Gangnam Style AI Video Prompt
Copy-paste prompts for Google Flow / Gemini, Kling, Seedance and Higgsfield. Each one is built from the same five blocks — subject reference, scene, camera, motion, and the negative list that stops the model from inventing a crowd.
The five blocks of a prompt that works
Framing
“Vertical 9:16 handheld phone footage” — locks the aspect and the phone-shot feel.
Reference
“The person from @Image1” — anchors identity instead of describing a face.
Scene
Night street, neon signs, wet asphalt, headlights. One sentence, concrete nouns.
Motion
Name the body parts: hand on imaginary reins, hips shifting, shoulders rolling, on the beat.
Exclusions
No text, no logos, no extra people in frame, natural skin texture.
Copy-paste prompts
Replace @Image1 with your own reference image in whichever tool you use.
Prompt 01 · Flow / Gemini · Kling · Seedance
Night street trio — the default look
Vertical 9:16 handheld phone footage at night on a busy city street. Three friends dance side by side in the middle of the road, performing the same loose horse-riding-style bounce on the beat: one hand on the imaginary reins, hips shifting, shoulders rolling. The person in the centre is the reference from @Image1. Neon shop signs and passing headlights glow behind them, wet asphalt reflects the light. Slight handheld shake, occasional push-in, natural skin texture, photorealistic, 24fps motion blur. No text, no logos, no watermark.
Prompt 02 · Flow / Gemini · Higgsfield
Rooftop skyline version
Vertical 9:16 handheld footage on a city rooftop at night. The person from @Image1 dances with two friends against a skyline of lit windows and distant traffic. Wide framing, city lights as soft bokeh, cool blue ambient light with warm practical lights behind them. The three dancers stay in sync, loose arms, weight shifting foot to foot, smiling at the camera. Subtle camera drift, photorealistic 8k, natural skin texture, no text overlays, no logos.
Prompt 03 · Kling · Seedance
Wet alley with neon reflections
Vertical 9:16 night footage in a narrow wet alley. The person from @Image1 and two friends dance in the middle of the frame, shoes splashing shallow water. Magenta and cyan neon signs reflect on the pavement, light rain visible in the beams. Handheld camera at chest height, slow push-in, energetic group choreography, cinematic contrast, photorealistic 8k, no text overlays, no logos.
Prompt 04 · Any image-to-video model
Solo dancer (easiest to keep consistent)
Vertical 9:16 handheld night street footage. The person from @Image1 dances alone in the centre of frame, doing the horse-riding bounce on the beat: one hand holding imaginary reins, other arm loose, hips and shoulders moving together. Neon storefronts behind, shallow depth of field, passing headlights creating light streaks. Natural skin texture, photorealistic, 24fps, no text, no logos, no extra people in frame.
Prompt 05 · Kling · Dreamina
Pet or cartoon character
Vertical 9:16 handheld night street footage. The character from @Image1 dances in the centre of frame doing a comedic horse-riding bounce, exaggerated and bouncy, on the beat. Neon signs and blurred city lights behind, shallow depth of field, playful energy, cinematic lighting, consistent character design across frames, no text overlays, no logos.
Prompt 06 · Kling · Runway
Comedy contrast — office worker at night
Vertical 9:16 handheld footage at night. A person in a business suit from @Image1 stands on a neon-lit city street, then abruptly breaks into the horse-riding dance, deadpan expression, perfectly on beat. Two friends in casual clothes join in behind them. Storefront signs glow, wet pavement reflects light, handheld camera shake, photorealistic 8k, comedic timing, no text overlays, no logos.
Negative prompt
extra fingers, deformed hands, duplicate faces, warped text, watermark, logo, subtitles, cartoon shading on a photoreal render, oversaturated skin, plastic skin, crowd of strangers in the foreground, camera cuts, split screen
Which model should you paste it into?
| Model | Mode | What it is good at |
|---|---|---|
| Google Flow / Gemini (Veo) | Image-to-video, 8s clips | Best prompt adherence; stitch two 8s clips for 16s |
| Kling 2.x | Image-to-video, 5–10s | Strongest dance motion; use the motion-strength slider low-to-mid |
| Seedance 2.5 | Image-to-video, up to 15s | Longest single take, good face retention |
| Higgsfield | Preset-driven | Fastest if you don't want to write prompts |
| CapCut / Dreamina templates | Still-image animation | Cannot generate a dancing body — avoid |
Five prompt mistakes
Describing the song instead of the shot
“Gangnam Style music video” pushes the model toward the original video's wardrobe, horse statue and elevator. Describe bodies, light and camera instead; the meme reads without the reference.
Forgetting the vertical framing
If you do not state 9:16, most models default to 16:9 and you lose the phone-shot feel. Say it twice: in the prompt and in the settings.
Asking for three faces without three references
One reference photo can only anchor one identity. Add @Image2 and @Image3 for friends, or the model invents two strangers and swaps them between frames.
Letting the model pick the crowd
“Crowded street” produces flickering extras. Specify “no extra people in frame” unless the crowd is the point.
Re-rolling instead of fixing the photo
If the face drifts on every render, the source photo is the problem — angle, blur or resolution. Five re-rolls cost more than one better selfie.
FAQ
What is the best prompt for a Gangnam Style AI video?+
Describe the shot, not the song: vertical 9:16 handheld night street, three friends dancing the horse-riding bounce on the beat, neon signage behind, wet asphalt, photorealistic, no text or logos. Copy the first prompt on this page and swap the scene details.
Which model handles this prompt best?+
Kling 2.x gives the most convincing dance motion, Seedance 2.5 holds a face longest in a single take, and Google Flow / Gemini follows detailed prompts most literally. Stitch two 8-second Flow clips if you need a longer cut.
How do I keep three faces consistent?+
Supply one reference image per person and name them explicitly in the prompt (@Image1, @Image2, @Image3). Keep the friends' photos in similar light and framing, and render a 5–8 second test before committing to a long take.
Should the prompt mention the original song?+
No. Mentioning it drags in the original video's styling and can trip safety filters. The format is recognisable from the dance and the night-street framing alone.
Can I use these prompts commercially?+
The prompts are free to use. What you generate is subject to the model provider's terms, and the song is still licensed separately — attach it through the platform's sound library rather than embedding it.
Prefer a walkthrough instead of a prompt? Read the five-step guide.
GANGNAM STYLE AI:
PUT YOURSELF IN THE NIGHT DANCE
Add three photos — left, center, right — and all three of you dance on a neon-lit night street. 15-second vertical clip — handheld camera, city glow, ready for TikTok, Reels and Shorts. Add the sound inside the app when you post.
Left
Center
Right
One person per photo, facing the camera. Each face stays separate in the clip.
Photos are used only to render your video; deleted in 24h. Privacy details
Scenes restyle the lighting and crowd around you. Same price.
Built for the night-city dance trend
One photo becomes a vertical clip of you dancing with two friends under neon signage. No editing timeline, no template hunting, no lip-sync pass.
Audio is not included. Add the track yourself from TikTok's licensed sound library when you post.
3 PHOTOSYOUR CLIP
Night-city street dance · vertical MP4
Illustrative demo render. We do not host or output the original song — add the track from TikTok's sound library.