A Bulin 47 AI prompt that actually holds
If you use a general tool that accepts a reference video and a photo, describe the scene, say exactly who gets replaced and what stays, and time the gestures second by second. Ask it to keep the reference audio unchanged. Without timing, most tools re-act the scene from the text, so moves drift and the audio gets re-sung. On this site you skip the prompt: one photo, and the timing is already done.
Last checked
All guidesGot a photo? Start here
Choose one photo and step into Bulin 47's spot in the Madrid freestyle. The first 2 seconds cost nothing.
Nothing uploads yet. You check everything in the booth before a video is made.
A prompt you can adapt
Paste this into a tool that takes a reference clip plus an image, then edit the bracketed parts. It is a starting point, not a guarantee; every tool reads prompts differently.
"Follow the reference video exactly: same framing, handheld camera, timing and every gesture. Night scene outside a red-lit restaurant entrance, staircase, security staff and a crowd filming. Replace only the bald man in sunglasses and a pink mesh shirt with [the person in the image], keeping [their face, hair and clothes] from the image. Everyone else stays unchanged, including the curly-haired man in the foreground holding a round black speaker and a phone. 0 to 7 s: raps straight at the phone, open hands moving with the beat. 7 to 12 s: leans toward the speaker, rubs the top of the head and laughs. 12 to 15 s: looks down at the phone smiling. Keep the reference audio as is: no new lyrics, no re-singing. Vertical 9:16, no captions."
What makes or breaks it
| Detail | Why it matters |
|---|---|
| Who gets replaced | Name the clothes and look of the performer or the tool may swap the fan instead |
| Timed gestures | Stops the tool from inventing its own moves |
| "Keep the audio" | Many tools regenerate sound unless told not to |
| A 15 s trim | Most tools cap reference clips around 10 to 15 s |
| A clear photo | Same as here: one face, front-on, good light |
Why one photo is simpler
This site already holds the trimmed clip and a tested, timed prompt; you add a photo and the original audio from Kevin Wai’s post is laid back after the render. It costs 10 credits at 480p. If you enjoy tinkering, the prompt route is fun; if you want the clip tonight, use the booth.
The doorway is waiting
Make my Bulin 47 videoFAQ
Asked and answered
Can ChatGPT or Gemini make the Bulin 47 video?
They can write a prompt, but the video still has to be made by a video tool that accepts a reference clip and a photo.
Does this site let me edit the prompt?
No. The prompt is fixed and tested so every video follows the clip.