Nodes/ComfyUI-vlo/vlo Memory Load Audio Batch
ComfyUI Node

vlo Memory Load Audio Batch

Feed audio to your graph from memory instead of a swelling input folder

By PxTicks·Created 3 months ago·Updated a day ago· 0
vlo Memory Load Audio Batch
    • audios
    audios
    disable_in_memoryfalse

    vloMemoryLoadAudioBatch is the audio member of the ComfyUI-vlo pack's memory-loader family. Where the image and video versions deliver frames you can look at, this one delivers clips you can hear - ordered AUDIO items sourced from an in-memory registry that an external app (vlo) fills per run, so nothing transient ever gets written to your ComfyUI/input folder just to pass through a graph.

    Think of the loader family as the answer to a specific annoyance: the ordinary ComfyUI flow of dropping files into input/ assumes the files have some lasting value and that a human chose them. When another program is generating many short-lived media items and handing them to ComfyUI by id, that assumption breaks. These nodes hold the media in RAM, reference it by id, and the batch variants add an ordered multi-select so you can hand a whole collection downstream in a defined sequence.

    Inputs, outputs, and the one toggle

    It's a lean node - two inputs, one output:

    • audios - your ordered multi-selection, rendered by a web extension that refreshes directly from vlo's registry.
    • disable_in_memory - set true and every selection loads from your normal input directory instead of the registry. This is the setting to flip when you're testing a workflow by hand in the ComfyUI UI, without the vlo app driving it. Nothing in the registry, or a stale id from a previous run, will bite you.

    The output is an ordered audios AUDIO list - genuinely a list, not one concatenated tensor. So the clips in a batch don't need matching durations, and each one stays a distinct clip you can address separately. A consumer that needs the whole collection at once has to accept ComfyUI list inputs; an ordinary node will execute once per list item. That's the same contract as the pack's image and video batch loaders, so once you've seen one you've seen them all - and the video variant additionally carries per-item audio flags, which the MiniMax H3 reference adapter can consume directly.

    Selections are capped at 100 items as a general safety bound.

    Install

    Whole-pack install, standard fare:

    cd ComfyUI/custom_nodes
    git clone https://github.com/PxTicks/ComfyUI-vlo.git
    

    then restart ComfyUI (or search "ComfyUI-vlo" in ComfyUI Manager). There's no requirements.txt - nothing extra to pip install. Two honest caveats if you didn't come here via vlo: without the vlo app feeding the registry, the node's interesting half is inert (use disable_in_memory and it becomes a slightly fancier audio-from-input loader), and the pack targets a recent ComfyUI, so if nodes fail to register, update ComfyUI before you debug anything else.

    Categoryaudio

    Inputs (2)

    NameTypeDefaultDescription
    audiosCOMBO0 options:
    disable_in_memoryBOOLEANfalseWhen true, load every selection from ComfyUI's normal input directory instead of the vlo in-memory registry.

    Outputs (1)

    NameTypeDescription
    audiosAUDIOOrdered audio list.