The gas station dance is a 15-second, three-shot clip: one person dances on an empty forecourt, a second films a walking selfie under the pump canopy, and a third steps out of a parked car. It comes from a 2024 road-trip video by Russian blogger Mikhail Litvin with his friends Ravshan and Adel, and reposts usually set it to "On and On" by the Romanian group Akcent. In India and on Instagram it is also called the petrol pump dance. The AI version keeps the original moves, camera and cuts and recasts them with three photos, one per shot.
This page: 3 photos, no prompt, 15 s, 9:16 or 16:9, original music · 17–481 credits per video; default Seedance 2.0 Fast 720p = 77 credits (~$3.85).
Photo 1 · Dancer
One person, face to the camera
Photo 2 · Selfie
One person, face to the camera
Photo 3 · Car exit
One person, face to the camera
JPG, PNG or WebP; 300–6000 px per side, up to 16 megapixels and 30 MB. Photo order is shot order: dancer, selfie, car exit. Use a sharp, front-facing photo of one person in each slot, full body for photos 1 and 3 if you can. Fewer than three people? Upload the same photo into more than one slot. Use only photos you have permission to use.
720p · 9:16 · 15 s MP4 · Original music
This video costs 77 credits. Charged at start; refunded if the job fails.
About $3.85 at the Studio credit rate. Packs start at $14.99; smaller packs cost more per credit.
Signup gift: 3 credits = 0 complete Mini 480p videos. Minimum: 17 credits. The $5 Trial cannot be used here.
Reposts describe the original as unplanned: Litvin and his friends stopped at a filling station at about 4 a.m. during a 2024 road trip and filmed it under the station lights. That is why the first version is a night clip. Most AI recasts moved it to a low golden-hour sun, so the orange sunset look you see everywhere now is a remake habit, not part of the original.
The order never changes: a forecourt dance (0:00–0:05.2), a walking selfie that opens with a whip pan across the sun (0:05.2–0:07.5), then the car exit between two parked SUVs, with a closer cut at about 0:09.5 and a fast pull back to the wide view at about 0:13.2.
It is not the Sunoco discount video. In summer 2026 a Sunoco station in Halfmoon, New York, took money off for customers who danced at the pump, set to Whitney Houston's "I Wanna Dance with Somebody". CNN, Fox and ABC covered it, and it fills most results for "dance at the gas station". That was a local promotion, not this AI trend.
It is not the "Gas Station Love" line dance either. That is a step-by-step line dance taught to a song of the same name, and searches for a gas station line dance or gas station love dance usually mean it. No photos or AI are involved.
ChatGPT and most guides name the same tools for this trend. The difference that matters is whether a tool copies the original clip's motion or tries to rebuild it from text. Checked in October 2026; plans change, so confirm prices on each tool's own site.
| Tool | What you give it | Keeps the three-shot edit and beat? | How you pay | Best for |
|---|---|---|---|---|
| Template generator (this page) | Three photos; one photo can fill several slots | Yes: the original clip is the motion reference | One-time credit packs, 17–481 credits per video, no subscription | The exact viral clip with your own faces |
| CapCut / Dreamina | Your own clips in CapCut; a photo and a written prompt in Dreamina (Seedance) | Only if you film or rebuild each shot yourself | Free editor with paid Pro features; Dreamina uses credits | Editing, captions and music on a phone |
| Kling | A photo plus a prompt, or a motion reference clip | Not by default; closest when you build it shot by shot | Free daily credits plus paid plans | Realistic body motion and a believable car exit |
| Google Flow (Veo) | Text and image prompts, one shot at a time | No; you assemble the shots and add the song yourself | Needs a Google AI plan | Cinematic single shots if you already pay for Google AI |
| Higgsfield | A photo plus a preset effect or a prompt | Depends on the preset; nothing guarantees this trend's cuts | Credit subscription | Trying many viral effects in one app |
| Gemini / ChatGPT | Three photos and one long prompt | Rarely; one prompt cannot carry timed choreography | Free tier with limits; paid plans for more video | A quick free experiment |
| Runway / Pika | Text or image prompts | No; better at stylised shots than at a fixed meme dance | Credit plans with limited free use | Cinematic or playful experiments |
Ozil Goal AI is not affiliated with any tool in this table. It describes how each tool works for this trend; it is not a side-by-side output test.
The trend works because of timing. The hip sway lands on the beat, the selfie's whip pan hides the first cut, and the car door opens as the music lifts. A text prompt can only describe that, so the model invents its own motion and pacing. The usual result is one long shot of someone swaying, or three stiff scenes that feel like zoomed photos.
A template avoids this by using the original clip as a motion reference. Here the model receives a copy of the clip whose three performers have been turned into faceless white mannequins, plus your three photos in shot order. It copies the motion and the cuts, takes faces, hair and clothes only from your photos, and the original track goes back on at the end.
If you prompt it yourself, generate each shot as its own clip, keep each clip to the shot's length, and cut them together on the beat in an editor. It is slower, but that is how people get close with Kling or Google Flow.
Photo 1 dances, photo 2 films the selfie, photo 3 steps out of the car. Fewer than three people? Put the same photo in two or three slots. Full-body photos work best for slots 1 and 3; a clear face matters most for slot 2.
Seedance 2.0 Fast at 720p is the default. Mini is the cheapest, and Pro and 2.5 add 1080p. 9:16 matches the original; 16:9 widens the scene for landscape posts. Every option is 15 seconds.
Credits are charged when the job starts and refunded automatically if it fails. Most videos finish in a few minutes, and you can close the page and come back. Download the MP4 with the original music.
| Shot | Time | Played by | What happens |
|---|---|---|---|
| 1 | 0:00–0:05.2 | Photo 1 | Dances alone on the empty forecourt and ends with both hands up |
| 2 | 0:05.2–0:07.5 | Photo 2 | Whip pan across the sun into a walking selfie under the canopy |
| 3 | 0:07.5–0:15 | Photo 3 | Steps out between two SUVs, walks up to the camera, then the camera pulls back wide |
Every option makes one 15-second MP4. The default (Seedance 2.0 Fast, 720p) is 77 credits; the cheapest is Seedance 2.0 Mini at 480p for 17 credits. Three different photos or one photo used three times cost the same.
| Model | Resolution | Credits per video | Value at Studio rate |
|---|---|---|---|
| Seedance 2.0 Mini | 480p | 17 | $0.85 |
| Seedance 2.0 Mini | 720p | 36 | $1.80 |
| Seedance 2.0 Fast | 480p | 36 | $1.80 |
| Seedance 2.0 Fast | 720p | 77 | $3.85 |
| Seedance 2.0 Pro | 480p | 61 | $3.05 |
| Seedance 2.0 Pro | 720p | 129 | $6.45 |
| Seedance 2.0 Pro | 1080p | 324 | $16.20 |
| Seedance 2.5 | 480p | 87 | $4.35 |
| Seedance 2.5 | 720p | 195 | $9.75 |
| Seedance 2.5 | 1080p | 481 | $24.05 |
USD values use the Studio pack's rate ($69.99 for 1,400 credits). They show what the credits are worth, not a separate per-video checkout.
Here you buy credits once and spend them per video: $14.99 for 250, $29.99 for 560 or $69.99 for 1,400. There is no subscription, and unused credits stay in your balance under the pack terms. A monthly plan elsewhere only pays off if you post several videos every month; for a single clip, a small pack costs less.
| Item | What applies |
|---|---|
| Signup gift | 3 credits, enough for 0 videos at the 17-credit minimum |
| Watermark | Videos made before your first purchase carry a moving site watermark; gift credits do not remove it |
| After a purchase | New videos export clean, and earlier watermarked videos unlock their clean original without generating again |
| $5 Ozil Mini Trial | Pays only for Ozil Mini; it cannot fund this tool |
| Failed video | Credits are refunded automatically |
| What you download | 15-second MP4, 9:16 or 16:9, original music, 480p to 1080p depending on the model |
Write one prompt per shot and generate three clips, attaching the matching photo to each prompt as the reference image. Leave the song out of the generation and add it in your editor.
Handheld vertical phone video at golden hour. The person from the reference photo dances alone on an empty asphalt gas station forecourt, parked trucks and low white buildings behind. They start with one hand on the hip, sway and bounce on bent knees, swing both arms across the body, then raise both hands above the head and turn. Same face and clothes as the photo, natural hands, warm low sun, slight camera shake.
A quick whip pan across the low sun, then a selfie filmed at arm's length by the person in the photo as they walk on brick paving in front of a station building and a red-striped pump canopy. They smile, point beside their head with the free hand, then fling that arm out to the side as the frame tilts. Phone-camera look, golden light.
Wide view from behind two parked SUVs, fields and power poles beyond. The door between them swings open and the person from the photo steps out, then walks slowly toward the camera holding a phone in both hands, smiling. Near the end the camera pulls back fast to the wide view. Handheld, sunset light, realistic proportions.
Replace "golden hour" with "after dark under bright white canopy lights, damp-looking asphalt, the station sign glowing against a deep blue sky". Keep the same three shots and timings.
Give shot 1 to one person and shots 2 and 3 to the other, or let both dance in shot 1 by adding "two people dance side by side, mirroring each other". In the template on this page, put one photo in slot 1 and the other in slots 2 and 3.
Crop each person into a separate photo first. Use the two dancers for shots 1 and 2 and the third person for the car exit. A prompt that points at "the person on the left" of a group photo is far less reliable than separate crops.
Expect several tries per shot with prompt-driven tools, and each try uses that tool's own credits.
Five credited recasts made with other generators, so you can see how the three shots read with different people and cars. None were made on this site, and your result will differ.
For the exact viral clip, a dedicated template generator, because it reuses the original motion and cuts instead of guessing them from text. Among general tools, Kling is the usual pick for realistic body motion and the car-exit shot if you build the video shot by shot. Runway leans cinematic and Pika playful; neither is aimed at a fixed meme dance.
You don't need one. This generator runs in the phone browser on iPhone and Android, and the MP4 saves to your camera roll. Install CapCut only if you want to edit the result, add captions or film your own version.
There are free routes, all with limits. Kling's daily credits or Gemini's free tier can give a rough version from prompts, and CapCut is free if you film the clip yourself. Here the signup gift (3 credits) covers 0 videos at the 17-credit minimum, and videos made before a purchase carry a watermark.
All three exist. Subscriptions suit people who post every week; per-video prices or credit packs suit a one-off clip. This page sells one-time packs from $14.99 with no subscription, and each video costs 17–481 credits depending on the model and resolution.
Split it into three prompts, one per shot, each with the matching photo attached: the forecourt dance (about 5 s), the walking selfie (2–3 s) and the car exit (about 7 s). Name the golden-hour light, the handheld phone camera and natural hands, and add the music afterwards. The full prompts are in the prompt section above.
Crop the photo into three single-person images first. With this template, put the two dancers in slots 1 and 2 and the third friend in slot 3. In a prompt tool, generate each shot from its own crop; prompts that point at people inside a group photo often mix up faces.
Use photos where the hands and full body are visible and not cut off, and avoid very baggy clothes that hide the body's shape. If Mini or Fast distorts the hands, try a higher tier (Seedance 2.0 Pro or 2.5). In prompt tools, add "natural hands, correct anatomy" and keep each shot short.
On this site the same photo-to-template idea runs the Ozil goal trend, Crazy Frog AI (the invisible-moped chase), Unforgettable Car (a 24-card reveal with your car) and Zombie Love Story. Each puts your photo into a fixed viral clip, so holiday shots with a clear face work.
Upload three photos, keep the default settings and get a 15-second gas station dance in a few minutes.
