The Smalltown Boy AI video is 21 seconds of controlled chaos: two subjects take turns screaming into an ultra-wide lens while the other dances behind, then both dance side by side and rush the camera, to Bronski Beat's 1984 song. This guide is about making your own with Smalltown Boy AI, based on the test renders we ran before putting it live.
If you are still wondering where the trend comes from (and whether it is the Netflix Lizzie Borden dance), read the Smalltown Boy trend explained first.
What you get
- The full 21.7 seconds of the original clip, at its own cuts: scream one, scream two, the side-by-side dance, the rush at the lens.
- Vertical 9:16, ready for TikTok, Reels and Shorts.
- The song from the original clip already on the file.
- 10 credits at 720p, 7 at 480p. A failed render is never charged.
Many AI tools give you an 8 to 14 second "duo dance". The meme only lands with both screams and the ending, so we kept the whole thing.
Step 1: decide who screams first
Photo 1 owns the first close-up (0 to 7 seconds), photo 2 owns the second (7 to 14). Whoever is in the background is the small dancer by the fence. Some pairings that work:
| Pair | Photo 1 | Photo 2 | Why |
|---|---|---|---|
| Two cats | The more dramatic one | The calmer one | The viral version: let the drama queen open |
| You and your dog | You | The dog | Your scream sets up the joke; the dog answering it is the punchline |
| Parent and kid | The parent | The kid | A nod to the 2024 parents' dance trend |
| Couple or friends | Whoever overreacts more | The other one | Caption it with what made you scream |
Use two separate photos, one subject each. A single photo of the two of you is not accepted for this effect; the tests below explain why.
Step 2: pick photos that survive an extreme close-up
Each scream is a face pressed right up to the lens, so the face carries the clip.
- People: front-facing, even light, nothing covering the face, one person per photo. A half-body or full-body shot helps, because the dance at the end shows everyone head to toe.
- Pets: a sharp photo where the face, eyes and markings are easy to see. A sitting pet is fine: it stands up on its own in the video.
- Mixing: a person and a pet works. The pet keeps its own fur and stands on its hind legs for the whole clip, like the viral cats.
Step 3: make it and post
Add the photos, keep the default (Standard, 720p, 21 seconds) and press make. It takes about four minutes. Download the MP4; the song is already on it. When you post on TikTok you can still pick the "Smalltown Boy" sound if you want your video on the sound page.
Our samples
Two cats, two separate photos: the tabby screams first, the tuxedo cat second.
A man and his golden retriever. The dog sits by the fence during the first scream, then stands up for its own.
All the people in our samples are AI-generated, not real customers.
What went wrong in our tests (and what we changed)
We ran ten test renders before launch. Four things came up:
- A cat turned into a child. In the person-and-cat test, the small dancer in the background of the first scream should have been the cat, and came out as a blond boy. The background figure in the original clip is a person, and the AI followed its shape. We now describe that background dancer as the animal itself, pet-sized, on its hind legs. The re-run had the cat dancing by the fence.
- Black fisheye corners. "Ultra-wide" sometimes came back as a round fisheye frame with black corners. We changed the wording and it mostly went away, though a soft dark corner still shows up now and then.
- One photo of a couple drew the same person twice. From a single photo of a man and a woman, the second scream showed the woman in front and the woman again dancing behind; the man was gone. It happened in both tries, even after we spelled it out in the prompt. From two separate photos it never happened, man first or woman first. So this effect only takes two separate photos.
- Borrowed shorts. In the mom-and-daughter render (made from one photo, before we turned that off), the daughter's lower half came out in the original dancer's khaki shorts and checked socks during the dance. The top half kept her own sweater. A full-body photo gives the AI more of your actual outfit to work from.
Two things never went wrong: the screams always landed on the right person, and the join between the two halves of the clip, at the 14-second cut, is invisible.
FAQ
Can I use two different animals? Yes. A cat and a dog, two dogs, a dog and a person: each photo plays its own part.
Does it work with one photo of both of us? No, use two separate photos. With one photo of a man and a woman, the AI sometimes drew the same person twice.
Is the original song included? Yes, timed to the cuts.
Why does it cost the same as a 15-second video? We render the clip in two halves on the standard model and join them at the original cut, which keeps the full 21 seconds at the normal price.
Ready? Make your Smalltown Boy AI video.