The Smalltown Boy dance AI trend is a 15-second clip in which one character lunges at the lens, clutches their head and screams while a second dances behind, then the two dance side by side and reach for the camera, set to Bronski Beat's Smalltown Boy. On videosi you can make the same clip starring you and a friend, or your two pets, from two photos, with the song already on it and no prompt to write. This guide answers the questions people ask after the basics.
Key takeaways
- The earliest posts we could date are from 3 October 2026 on TikTok; the most-viewed one we found is from 8 October, so the trend is still spreading.
- The song is Smalltown Boy by Bronski Beat (1984), usually the remastered version, and it is not the same format as the 2024 "dance like it's the 80s" challenge.
- Photo 1 is the screamer in front and photo 2 the dancer behind; each can be a person or a pet.
- A 15-second Standard 720p video is 13 credits, and failed renders return their credits.
- Nobody has confirmed who made the first version, and videosi is not affiliated with TikTok, CapCut or Bronski Beat.
In this guide
- What is the Smalltown Boy dance AI trend, exactly?
- Where and when did it start?
- Which Smalltown Boy version is used, and is it the 2024 trend?
- Who screams and who dances?
- How do I make it from my own photos?
- Does it work with cats, dogs and other pets?
- Which photos work best?
- Which AI model makes it, and can I use other tools?
- How much does it cost?
- Why did mine come out wrong?
- What should I know before posting it?
What is the Smalltown Boy dance AI trend, exactly?
It is a comedy dance in two acts, filmed from a low, wide-angle camera as if someone were holding a phone at arm's length. In the first act a character shoves their face toward the lens, presses both hands to their head and screams with their mouth wide open. Far behind them a second character dances with their arms up and keeps going the whole time. In the second act the shot pulls wide, the two stand side by side and sway in time, and the clip ends with both leaning in and reaching for the camera.
The scream is silent. What carries the video is the music, the faces and the contrast between the front character losing it and the back one dancing calmly. The structure is fixed, which is why it has been remade so many times: any two characters can fill the two roles. Our Smalltown Boy dance AI page shows the example clips and the beat-by-beat timeline.
Where and when did it start?
We found the format on TikTok in the first days of October 2026. The earliest dated posts we could check, a cat version and a live-action version with two fighters, are from 3 October. The most-viewed post we found, two cats in a living room, is from 8 October. Template and tutorial posts for CapCut followed within days, and later versions use actors in costume, football players, other cats and dogs.
Nobody has confirmed who posted the first one, and we do not know whether the cat version or the fighters' version came first. That is why the page marks the origin as unconfirmed. As of 1 November 2026 new versions were still appearing, so we call the trend spreading rather than peaked.
Which Smalltown Boy version is used, and is it the 2024 trend?
The clips use Smalltown Boy by Bronski Beat, the 1984 single, in most cases the remastered release. The song also had a TikTok moment in May 2024, when kids played it for their parents and asked them to dance like it was the 80s; Stereogum covered it when the track turned 40. The 2026 trend is a different format: it is an AI-style scream-and-dance clip with two characters, not parents dancing in a kitchen.
For the song's background, Prospect magazine's piece on the 40th anniversary looks back at the record's history. The song itself is about leaving a small town, and the 2026 dance does not refer to the lyrics, so treat the meme as a joke that happens to use a famous track.
We compared the audio of the most-viewed cat clip with the fighters' clip, another AI cat version and a costume version, and all four matched, so there is one track for the whole trend. The videosi result has that track already mixed in, lined up with the moves.
Who screams and who dances?
The front character gets the close-up, the lunge and the scream. The back character gets the arms-up dance, and after about 7 seconds the two switch to dancing together. On videosi photo 1 is always the front character and photo 2 the back one, so put the face you want in the long close-up first. If you upload one photo of both of you instead, the person on the left is treated as the screamer.
The reference clip shows two different performers in the opening seconds, one lunging first and a second taking over the front. The videosi version keeps the part that repeats in every copy: one character screams in front, one dances behind, then both dance and reach for the lens.
How do I make it from my own photos?
Open the Smalltown Boy dance AI page, upload two photos, choose the model and the format, and press Generate. The 15-second clip renders in a few minutes. It is an MP4 without a watermark and with the song on it, ready for TikTok, Reels or Shorts at 9:16, or 1:1 and 16:9 if you prefer.
There is no prompt to write and nothing to choose except the photos, the model tier and the format. The motion, the room and the timing are built in, so the trend looks like the originals instead of like a loose interpretation.
Does it work with cats, dogs and other pets?
Yes, and cats were the most-viewed version of the trend. Either photo can be a pet: it stands on its hind legs, screams or dances, and keeps its breed, coat and markings. A pet in front and a person behind, or the other way round, works too. Use a sharp, well-lit photo of the whole animal with its face clear.
Very small or very dark animals are harder, because fine markings and eye detail get lost during the lunge at the lens. If your pet has distinctive markings, pick the photo where they are easiest to see.
Which photos work best?
The front character is seen from the waist up in a close-up and then full length, so a photo that shows the face and shoulders, with arms visible and clothes you want to keep, works best. Avoid sunglasses, hats over the eyes and heavy filters. Faces turned straight to the camera keep their shape through the lunge; a strong side profile is more likely to drift.
Use one person or one pet per photo. A group photo can confuse the model, which expects exactly two characters in the clip. If you want a single image of you both, switch to the one-photo mode and put the person who should scream on the left.
Which AI model makes it, and can I use other tools?
videosi uses ByteDance's Seedance video model: Seedance 2.0 Mini by default, with Fast, Pro and Ultra tiers on Seedance 2.0 and 2.5. It is given your photos, a short motion reference and a written prompt that tells it to replace the original two performers. Other tools such as Kapwing or CapCut templates offer similar results; they change often, so check them in their own apps.
You can try the trend in a general video model with your own prompt, but it usually takes several paid attempts. The usual problems are the two faces blending and one of the original performers staying in the shot when a face fills the frame.
How much does it cost?
It is not free. A standard 15-second video (Standard model, 720p) is 13 credits; better models and higher resolutions cost more. Credit packs start at $6.99 for 10 credits and never expire, and plans start at $29.90 a month. A failed, timed-out or blocked render returns its credits automatically. There is no free generation and no free trial.
Why did mine come out wrong?
The usual causes are a tight face crop, sunglasses, more than two people in a photo, or a dark and blurry image. Close-ups are also where invented detail shows first: hands pressed to the head and the final reach toward the lens can look slightly off, and likeness is not guaranteed in a 15-second clip. Try a sharper photo, with the face centred and the upper body visible.
If you used a pet, check that the photo shows the whole animal; a cropped face gives the model too little to hold on to during the screaming close-up.
What should I know before posting it?
Turn on the AI-generated label on TikTok, Instagram or YouTube before you publish; platforms ask for it and can take down unlabelled synthetic video. The clip already carries Bronski Beat's track, so post it as it is, or swap the sound inside the app if you prefer. Use only photos of adults who agreed to it, or your own pets: videosi does not accept celebrities, public figures or minors.
The trend is still moving, so expect new twists on the same two-act format. If you want to see where else the same two roles apply, our AI cat dance page and the dancing dog page use pets in the same way.
