The generator above uses one built-in clip: the 15-second sunset template most AI versions of this trend copy. Photo 1 dances on the forecourt, Photo 2 films the walking selfie, Photo 3 climbs out of the car and walks toward the camera. The moves, the cuts and the music come from the template; faces and clothes come from your photos.
What the gas station dance is, and what it isn’t
The viral gas station dance began as an ordinary phone video, not a choreographed challenge. The AI wave that followed rebuilds the same three shots with other people.
| Original video | Russian blogger Mikhail Litvin with his friends Ravshan and Adel. Reposts describe it as filmed unplanned at a gas station around 4 a.m. during a 2024 road trip. |
|---|---|
| Song people pair with it | “On and On” by the Romanian group Akcent, as reposts usually label it |
| The three shots | A dance on the forecourt (0–5.2 s), a walking selfie under the canopy (5.2–7.5 s), a slow step out of a parked car (7.5–15 s) |
| The AI version | A reference clip recast with new people, one photo per shot, usually in warm sunset light instead of the original’s night lighting |
| Also called | Gas station dance trend; the petrol pump dance trend in India, where Navbharat Times covered it on 4 October 2026 |
| Not the same as | The 2026 Sunoco station in Halfmoon, New York, that gave customers a discount for dancing to Whitney Houston ( CNN ). Searches for “dance at the gas station” often land on that news story. |
Because the edit is fixed, a written prompt alone does not reproduce it: a text-to-video model invents its own dance, its own timing and usually a single shot. A gas station dance video made from the reference clip keeps the timing that makes the cuts land on the beat.
Casting planner: who plays which shot
Start from the shots, not the photos. Each shot shows one person and asks for a different kind of picture.
| Slot | Shot and time on screen | Best photo for it |
|---|---|---|
| Photo 1 | The forecourt dance, 0–5.2 s, full body in a wide frame | Full length, so trousers and shoes are not left to the model |
| Photo 2 | The walking selfie, 5.2–7.5 s, a close-up at arm’s length | A sharp, front-facing face; this is the closest shot of the video |
| Photo 3 | The car exit, 7.5–15 s, the longest shot and the ending | Full length again; give it to the person who should get the reveal |
Then match the slots to the people you have:
| You have | Slot 1 · dance | Slot 2 · selfie | Slot 3 · car exit |
|---|---|---|---|
| One person | You | You | You |
| A couple | Partner A | Partner B | Partner A |
| Three friends | The best mover | The clearest face photo | Whoever should end the video |
| One group photo of three | Crop each person into their own image (at least 300 px on the short side) and place the crops in the order above | ||
| A team of four or more | One take per three people: six staff make two takes, twelve make four. Post them as a series, or open with the boss in slot 3 | ||
The template casts people only, one adult per photo. Two people dancing side by side, or a pet, would need a different clip; you can upload your own footage to the main editor and add up to nine photos there.
Gas station dance examples
Five recasts of the same template. They show how far outfits, builds and shared themes can change while the three shots stay put.
A team in matching kit
Three football jerseys, one per shot: number 07 dances, number 10 films the selfie, number 11 climbs out of the car. Shirts like these are how a squad or staff version reads as one group.
Three friends, one colour
A white blazer, white linen and a white blazer dress: a shared dress code ties three different people together across the cuts.
Each person keeps their own outfit
Two leather jackets and a brown suede one; the last performer steps out holding a motorcycle helmet. Clothes come from each photo, the moves from the template.
Mixed group in tracksuits
Black, navy and olive velour tracksuits: three people with different builds and hair, each holding their own shot.
A deeper orange sunset
A take with a lower sun, a different car and close-ups in the last shot; it runs about 17 seconds instead of 15.
The team version works because three people in matching kit read as one group across the cuts. A shared colour or theme does the same for friends.
What a gas station dance costs: 1, 10 or 20 takes
One take is one complete 15-second video, so the price depends only on the model and resolution. Credits come from the same function the generator charges with; the dollar column uses the Pro monthly plan, the cheapest monthly credit.
| 模型 | 1 take | ≈ USD | 10 takes | 20 takes | 20 takes ≈ USD |
|---|---|---|---|---|---|
| Seedance 2.0 Mini · 480p | 23 | $0.86 | 230 | 460 | $17.10 |
| Seedance 2.0 Mini · 720p | 48 | $1.78 | 480 | 960 | $35.69 |
| Seedance 2.0 Fast · 480p | 48 | $1.78 | 480 | 960 | $35.69 |
| Seedance 2.0 Fast · 720p | 104 | $3.87 | 1,040 | 2,080 | $77.33 |
| Seedance 2.0 · 480p | 81 | $3.01 | 810 | 1,620 | $60.23 |
| Seedance 2.0 · 720p | 174 | $6.47 | 1,740 | 3,480 | $129.38 |
| Seedance 2.0 · 1080p | 436 | $16.21 | 4,360 | 8,720 | $324.21 |
| Seedance 2.5 · 480p | 117 | $4.35 | 1,170 | 2,340 | $87.00 |
| Seedance 2.5 · 720p | 262 | $9.74 | 2,620 | 5,240 | $194.82 |
| Seedance 2.5 · 1080p | 464 | $17.25 | 4,640 | 9,280 | $345.03 |
| Free credits | New accounts get 3 credits, valid for 7 days. The cheapest take is 23 credits, so signup credits alone do not make a video. |
|---|---|
| Smallest purchase | The $5 Trial Pack adds 24 credits: one Mini take at 480p (23 credits). |
| Watermark | Until your first purchase, videos carry a moving genjutsu-ai.app watermark. Any purchase unlocks the clean file of videos you already made and of every new one. |
| Failed jobs | Refunded automatically, including photos that Seedance 2.5’s review rejects. A finished take you decide to redo is charged again. |
For a batch, plan one test take on the model you intend to use, check every face, then run the rest. Mini at 480p (23 credits) is the cheapest way to check a casting; the default Fast at 720p (104 credits) is sharper for posting. See all plans on the pricing page.
Using it for a business, team or brand account
Restaurants, gyms, shops and agencies use the trend because the format is instantly recognisable and needs no filming. Four things separate a safe company post from a risky one.
- Written consent from everyone pictured. Staff included. Agree in advance which photo is used and where the video will appear, leave out anyone under 18, and remove the video if someone asks later.
- Music rights. The download carries the template’s music. Posting it as an organic trend video is how most people share it; a paid ad needs a track you have a licence for, and buying credits is not one.
- Label it as AI. TikTok, Instagram and YouTube all ask creators to disclose realistic AI-generated video. Use the app’s AI label as well as a line in the caption.
- Keep brands real. The forecourt is a generic station. Do not edit a real fuel brand or another company’s logo into it, and do not suggest that a public figure took part.
A workable schedule for 20 takes: collect consent and photos first, write a sheet with three names per take, generate one test, then run the rest one after another. Each job usually takes a few minutes and appears in the recent-videos list under the generator, where you can download it again.
Export and posting settings for TikTok, Reels and Shorts
| Where | Frame to generate | Resolution | Tips |
|---|---|---|---|
| TikTok | 9:16 | 1080p for 1080 × 1920; 720p is fine on phones | Upload the MP4 itself; keep captions above the bottom fifth of the screen |
| Instagram Reels | 9:16 | 1080p or 720p | The grid crops the cover to the centre; the dancer sits in the middle |
| YouTube Shorts | 9:16 | 1080p or 720p | 15 seconds is well inside the Shorts limit |
| YouTube video, website, presentation | 16:9 | 720p or 1080p | 16:9 widens the forecourt and cars around each shot |
Avoid screen recordings and repeated re-exports through other apps: each pass compresses the faces again. If you add text in an editor, keep the template’s music as it is so the cuts stay on the beat.
Photo checklist before you spend credits
Most disappointing takes trace back to the photos. Check these before you press generate.
| If you want to avoid | Check this in your photos |
|---|---|
| A face that looks like someone else | A recent, sharp, front-facing original: no sunglasses, beauty filters or heavy shadows, and the face large in the frame |
| Made-up trousers, shoes or hands | Full-length photos for slots 1 and 3, with both hands visible |
| Two people blended into one | One person per photo; crop group shots before uploading |
| A blurry selfie shot | The clearest face photo goes in slot 2; pick 720p or 1080p |
| Dance and music out of step | Nothing in the photos: keep the template’s own track and do not swap in a remix afterwards |
| A rejected Seedance 2.5 job | Clear photos of adults, fully clothed; or choose a Seedance 2.0 model. Rejected jobs are refunded |
More group-friendly templates: the two-person Wipe Me Down dance AI, the Raindance AI template for couples, the Alors on Danse AI video duo and the zombie love story AI video generator.
最近更新: 2026-10-23