Nodes/FunUI/FunUI Load Media
ComfyUI Node

FunUI Load Media

An ID, Not a Filename

By digital-garbage·Created about a year ago·Updated a day ago· 21
FunUI Load Media
    • image
    • frames
    • audio
    ◄media_id►
    ◄blank_width0►
    ◄blank_height0►
    ◄blank_frames0►
    ◄fullfalse►
    ◄prepare►

    Every other loader in ComfyUI wants a filename from a folder. This one wants an ID from FunUI's media library - the thing you get when you drag a file into the Media bin in the studio at /fun/. That's a strange-looking input for a ComfyUI node, and it's deliberate: the app writes the ID onto the node before a run queues, the same way a scene's prompt text lands on a text-encode node.

    So if you open this node in a normal graph and see an empty string, nothing is broken. You're the one who has to fill it in, and there's no file browser to help you.

    What it actually does

    Given a media ID, it looks the file up in the library, works out what kind of thing it is, and fills whichever outputs make sense:

    • A picture gives you the image; the frames and audio outputs get the same single frame and a placeholder silent track.
    • A video gives you its first frame as image - which is all a picture input (a scene's shape reference, for instance) needs - and, with full on, all the frames at 24 fps plus the soundtrack.
    • An audio file gives you the sound and a tiny grey 64×64 picture on the image outputs, because those simply won't be wired.

    That last trick is worth noting: every output always has something on it. There's no "None means crash later" - the wrong outputs for your file type carry harmless filler. And when the file has no sound, the audio output is a single silent sample rather than nothing.

    The other escape hatch is blank_width/blank_height. Give both a size and, with no file picked, the node manufactures a plain mid-grey picture of that size instead of refusing. That's for text-only starts: on H3's preset the "scene picture" exists only to give the pipeline a shape, so a grey stand-in is exactly right.

    Inputs and outputs

    • media_id - the ID string, normally written by the app. Non-negotiable if you want a real file.
    • blank_width / blank_height - both must be above 0 for the grey fallback; 0 means refuse.
    • full - off by default. Off reads only the first frame, which is fast and is all a picture input wants. Turn it on when you need the whole video and its sound.

    Outputs: image (the picture, or a video's first frame), frames (a video's frames at 24 fps), audio (soundtrack or audio file). Wire the one your downstream node asks for; the others are filler.

    Installing it

    Part of the FunUI pack:

    cd ComfyUI/custom_nodes
    git clone https://github.com/digital-garbage/FunUI
    cd FunUI && pip install -r requirements.txt
    

    Or ComfyUI Manager → search FunUI. Restart ComfyUI afterwards. (The pack page you're on says olivv-cs/ComfyUI-FunPack; the project renamed itself to FunUI and lives at digital-garbage/FunUI. If your clone folder kept the old name, node info will show custom_nodes.ComfyUI-FunPack - cosmetic.)

    Video and audio decoding goes through PyAV, which ComfyUI already depends on, so there's nothing extra to install for that. FunUI's own module settings can add spaCy and NVIDIA VFX later, but neither touches this node.

    Then: open /fun/, drop files into the Media bin, and they get IDs. That's the workflow the node expects.

    Where people get burned

    Typing a filename into media_id. It won't resolve. The string is an opaque library ID, not a path; the node has no fallback to ComfyUI/input. If you want the classic file picker, use a normal Load Image / Load Video node instead - nothing in the pack requires you to use this one.

    "Media not in the library any more." That's a deleted or moved file, or a workflow opened on a different machine where the library doesn't exist. The ID means nothing outside the project that made it - this is not a portable loader, and it's the single biggest trap here for people sharing workflows.

    24 fps is hardcoded. frames always come back at 24 fps, a constant in the source aimed at the H3 reference-video input. If you save at a different fps in FunUI Save Video, your clip will run fast or slow compared to the source. Keep 24 end to end unless you've thought about it.

    Forgetting full. The number one confusion: you load a video, get one frame, and conclude the node is broken. It's doing what it says - first frame only, unless you ask for everything.

    A grey 64×64 sneaking downstream. If you accidentally wire the image output of an audio file into a conditioning node, you'll get a valid but meaningless picture. The console says "loaded X (audio)"; nothing downstream will.

    CategoryFunUI/Media

    Inputs (6)

    NameTypeDefaultDescription
    media_idSTRINGA media library id, picked in the Media bin.
    blank_widthoptINT00–16384With nothing picked, make a plain grey picture this wide instead of refusing. 0 = refuse.
    blank_heightoptINT00–16384With nothing picked, make a plain grey picture this tall instead of refusing. 0 = refuse.
    blank_framesoptINT00–100000With nothing picked and a size given: how many grey frames the stand-in video has. 0 = one.
    fulloptBOOLEANfalseRead the whole video and its sound. Off: only the first frame, which is all a picture input needs.
    prepareoptSTRINGWith full on: how to get the video ready, as JSON. Keys: fps, skip_first, cut_last, every_nth, frame_cap, width, height, fit (crop|pad|stretch), crop_left/right/top/bottom (percent), multiple_of. Empty: all frames at 24 fps, size untouched.

    Outputs (3)

    NameTypeDescription
    imageIMAGE—
    framesIMAGE—
    audioAUDIO—