In this look, three people dance against a petrol pump background at night. To try it in Gemini Omni, keep a reference video ready along with three clear adult photos. A reference video is the clip that shows the movement and camera framing. Getting the exact same dance from only photos and a prompt is not guaranteed.

Keep these ready first
Use three photos and a dance clip of yourself, or of people who have given permission. Each photo should clearly show the face of a different person. Choose a vertical clip: 9:16 means the tall frame of a phone screen. Treat this setup as an AI recreation. You do not need to shoot a dance at a real petrol pump.
According to Google’s official video guide, a personal account needs a Google AI plan and you must be 18+. First check your account’s video access and remaining limit. Do not assume it is free for everyone.
The dance video prompt
A prompt is the instruction you give to the AI. Upload the three photos in this order: left dancer, right dancer, then the person near the car. The text below is a suggested starting point. Check the faces and movement in every result.
🎨 Customize prompt size, colour, glasses
Check the result before you use it. Features, limits and cost depend on your AI app account.
How to try it in Gemini on a computer
1. Sign in at gemini.google.com. Open the sidebar and choose Create video.
2. Use Add video to add your reference dance clip that you have permission to use.
3. Use Add image to add the three photos in the right order. The official guide supports one video and up to five images.
4. Paste the prompt above. If you want a vertical output, keep the reference clip vertical too.
5. Submit and wait for the result. Generation can take a few minutes.
6. Play the video: check for three different faces, hands and feet, clothes and dance continuity.
7. If the result is right, choose Share → Download video below the video. Play the file and check the dimensions and audio. Writing HD in the prompt does not set the actual resolution.
If something goes wrong in the result
If the faces mix, reply: Keep the three identities separate. Reduce the choreography to simple steps with clear spacing. If the hands or feet look wrong, ask for small movements instead of complex spins. Corrections can also count against your account’s generation limits.
If the full sequence of getting out of the car does not fit, first try a simple version where the person standing near the car joins the dance. The exact choreography of the original reel and a perfect identity match are not guaranteed.
These steps are based on Google documentation, checked on 6 October 2026. A TricksFunn demo of this prompt is not available yet.





