Wan2.7: Video Edit
Upload a video and corresponding images to edit and replace characters or scenes using the Wan2.7 model.
简介
This ComfyUI workflow demonstrates reference-guided video editing with Wan2.7. You load a source clip with LoadVideo, provide one or more reference images with LoadImage, and let the Wan2VideoEditApi node apply character or scene replacements consistently across frames. The node calls the Wan2.7 model to infer identity, attire, and style cues from your references, then synthesizes edited frames that preserve the original camera motion and timing. Finally, SaveVideo assembles the frames into a finished MP4/WEBM with your chosen frame rate and quality.
你需要提供
- 正在读取工作流
在 ComfyUI 中运行。 此工作流无法在本页面运行。你可以在 Comfy Cloud 中打开（已预先加载所需内容），或下载后在自己的 ComfyUI 中使用。
在 Cloud 中试用常见问答
The Wan2VideoEditApi node typically calls a remote Wan2.7 service, which may require credentials and an active internet connection. Check the node’s settings or your ComfyUI extension docs to configure any required keys.
Use high-resolution reference images that match the video’s angle and lighting, keep prompts specific (describe attire, colors, and style), and tune the edit strength to avoid overdriving the look. If your node exposes a seed, lock it for reproducibility.
Yes. If LoadVideo or the API node exposes a frame range, use it to test or limit the edit. Otherwise, trim the source clip externally before loading it, then process just that segment in ComfyUI.
SaveVideo in ComfyUI generally exports silent video. To keep audio, export the edited clip and then remux the original audio track using a video editor or a tool like FFmpeg.