How to make a Gangnam Style AI video

Every method starts from photos, and the photos decide most of the result. Get these right once and you can reuse them for all three methods.

• One person per photo. Crop group shots so only the person you want is left in the frame.
• Facing the camera. A straight-on face with both eyes visible works far better than a profile or a tilted head, especially for the close-up at the end.
• Waist-up or full body. Each person keeps their own face, hair and clothes in the video, so include the outfit.
• Good, even light. Skip heavy filters, sunglasses and hands over the face.
• A clear original. Use the photo itself, not a blurry screenshot of it. JPG, PNG or WebP up to 10 MB; HEIC works in Safari.
• Only people who agreed. Use photos of people who are happy to be in it. No public figures and no minors.

Decide now who goes Left, Center and Right. Each spot has its own moment: Center opens and closes the clip in close-up, Left sings and gestures alone after the first whip pan, and Right sings and dances alone in front of a glass building.

The generator recreates the original clip with your three people in it: same camera moves, same close-ups, same timing. Each person keeps their own face, hair and clothes, and the original audio stays in.

1. 1Add three photos

   Open the generator on the home page and sign in with Google. Add one photo each for Left, Center and Right: one person per photo, face clearly visible.

   

2. 2Generate the video

   Press Generate video. It costs 300 credits and takes about 10 minutes. You can leave the page; the video is saved in My videos. If a video fails, the credits are refunded automatically.

   

3. 3Download your video

   When it is ready, download the 15-second MP4 and share it with your friends.

   

Time

About 10 minutes

Output

16:9 MP4, 15 seconds, original audio

Cost

300 credits per video

Watermark

None

Many Gangnam Style AI clips on TikTok carry a CapCut template link. A template is an edit someone else built: you add your own photos or Live Photos and CapCut places them into it, so the motion, cuts and timing are decided by whoever made the template.

1. Find a Gangnam Style AI video on TikTok or in CapCut that shows a template link.
2. Tap Try this template to open it in CapCut.
3. Add your photos or Live Photos in the slots the template asks for.
4. Check the result, then export and share.

The catch

• The result is the template creator's edit, and templates differ a lot in how closely they follow the beats.
• Free exports can carry a CapCut watermark or end screen.
• Length, cuts and framing are fixed by the template, so you can't re-time it or change the setting.
• Templates come and go. If one disappears, look for another video with a template link.

If you would rather build it yourself, current video models can get close to the look. Kling, Veo and Sora all take a text prompt plus an optional start image. Starting from a photo of your three people (image-to-video) keeps their faces far closer than text alone.

Start with this prompt, which describes the night terrace of the original clip, and change the description of the three people:

Versions with other settings, such as a stadium, a night rooftop or a castle, plus a negative prompt and settings, are on the Gangnam Style AI prompts page.

Expect randomness. Models often skip a beat, mix up who has which solo moment, or keep the camera still instead of whip-panning. Budget several attempts and keep the best one.

The face doesn't look like the person

Use a sharper, front-facing photo with even light, and crop so the face is larger in the frame. Remove sunglasses and heavy filters.

The wrong person gets the wrong moment

In the generator, the slot decides the moment: Center gets the close-ups at the start and end, Left the first solo, Right the solo in front of the glass building. Put each photo in the slot you want. In a prompt, describe each person by position, for example "on the left, a woman in a red jacket".

The camera doesn't move

This happens with prompts. Describe each camera move in order (close-up on the center person, whip pan to the left person, cut to the right person, back to the center close-up) and add "static tripod shot" to the negative prompt.

The moves are out of sync with the music

Templates and prompt videos are not always timed to the song. Line up the start in your editor, or add the sound in TikTok and nudge it. The generator keeps the original audio and timing.

Short version: if you want the beats in the right order with your own three people, use the generator. If you want something quick and free and don't mind whatever the template does, use CapCut. If you enjoy tinkering, write a prompt.