ComfyUI Node

Video Stream Data

Five metadata outputs, all hardcoded

By Immac·Created about a year ago·Updated about a year ago· 2
Video Stream Data
    • frames_per_second
    • frame_count
    • duration
    • width
    • height
    video_stream

    Video Stream Data is the metadata probe of ComfyUI Core Video Nodes (Immac/ComfyUI-CoreVideoMocks). In a real video pipeline, this is the node you'd use to find out what you're actually working with - framerate, length, resolution - before you commit to decode settings. Its claim to fame in this mock: it's the node with the most outputs in the whole pack, and every single one is a hardcoded number.

    One input, video_stream (a force-input STRING), and five outputs:

    • frames_per_second (FLOAT)
    • frame_count (INT)
    • duration (FLOAT)
    • width (INT)
    • height (INT)

    Sound useful, right? The source literally returns 30.0, 300, 10.0, 1920, 1080 every single time. Thirty frames per second, three hundred frames, ten seconds, 1920×1080. That's not metadata from your video - that's a sample frame for a 10-second 1080p30 clip, hardcoded as a constant. Feed it a 2-second vertical phone video and it will cheerfully tell you it's a 1920×1080, 10-second, 300-frame clip.

    Why this one's a trap

    The outputs are real types (FLOAT, INT), so other nodes will happily consume them. If you wire width and height into something that sizes your output, or frames_per_second into a timing calculation, you'll get the same plausible-looking wrong numbers every time. There's no error - just confident misinformation. It's the quiet sibling of Decode Video Stream's fake IMAGE output: no crash, but quietly useless.

    Install

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

    Restart ComfyUI. No pip dependencies, no models, no ffmpeg - the pack is pure Python on ComfyUI's own folder_paths, also available via ComfyUI Manager under "ComfyUI Core Video Nodes."

    The honest verdict

    Like the rest of this RFC pack - two GitHub stars, no community footprint, untouched since March 2025 - the value is in the API shape, not the output. The five-output metadata probe is a reasonable guess at what a core video node should expose; you could copy this interface and implement it for real over PyAV or ffmpeg in an afternoon. But don't trust the numbers this one hands you for anything. For actual metadata probing in ComfyUI today, VideoHelperSuite (VHS) is the tool with real answers.

    Categoryvideo

    Inputs (1)

    NameTypeDefaultDescription
    video_streamSTRING

    Outputs (5)

    NameTypeDescription
    frames_per_secondFLOAT
    frame_countINT
    durationFLOAT
    widthINT
    heightINT