Nodes/IG Interpolation Nodes/πŸ“‚ IG Load Images
ComfyUI Node

πŸ“‚ IG Load Images

Load a whole folder of frames as one batch β€” with masks to match

By IDGallagherΒ·Created 3 years agoΒ·Updated about a year agoΒ· 3
πŸ“‚ IG Load Images
    • IMAGE
    • MASK
    • INT
    β—„folderβ€”β–Ί
    β—„image_load_cap0β–Ί
    β—„skip_first_images0β–Ί
    β—„select_every_nth1β–Ί

    IG Load Images is the pack's batch loader: point it at a folder and it reads every supported image file in it, sorted, and hands you a single batched IMAGE tensor, a stacked MASK tensor, and a count. It's the standard ComfyUI "Load Images from folder" workflow, which is why this is one of the few IG nodes with any search impressions at all - people find it while looking for exactly that.

    The useful twist is the extras. Three optional inputs control which images you get:

    • image_load_cap - stop after N images (0 = no limit).
    • skip_first_images - drop the first N from the sorted list.
    • select_every_nth - keep every Nth image (1 = keep all). Combine this with the cap and you've got a quick frame-sampling tool for a video dump.

    Sorting is lexical, so the usual advice applies: zero-pad your filenames (frame_001.png, not frame_1.png) or the order will surprise you. It honors EXIF orientation on load, and if a file has an alpha channel, that alpha becomes a proper mask - 1 - alpha, which is the ComfyUI mask convention where the mask marks what to inpaint or remove. Images without alpha get a dummy 64Γ—64 zero mask so the batch shapes always match. Outputs are IMAGE, MASK, and an INT count of how many were loaded.

    One gotcha to file away: the output is a single batched tensor, not a list of tensors. Several other nodes in this pack - the stitchers in particular - declare INPUT_IS_LIST and want a Python list of images. If you feed them a Load Images batch directly you'll get a type error, so plan to split the batch first. That mismatch has bitten more than one person, and the README being literally :) doesn't help.

    It also does a couple of things under the hood that matter for re-runs: IS_CHANGED hashes every file in the folder, so ComfyUI re-executes the node when the folder's contents change, and it raises a clear FileNotFoundError if the folder doesn't exist or has no images - which is the failure mode to check first when nothing loads.

    Install is the pack standard:

    cd ComfyUI/custom_nodes
    git clone https://github.com/IDGallagher/ComfyUI-IG-Nodes
    cd ComfyUI-IG-Nodes && pip install -r requirements.txt
    

    Or via ComfyUI Manager - search "IG Interpolation Nodes" and Install, then restart. No model downloads; this node is pure image I/O. It pairs naturally with IG Folder (to build the path) and IG Analyze SSIM (to chart the frames you just loaded).

    CategoryπŸ“ IG Nodes/IO

    Inputs (4)

    NameTypeDefaultDescription
    folderSTRINGβ€”
    image_load_capoptINT0β€”
    skip_first_imagesoptINT0β€”
    select_every_nthoptINT1β€”

    Outputs (3)

    NameTypeDescription
    IMAGEIMAGEβ€”
    MASKMASKβ€”
    INTINTβ€”