The freestyle trend, explained

What happened

The original clip is a single, unplanned take. A street performer stops a passer-by, plays a beat from his phone through a small speaker, and invites him to rap. The passer-by goes for it on the spot. A crowd forms, phones come out, and the moment is over in under twenty seconds.

That's what made it travel: one continuous shot, a clear beginning, middle and end, and a performance that feels completely unrehearsed. Short enough to rewatch, loose enough to imitate.

Why people recast it

Soon after the original spread, AI versions started appearing in which someone else takes the lead role: friends, family members, sports stars, cartoon characters. The format works because everything except the performer stays the same. The joke, or the flex, is entirely in who is standing in the middle of the crowd.

Recreating the scene for real isn't practical. You'd need the location, the lighting, the crowd and the timing. Video models solve that: give them the original take as a reference and a photo of someone new, and they re-render the scene with the new person in it.

How the recast works

1. You upload one photo. A clear, front-facing photo with the upper body visible works best.
2. The video model gets two references. Your photo and the original clip. The instruction is to keep the scene, camera, crowd and timing, and to replace only the lead performer.
3. You get the full take back. Same length, same framing, original sound, with you as the lead.

The model behind it is Seedance, ByteDance's video-generation model, which can take a reference video and a reference image together and hold a character's likeness across a full clip.

Doing it responsibly

• Only recast yourself, or adults who've agreed to it.
• Don't present the result as real footage.
• Credit the original when you post, and respect takedown requests from the people in it.

Make yours

Head to the generator, upload a photo and you'll have your version in a few minutes.