Save Video

Save Video is the endpoint node of the ComfyUI Core Video Nodes pack (Immac/ComfyUI-CoreVideoMocks) - the thing that's supposed to take your finished video stream and drop a real file into ComfyUI's output folder. The punchline: it does write a file, with a proper .webm name, into ComfyUI/output. The file's contents are the text "Mock video content for stream: ...". Your "video" is a text file with a video extension.

That's the whole story of this pack in miniature. GitHub describes it as "A set of mock nodes for RFC purposes," and Save Video is the mock's final act: it exercises the real output plumbing - writing to the output directory, returning a saved_video UI payload so the file shows up in ComfyUI's "Output" tab - while faking the actual encoding.

The inputs

Two, both simple:

• video_stream - a STRING, force-input only, so it has to be wired from upstream (Combine Video is the natural source).
• filename_prefix - a text field, default video_output, with the tooltip "The prefix for the saved video file." Set it to something meaningful like final_clean and you'll get final_clean.webm in your output folder.

There are no schema outputs - it's an output node, so the result appears as a file, not as a port you can wire onward.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/Immac/ComfyUI-CoreVideoMocks

Restart ComfyUI. No pip requirements, no model files, no ffmpeg - the whole pack is pure Python on top of ComfyUI's own folder_paths. Manager works too: search "ComfyUI Core Video Nodes."

The honest verdict

If you want to save a real video from ComfyUI - which is the point of the real node this one imitates - you want VideoHelperSuite (VHS), whose save nodes hand an actual encoded file to your output folder. This mock is a placeholder: useful for prototyping a workflow shape, useless for production. The whole pack has roughly two stars and zero community chatter; it's a spec proposal for what native core video nodes could look like, and it's been sitting untouched since March 2025. Don't build a pipeline on it. Do steal its mental model - stream in, prefix in, file out - because that's genuinely how a good video exporter should read.