Seedance2.5: Video Editing is a text-driven video-to-video editor built in ComfyUI. It loads your source clip with LoadVideo, optionally trims it with Video Slice, and sends the frames to ByteDance2ReferenceNode running Seedance 2.5 in video_editing mode. You describe the change in plain language—such as “replace the background with a minimalist white studio” or “remove the product label and keep everything else the same”—and the model returns a single, temporally consistent edited clip.

Technically, ByteDance2ReferenceNode handles the model/API call and enforces frame-to-frame coherence so objects, lighting, and motion remain stable as the edit carries across time. The workflow inherits the aspect ratio and duration from your input (per the video_editing mode behavior and author notes), preserving original composition and timing while applying local object removal, background swaps, scene replacements, or style transfers. MarkdownNote provides in-graph guidance, and SaveVideo assembles the processed frames into your final output video.