Burning Bridges AI prompt

Video-to-video tools such as Higgsfield's Genjutsu take a clip of the music video plus photos of the new person, and the prompt tells them what to change. The first two sentences below are from the Reddit thread from September 30, 2026 that traced the trend to Genjutsu. The rest fills in what the thread cut off.

Replace Drake in this video with the person from my reference photos. Keep everything else exactly the same: movements, gestures, timing, camera motion, lighting, background, and other people in the scene. Match their face, skin tone, hair and body proportions from the reference photos in every frame, including when a hand covers the eyes. Keep the metal cup in their hand. No text, no captions, no watermark.

The same thread says to upload photos from several angles, including the back of your outfit.

Two people in that thread hit "restricted content", and one asked whether the clip or Drake's name in the prompt caused it. Nobody there confirmed the cause. Writing "the man in the centre" instead of Drake's name is an easy first thing to try.

Image-to-video tools start from one photo and no clip, so the prompt has to describe the room and every move. This one follows the take in order:

One continuous handheld shot in a packed private dining room under crystal chandeliers, warm low amber light, a crowd in dark evening clothes cheering. The person from the photo stands in the centre, closest to the camera, holding a plain metal cup. They dance toward the camera, push the cup right up to the lens, swing both hands down on the beat, grab their collar and tug it with a shrug, point at friends who point back, then put a flat hand over their eyes and scan the room before waving at the camera with the cup raised. Keep their face, hair and clothes from the photo in every frame. No text, no captions, no watermark.

Check how long a clip your tool allows before you paste the whole thing. Kavel and Renoise, for example, stop at 15 seconds, which fits about half of the take, so cut the moves you don't need from the end.

Prompts get less reliable with every extra face: friends swap faces, merge into one person or turn into strangers. If your tool takes several reference images, give each person a place, for example "the person from image 2 stands on the host's left, the person from image 3 on the host's right", and ask it never to copy one face onto another.

Most tools give you a silent clip. In TikTok, tap Add sound, pick "original sound - cazetv", the most reused sound for this trend, and slide it until the hand over the eyes lands on "where she at".

On this site there's nothing to write. You add a photo of yourself and up to 6 friends, and the room, the moves and the timing are already set up for seven people. See how it works, step by step.