RUN THIS WORKFLOW NOW ON FLOYO!

ABOUT THE WORKFLOW

Edit a video from a prompt
Upload a source video, add one or more reference images, and describe what you want changed. Bernini reads the clip and your instruction together, then returns a new video with your edit applied and everything you did not mention left intact. It also generates video from a prompt or an image when you start without source footage.

This is the quantized GGUF build, split into a high-noise and low-noise stage, so it runs lighter and faster than the full model while keeping the same editing behaviour.

Model

• Bernini by ByteDance. An open-source video model that pairs a semantic planner with a renderer built on Wan, strong at instruction-based edits that hold identity and detail across the whole clip.