Load Image From Directory

ComfyUI's stock LoadImage has a one-file problem: you pick a single image from input/, and that's it. Load Image From Directory is the escape hatch - point it at a folder and it loads every .jpg, .jpeg, and .png in it, handing each one to you as a separate item on a list output. If you've ever batch-processed reference images or fed a folder of sources into a loop, you know the pain this removes.

It's the "individual" flavor of FairLab's folder loading. The sibling Load Image Batch From Directory stacks everything into one batched tensor; this one keeps them as a list of separate images, which is what you want when each image needs to travel to a different place, get its own filename treatment, or feed a node that iterates over list inputs.

Inputs and outputs

• directory (STRING) in - the folder path, absolute, on the machine running ComfyUI.
• recursive (BOOLEAN) - if true, walks subfolders too; default false.
• channels (RGB / RGBA) - which channel layout to load into; default RGB.
• image (IMAGE, list) out - one image per file.
• directory (STRING, list) out - the folder per file.
• name (STRING, list) out - filename per image.

All three lists are index-aligned, so name[i] matches image[i].

How it works

It scans the folder (walking recursively if you ask), filters to image extensions, and converts each file to a ComfyUI IMAGE tensor, with a progress bar so you can watch a big folder churn through. The recursive toggle is the one that changes behavior most: off means a flat folder, on means every nested directory gets swept in, so be deliberate about it on a big tree.

Where it fits

Batch img2img, folder-of-references workflows, dataset inspection, and pairing with the pack's Load String From Directory for caption sidecars. Because names come back on wires, you can also wire the filename into a save node and keep output names matching your inputs.

Install

FairLab is one pack: ComfyUI Manager → search ComfyUI-FairLab, or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/yanhuifair/ComfyUI-FairLab.git
cd ComfyUI-FairLab
pip install -r requirements.txt

Restart, then find it under Fair/image. Loading is done with PIL via the backend, so no extra dependency beyond what ships with ComfyUI.

Gotchas

• The folder must be reachable by the ComfyUI process. In a Docker or remote setup, input/ on your browser machine is not the same filesystem - use a path the server can see.
• If the directory is invalid, the node raises a hard error (folder_path is not valid: ...) rather than silently doing nothing. That's actually the friendlier failure - the message tells you the exact path it rejected.
• Non-image files are skipped quietly. A folder with mixed content loads only the .jpg/.jpeg/.png files.