Back to home

How to Make the Viral Smalltown Boy AI Dance Video

Step-by-step: upload two portrait photos (left and right), generate a ~15s · 9:16 · 720p Smalltown Boy AI clip on this site — Bronski Beat audio included, no long prompt.

Last updated: 2026-10-14

The Smalltown Boy AI video format is a two-person scream-and-dance clip set to Bronski Beat. On this site you can make one in a few minutes with two portrait photos, a built-in ~15-second reference, 9:16 output, then preview and download.

Start here: Homepage generator

Note: Smalltown Boy AI is an independent tool inspired by the 2026 trend — not an official Bronski Beat or label product.

Easiest method (this site)

  1. Open the homepage generator and sign in.
  2. Upload two clear portraits — one for each performer slot (see order below).
  3. Output is 9:16, 720p, and about 15 seconds (production).
  4. Tap Generate — you get a clip with our built-in Smalltown Boy dance reference (no prompt box on the homepage).
  5. Preview the full clip, then download MP4. The reference audio track is included; review before posting.

You do not need to upload the original Instagram reference or paste a long Wan prompt when using this site.

Read this first: upload order is the casting plan

Upload order maps to performers in the reference clip:

| Slot | Role in reference | Tip | | ----------- | --------------------------------------------------- | ------------------------------------- | | Image 1 | Dancer who starts on the LEFT in the reference | Clear face; matches opening layout | | Image 2 | Dancer who starts on the RIGHT in the reference | Pair with Image 1 for balanced energy |

Want a solo-style joke? Use two different photos of yourself — still two uploads.

Why reference motion matters

Generic image-to-video with a text prompt rarely copies the scream timing, swap, and courtyard dance that make the meme recognizable. The trend is tied to a specific performance edit. Here that reference is already applied — upload two photos and generate.

What you need before you start

  • Two images — clear face or upper-body photos; JPEG/PNG/WebP as the uploader allows
  • Account & credits — sign in; trial credits apply when enabled on your account
  • Consent — only upload people (or pets) you have rights to use
  • Expectations — results vary; review identity, hands, and overlap every time

Step 1: Pick strong photos

Do

  • Front or three-quarter view, eyes visible, minimal motion blur
  • Even lighting; avoid heavy beauty filters
  • One subject per file — head-and-shoulders works well

Avoid

  • Group shots in a single slot (two faces confuse casting)
  • Sunglasses or masks covering key features
  • Extreme crops that cut off the jaw or forehead

More notes: Smalltown Boy prompt guide.

Step 2: Open the generator

  1. Open the homepage and go to Generate (/#generator).
  2. Sign in if prompted.
  3. Upload two photos in the labeled order.

You do not paste a prompt on the homepage; the Wan reference and performer mapping are applied for you.

Step 3: Generate and review

| Setting | On this site | | ------------ | ------------------------------------------------------------ | | Aspect ratio | 9:16 (fixed on homepage) | | Length | ~15 seconds (production) | | Resolution | 720p (marketing); backend may use 480p for the model run |

  1. Confirm credits on the Generate button.
  2. Wait for the job to finish; open Preview.
  3. Check both identities, scream beats, foot contact, and audio sync.
  4. Download when satisfied.

Step 4: Share responsibly

  • Label AI-generated content if your platform or audience expects it.
  • Do not imply Bronski Beat, the original dancers, or a label endorsed your clip.
  • See examples for gallery-quality references.

Troubleshooting

| Issue | What to try | | ---------------------- | ------------------------------------------------------------------------- | | Faces swap mid-clip | Re-upload with clearer, more matched lighting; avoid heavy filters | | Wrong “performer side” | Swap Image 1 and Image 2 — order defines left/right mapping | | Weak scream timing | Use photos with visible forehead and hands; avoid tight crops | | Audio feels off | Re-generate once; if still wrong, treat as a failed run and adjust inputs |

Learn the trend context