Gas Station Dance AI prompt, and how to make it

The prompt to copy

Use it with a model that takes a reference image and a text prompt. Attach one full-body photo of one person as the reference image, set the output to 9:16 and about 11 seconds, and paste this.

Photoreal 9:16 vertical phone video, 11 seconds, filmed handheld at night on a gas station forecourt under a bright canopy, with a fuel pump close by, parked cars behind and a car door standing open. The person from the reference image wears the clothes from the reference image in every shot, and their face, hair and outfit stay the same in every frame. They step away from the open car door, clap, throw both arms up, bend low, reach down with both hands and finish on a flex, one move straight after the other. The camera has a slight handheld shake and never cuts. No subtitles, captions, logos, watermarks or on-screen text, and no logo on the pump sign.

What the prompt asks for, and where it falls short

The moves are the ones in the clip: out of the car, a clap, both arms up, a low bend, a reach down and a flex. A prompt can only describe them, so how closely a model follows the order is up to the model.

SameButMe works differently. The template holds the original clip with its dancer already replaced by an AI character, and your photo replaces that character, so the moves, the camera and the timing stay as they are.

• Keep the clothes line if you want your outfit in the video.
• Keep the no-text line. It asks the model to leave captions and logos out of the frame.

The one-photo way, step by step

No prompt is needed here. The photo is the only input.

1. 1Open the Gas Station Dance template.
2. 2Add one full-body photo of one person, standing and facing the camera.
3. 3Choose 720p for the clip you post or 480p for a test.
4. 4Tick the box that the photo is of you or of an adult who agreed.
5. 5Press Make my video. Create a free account if you're asked to, and top up if you're short of credits; the photo stays on the page.
6. 6Wait about 6 minutes, then download the vertical MP4.

The photo that works

The routine uses the whole body, so a photo from head to shoes gives the most to work with. Behind the scenes your photo becomes a character reference, a close view of the face plus a full-length view, that keeps your hair, glasses and clothes.

• Stand tall with your arms away from your body. The claps and the flex need room.
• Face the camera in even light, with nothing over your face.
• Wear what you want to dance in. The dancer wears whatever you wear in the photo, dress or not, and bright colors show up well against the night.
• If the photo stops at the waist, the legs and shoes are filled in to match the outfit.

720p or 480p?

Both give the same 11-second vertical clip. 720p costs 170 credits and 480p costs 80, and it is usually ready in about 6 minutes.

A 480p test followed by a 720p final costs 250 credits in all.

Posting it to TikTok or Reels

The clip is already 9:16, so it posts to TikTok and Reels as it is. A free 1:1 copy and a silent copy are there too, and none of them costs credits.

The original sound is in the file. Download the silent copy if you want to add a sound inside the app.

Mark the post as AI-generated and keep it out of ads and paid posts: the clip and its sound belong to others.

Is it free?

The prompt is free to copy. Making the video on SameButMe is not: an account is free but starts with no credits, so you buy a plan or a pack first. The cheapest way is 480p, at 80 credits.

See pricing

If something goes wrong

The problems you can run into, and what to do about each.

• The photo is refused. The site declines celebrities, public figures, minors and nudity, so use a photo of yourself or of an adult who agreed.
• The button reads Top up to make it. Your balance is short of the price. Top up and your photo is still there when you come back.
• The moves look cramped. Retake the photo standing tall, with your arms away from your body and your whole body in the shot.
• A render fails. You get the credits back and can run it again.