Home / Prompt

Burning Bridges AI video prompt

Four prompts, all written so the cup stays in one hand, the background guests stay themselves, and the ending lands on the hand-over-eyes move. Copy, paste, change "LEFT" if you want the cup on the other side.

1. Swap yourself into the real scene

For Higgsfield Genjutsu, Kling Motion Control or any tool that takes a reference video plus an image. @Image1 is your photo, @Video1 is your trimmed clip of the party scene.

The person in @Image1 replaces the man waving at the camera in @Video1, matching his position, pose, wave, cup and timing exactly. Keep everything else in @Video1 unchanged: the dining room, the chandeliers, the warm light, the other guests, the handheld camera movement and the cuts. Keep the face, hair and clothes from @Image1. Only the central performer changes; do not alter any background face.

2. Build the scene from one photo

For text-plus-image video models (Seedance, Kling, Veo and similar) when you do not have a reference clip. Longer, because the prompt has to carry the whole routine.

A 20 second vertical 9:16 music video clip. The person from the reference photo is the lead performer at a packed late-night party in a wood-panelled private dining room lit by crystal chandeliers, warm amber light, dark wood, red-orange halation and film grain, shot like an 85mm music video with shallow depth of field and a handheld camera.

The lead holds a silver cocktail shaker in the LEFT hand for the whole clip and never switches hands. Routine in order: walk out of the crowd pointing at the lens; wave at the camera and raise the shaker beside the head; cross arms and sway to the beat; extend one arm and point the shaker at dancing guests; flick fingers at the lens; then raise the RIGHT hand flat over the eyes and turn the head slowly as if searching the room. The mouth keeps moving as if rapping, including during the hand-over-eyes move.

Keep the face, hair, skin tone, clothes and jewellery exactly as in the reference photo. Background guests are distinct people, not copies of the lead. No text, no captions, no watermark.

3. Pet version

A 20 second vertical 9:16 clip. The pet from the reference photo stands upright on two legs in the middle of a chandelier-lit wood-panelled party, warm amber light, film grain, handheld music-video camera. It holds a silver cocktail shaker in its LEFT paw and uses its RIGHT paw to point at the lens, wave, and finally cover its eyes while turning its head as if scanning the room. Keep its real species, face, fur colour and markings. Real animal paws, not human hands. Its mouth moves as if rapping. Background guests are human and distinct. No text, no watermark.

4. Still image first (two-step method)

Some people get cleaner results by generating a film still first, then animating that still with a motion reference. This is the still prompt; animate it with prompt 1.

Photorealistic vertical 9:16 film still. The person from the reference photo stands in a crowded wood-panelled private dining room under crystal chandeliers, warm amber light, shot on an 85mm lens with creamy bokeh and light film grain. They hold a silver cocktail shaker up in their LEFT hand and point at the lens with their RIGHT hand. Same face, hair, clothes and jewellery as the photo. Background guests raise glasses. No text.

The beat list the prompts follow

Tweaks that matter

Keep going