Burning Bridges AI Prompt: Recast the Lead, Keep the Scene

Start with the kind of video you are making

A text-only party prompt asks a model to invent a scene. A reference-video edit asks it to change a person inside an existing performance. The latter fits this template: the supplied video already contains the movements, camera work and surrounding guests.

A long description of a different nightclub, new camera cuts or a new supporting cast can conflict with that reference. Keep the instruction focused on the one replacement and on the details that must continue through the original clip.

An original reference-edit prompt example

Replace only the foreground lead in the reference video with the person in the uploaded photo. Use the photo’s face, hair and outfit, and keep that identity on the same performer through turns and hands covering the face. Preserve the silver cup, hand gestures, background guests, room, camera movement, framing and timing. Do not add people, captions or new shots.

This is an explanatory prompt you can adapt in a tool that accepts a reference video and portrait. Attachment syntax varies by tool. In the EasyVideoAI studio, the template already provides the role mapping and scene instructions; there is no text box to paste this into.

Open the Burning Bridges AI studio

Use opening features to identify the role, not the final outfit

The original lead starts in a black tank top, light trousers and a necklace. Those details identify which person to replace. They should not force your result to retain the original face, hair or clothes when the uploaded photo shows a different appearance.

A photo with visible hair and upper-body clothing makes that distinction easier. Keep background people outside the replacement instruction. Asking for everyone to look like the photo can spread the same identity across the crowd.

Preserve contact and identity through occlusion

The silver cup belongs to the scene, and its position should stay tied to the lead’s hand. The most useful review points are the grip, gestures near the lens and the moment the hands cover the face. The face should return as the same person afterward.

Instructions can guide these details, but they do not guarantee flawless fingers or likeness. A clear portrait and a complete review of the rendered clip remain necessary. Repeating identity adjectives in a prompt cannot recover facial detail that is absent from a blurry upload.

Treat the soundtrack separately from the visual prompt

This template restores the matching source audio after the visual edit. The portrait does not create a new vocal performance. A prompt asking for new lyrics or a different singer would describe a different workflow.

Choose the scene length in the studio instead of asking the prompt to extend it. The available choices are the first 15 seconds and the full approximately 29.08-second clip, both in the original near-4:3 frame.