Burning Bridges AI prompt: how the video is made

Many people search for a prompt to recreate the Burning Bridges party video themselves. Here is the approach behind our tool, with its limits.

The idea: reference video + your photo

Instead of describing the whole party scene in words, a reference-to-video model receives two inputs: a short reference clip that fixes the room, camera movement and performance, and a portrait that fixes who the lead is. The text prompt only explains which input does what.

The prompt we use

Use the supplied portrait as the identity and clothing reference for the foreground lead only. Follow the supplied video's performance, camera path and timing. Preserve the other people and the room. Keep the new lead recognizable through hand and cup occlusions.

Version a-2026-10-12-short. Short instructions that assign roles to each input tend to follow the reference more closely than long scene descriptions. We change this text only after a side-by-side test, and the version above is the one running on the site.

Why a prompt alone is not enough

• You need a clean reference clip you are allowed to use, with the lead large enough in frame.
• Results vary: hands and cups covering the face are the hardest moments.
• The output length is limited by the model (up to 15 seconds for the model we use).

If you would rather skip the setup, upload a photo here and get the finished MP4.