Nodes/ComfyUI_Isulion Random Prompt Generator/๐Ÿ“ Isulion Load Images from Directory
ComfyUI Node

๐Ÿ“ Isulion Load Images from Directory

Feed a whole folder into a workflow โ€” the Isulion directory loader explained

By IsulionยทCreated 2 years agoยทUpdated 7 days agoยท 41
๐Ÿ“ Isulion Load Images from Directory
    • images
    โ—„directory./inputโ–บ
    โ—„target_row_height300โ–บ

    Here's the thing about the Isulion Load Images from Directory node: it's the odd one out in a pack of prompt generators. Everything else in ComfyUI_Isulion spits out text; this one loads image files. It exists because the pack's Image Collage node needs a batch of images, and the pack author got tired of wiring up twenty individual Load Image nodes. You load a folder, get a batch tensor, done.

    It's a straightforward utility, but there's a subtle design choice in how it processes images that's worth understanding before you blame it for bad results.

    How it works

    Point it at a directory and it reads every .png, .jpg, .jpeg, .bmp, .gif, or .webp in there, sorted by filename so the order is stable. Each image gets its EXIF orientation corrected (so phone photos don't come out sideways), converted to RGB, and resized to a target row height while keeping its aspect ratio. Then the clever bit: images of different widths can't be stacked into one tensor, so it pads the shorter ones with black to match the widest, and stacks the lot into a single batched IMAGE (B, H, W, C).

    That padding is the thing to remember. Your images all come out the same height, but if one is a wide panorama and another is a portrait, you're getting black bars. For the pack's own collage node that's exactly what it wants - uniform rows to arrange. For feeding a batch into img2img or an IPAdapter, the bars might surprise you.

    The inputs that matter

    Only two, and you'll mostly touch the first:

    • directory - default ./input. This resolves through ComfyUI's own folder logic, so a bare name lands in ComfyUI/input/, and an absolute path works too. Drop your folder inside ComfyUI/input, type its name, and you're set.
    • target_row_height - default 300, range 100โ€“1024. This is the height every image gets resized to. Lower = smaller batch tensor, faster processing.

    Output

    One output: images (an IMAGE batch). It feeds straight into any node that takes a batch - the Isulion Image Collage from the same pack, a sampler, or batch-aware postprocessing.

    Installing it

    Same story as every Isulion node - it ships in the one pack:

    1. ComfyUI Manager โ†’ Custom Nodes Manager โ†’ search Isulion โ†’ Install โ†’ Restart.
    2. Or clone it manually:
    cd "your_ComfyUI_install_dir/custom_nodes"
    git clone https://github.com/Isulion/ComfyUI_Isulion
    

    No requirements.txt, no model downloads - just the pack, and ComfyUI's own torch/PIL underneath.

    Common issues

    The usual complaints, in order: "Directory does not exist" (typo, or you put it somewhere ComfyUI isn't looking - use a path under ComfyUI/input), and "No images found in directory" (wrong name, or it's a subfolder - the loader does not recurse, only the top level). Because the node hashes the file list to detect changes, adding or removing a file re-triggers the workflow, which is nice but means a busy folder will rerun often. And if your images come out with weird black borders, that's the padding - resize your source set to a consistent aspect ratio or just accept it.

    CategoryIsulion/Image

    Inputs (2)

    NameTypeDefaultDescription
    directorySTRING./inputโ€”
    target_row_heightINT300100โ€“1024โ€”

    Outputs (1)

    NameTypeDescription
    imagesIMAGEโ€”