Nodes/ComfyUI-FolderBatch/FolderBatch Video Queue
ComfyUI Node

FolderBatch Video Queue

Work through a folder of videos one clip at a time — without rebuilding the graph

By aiai-r·Created 8 months ago·Updated about a month ago· 3
FolderBatch Video Queue
    • video_path
    • file_name
    • video_count
    folder
    extension*.mp4
    start_at0
    auto_queuetrue
    sort_byName
    order_byA-Z
    video_count0
    progress0.000

    Video work in ComfyUI is where batch tooling stops being a luxury and becomes the whole job. You don't process one clip and call it done - you've got a folder of source footage to upscale, a batch of clips to run through frame interpolation, or a stack of renders to convert. FolderBatch Video Queue is the iterator for exactly that: it scans a folder for videos, hands you one video_path per execution, and auto-advances through the list so your graph runs once and then just... keeps going.

    It works identically to its image sibling, and that's a feature. On first run it globs the folder, sorts the results, and caches the list. Each execution returns the clip at start_at. The frontend then bumps start_at and, with auto_queue on, queues the next prompt - land on the final clip and it resets to zero. Run, advance, re-queue, repeat, until the folder's done.

    Inputs that matter: folder takes an absolute path (paste it; there's no picker). extension defaults to *.mp4 and takes glob patterns - separate several with semicolons if you have mixed containers. start_at is your resume point. sort_by (Name / Date / Random) with order_by (A-Z / Z-A) controls the order, and Date is your friend for "process newest first." The video_count and progress optional inputs are UI display only, updated by the node's own frontend.

    Outputs: video_path (full path of the current file), file_name (base name without extension - ideal for deriving output filenames), and video_count (total matches). The canonical wiring is video_path into FolderBatch Load Video Frames, whose images batch then feeds whatever per-frame processing you're doing.

    Installing is the pack-level story, so it's short: ComfyUI Manager, search "ComfyUI-FolderBatch", restart. Or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/aiai-r/ComfyUI-FolderBatch
    

    No model downloads, no extra Python deps declared - the pack leans on PyAV, which ComfyUI already ships.

    The gotchas are the queue-family ones. The file list is cached per node instance keyed on folder + extension + sort order, so clips added mid-run won't appear until the cache resets. auto_queue requires the web UI - it's literally the frontend calling queuePrompt(), so pure API/headless runs won't self-advance unless you drive start_at. And if the folder's empty you get an empty string and a zero count rather than a crash. One video-specific note: this node only hands you paths; the actual frame decoding happens downstream in the loader, so a long clip's memory cost lands there, not here.

    CategoryFolderBatch/Video

    Inputs (8)

    NameTypeDefaultDescription
    folderSTRING
    extensionSTRING*.mp4
    start_atINT0
    auto_queueBOOLEANtrue
    sort_byCOMBOName3 options: Name, Date, Random
    order_byCOMBOA-Z2 options: A-Z, Z-A
    video_countoptINT0
    progressoptFLOAT0.0000–1

    Outputs (3)

    NameTypeDescription
    video_pathSTRING
    file_nameSTRING
    video_countINT