, 19 seconds, landscape, with an original beat. Each video costs 100 credits with Standard 480p or 220 credits with HD 720p (credit packs start at $9.99 for 100 credits), and you see the cost before you render.

GangnamStyleAIVideo.org is an independent tool. It is not affiliated with, endorsed by or sponsored by PSY, his record labels or TikTok; their names are only used to describe the trend.