Gangnam Style AI Video Prompts: Three Friends, One Selfie and Pets

These are Gangnam Style AI’s own reference-editing directions, not a disclosure of gangnamstyle.io’s private prompt. The template uses Seedance 2.5 with a supplied trio clip. Preview the scene in the Gangnam Style studio before reading the time map: the sequence is a group shot, three solos and a final group shot.

The prompts below use @Video1 for that reference clip. It supplies the background, camera, framing, actions and shot sequence. The uploaded images supply the new performers’ appearance. Merely pasting the words into a text-to-video tool does not attach those assets or reproduce the reference scene.

Friends and pet modes use three separate photos. The selfie mode uses only @Image1, repeated across all three roles. In another editor, attach the assets in the matching order and check its supported reference syntax and editing mode before submitting.

The instructions identify performers by screen position and time, rather than by the original cast’s clothes. This matters when a solo fills the frame or someone returns after a cut. Left and right always mean the viewer’s left and right.

These time codes belong to this particular reference clip. If you use another video, watch it and rewrite the shot boundaries and role assignments. For the selfie version, every @Image2 and @Image3 assignment becomes @Image1.

Use this version with one photo per person. Upload the center performer first, the viewer’s right performer second and the viewer’s left performer third. Each photo is the reference for that person’s face, hair and clothes throughout the clip. Select and copy the complete text below when using a compatible editor.

Edit @Video1. Replace ALL THREE original foreground performers from the first frame through the final frame. Assign identities by screen position and time, not by the original performers' clothes or hairstyles. Left and right mean the viewer's left and right. Opening group shot, 0-2 seconds: screen-left is @Image3, center is @Image1, screen-right is @Image2. Solo close-ups: 2-4.9 seconds is @Image1; 4.9-8.4 seconds is @Image2; 8.4-12.3 seconds is @Image3. Closing group shot, 12.3 seconds to the final frame: screen-left is @Image3, center/front is @Image1, screen-right is @Image2. Keep these assignments through camera transitions and whenever a performer re-enters the frame. Each assigned image is the sole source of that person's face, hair and clothes. Keep the original background, camera movement, framing, timing, expressions and actions unchanged. Remove all audio tracks. No audible output, added captions or watermarks.

For Gangnam Style AI, select Gangnam Style trio and put each image into its labeled slot; the prompt is already configured. Use the photo preparation guide to avoid unclear faces, group photos and unsuitable crops.

Use one clear adult portrait as @Image1. This version asks for three separate copies of the same person, so all group positions and every solo use the same identity. It also explicitly preserves that person’s face, hair and clothes across the copies.

Edit @Video1. Replace ALL THREE original foreground performers from the first frame through the final frame. Assign identities by screen position and time, not by the original performers' clothes or hairstyles. Left and right mean the viewer's left and right. Opening group shot, 0-2 seconds: screen-left is @Image1, center is @Image1, screen-right is @Image1. Solo close-ups: 2-4.9 seconds is @Image1; 4.9-8.4 seconds is @Image1; 8.4-12.3 seconds is @Image1. Closing group shot, 12.3 seconds to the final frame: screen-left is @Image1, center/front is @Image1, screen-right is @Image1. Keep these assignments through camera transitions and whenever a performer re-enters the frame. Create three separate copies of the same adult from @Image1. Each copy keeps the uploaded face, hair and clothes throughout. Keep the original background, camera movement, framing, timing, expressions and actions unchanged. Remove all audio tracks. No audible output, added captions or watermarks.

In Gangnam Style AI, choose Three of you and upload one selfie. You do not need to upload the same file three times. The single-selfie guide explains the photo selection and review steps.

Attach one individual animal photo to each image reference, using the same Middle, Right and Left order. The pet version preserves species, face, fur markings and size. It asks the model to interpret human expressions and hand gestures as head and front-paw movements with coherent animal anatomy.

Edit @Video1. Replace ALL THREE original foreground performers from the first frame through the final frame. Assign identities by screen position and time, not by the original performers' clothes or hairstyles. Left and right mean the viewer's left and right. Opening group shot, 0-2 seconds: screen-left is @Image3, center is @Image1, screen-right is @Image2. Solo close-ups: 2-4.9 seconds is @Image1; 4.9-8.4 seconds is @Image2; 8.4-12.3 seconds is @Image3. Closing group shot, 12.3 seconds to the final frame: screen-left is @Image3, center/front is @Image1, screen-right is @Image2. Keep these assignments through camera transitions and whenever a performer re-enters the frame. Each assigned image is the sole source of that pet's species, face, fur markings and size. Translate expressions and hand gestures into head and front-paw movements with coherent animal anatomy; do not turn pets into people. Keep the original background, camera movement, framing, timing, expressions and actions unchanged. Remove all audio tracks. No audible output, added captions or watermarks.

Choose Pet trio in Gangnam Style AI to apply this version automatically. Review paws, limbs and species consistency in the result; the instruction expresses the intended behavior but does not guarantee perfect anatomy. See the pet video guide for suitable reference photos.

The current Gangnam Style AI template uses Seedance 2.5 reference editing, fixed 16:9 landscape framing and a roughly 15-second reference. You choose 480p or 720p. The reference provides the composition and timing; higher resolution changes output size, not the identity map or the intended shot order.

“Remove all audio tracks” applies to the model’s video-generation stage. Gangnam Style AI requests a silent visual edit and then uses an FFmpeg processing step to merge the reference clip’s soundtrack into the finished video. New completed Gangnam Style AI results therefore include sound. The prompt does not ask the model to recreate the music.

If you copy these prompts into another tool, that tool will not automatically run Gangnam Style AI’s soundtrack step. Add the intended audio separately using its editing controls or your own post-production workflow. The soundtrack guide explains playback and the audio included in this template.

1. Choose the version that matches your cast: three people, one repeated adult selfie or three pets.
2. In a compatible video editor, attach the reference clip and photos. Use its reference-editing workflow and match the asset labels to the prompt.
3. Keep the time map only if you are using this exact reference. For a different clip, update the positions and solo timings after watching it.
4. Generate the visual edit, review the entire clip and add the desired soundtrack separately if your tool does not do that for you.

In the Gangnam Style AI studio, the reference, matching prompt and soundtrack processing are already set up. Choose your mode, photos and quality, review the displayed credits and generate. There is no prompt text box. The step-by-step walkthrough covers the available controls.

• Opening and closing groups: are all three foreground performers replaced in the intended positions?
• Each solo: does the correct person or pet appear, including immediately after a camera transition?
• Appearance: do faces, hair, clothing or fur markings stay recognizable throughout?
• Scene and sound: does the clip retain the reference sequence, and does the downloaded final file play with its soundtrack?

An instruction is a target, not a guarantee for every generated frame. If a solo has the wrong person, inspect the image order and role map. If a face is unclear, improve the source photo. If using another editor, confirm it actually received the video reference. Adding more adjectives does not resolve those input problems.