Bernini-R: Video Edit

Re-light and restage a clip without changing its motion: Bernini-R edits the footage frame by frame from a written instruction.

Input

What the scene should become; motion is kept.

The footage to edit.

Advanced settings

Leave empty for a new result each run. Enter a number to repeat one.

Output
00:00
Example

試試範例

關於此工作流程

Bernini-R edits footage instead of regenerating it: the source clip is carried into the conditioning as the motion and structure reference, and the prompt changes what fills it — a grey studio backdrop becomes a daytime street, an interior becomes night. Because the timing comes from the source, people keep walking the way they walked and the camera keeps its move; only the scene around them is re-rendered, with the new light falling consistently on the subject. The template slices the first 5 seconds, runs a distilled two-stage sampler at 3 steps, and writes an mp4 with the source audio. Width and height are fixed at 480x832 in the graph rather than inherited from the clip, so a source in another shape is fitted rather than cropped.

Graph: LoadVideo -> Video Slice (0-5 s) -> GetVideoComponents -> BerniniConditioning (source video, length from the slice) -> two-stage SamplerCustom (3 steps, res_multistep) -> VAEDecode -> CreateVideo -> SaveVideo (mp4).