Load Image Batch From Dir

Dragging fifty images into ComfyUI one at a time is a slow way to spend an afternoon. wcx_LoadImagesFromDirBatch is the node for when you have a folder full of files and you want them all as one batch in a single node - for a batch img2img pass, a style transfer run, a comparison set, whatever. It points at a directory and turns its contents into an IMAGE batch, a MASK batch, and a count.

What it actually does, from the source: it lists the directory, keeps the image files (.jpg, .jpeg, .png, .webp - plus .jxl if you've installed pillow-jxl), optionally sorts them, skips ahead by start_index, and loads up to image_load_cap of them (0 means no limit). It transposes EXIF orientation as it goes, so your phone photos come out the right way up - a small touch that a lot of loaders skip. For each image it emits an RGB tensor and a mask built from the alpha channel (inverted per ComfyUI convention, so transparent areas read as masked). Images with no alpha get a tiny 64×64 empty mask. The third output is the INT count of images loaded.

The knobs that matter:

• directory - the only required input. Give it an absolute path; it won't browse your folders for you.
• sort_method - None, Alphabetical, Numerical, or Datetime, each ASC/DESC. Numerical is the one you want for frame_001.png…frame_120.png sequences, since alphabetical would order them frame_1, frame_10, frame_2. start_index applies after sorting, so sorting first means "skip the first N in this order."
• load_always - when on, the node reruns every execution, picking up files added since last run. It does this with the IS_CHANGED → NaN trick the plumbing KB doc warns about: NaN never equals itself, so the cache is defeated. Useful for a folder you're actively dropping files into; annoying if you leave it on, because it poisons the cache for everything downstream of the loader.

The traps are real and mostly silent:

• Mixed-resolution folders get squashed. If the images aren't all the same size, everything after the first is resized (bilinear) to match it. That's often what you want for a batch pass, but it's automatic - feed in a folder with one odd-sized image and it just quietly resizes.
• Missing or empty directory is a hard error. It raises FileNotFoundError rather than passing through an empty batch. A folder that exists but has no valid images does the same.
• The mask is only meaningful for PNGs with transparency. For everything else you get a dummy 64×64 mask. Don't wire it into a mask workflow expecting real data unless your files actually have alpha.

Install is the pack's standard light route:

# ComfyUI Manager: search "ComfyUI-Practical-Tools" → Install → restart

# or manually:
cd ComfyUI/custom_nodes
git clone https://github.com/wenchengxiang/ComfyUI-Practical-Tools
# restart ComfyUI

No model files, no requirements file. For .jxl support you'd add pip install pillow-jxl yourself, but that's optional and most people never need it. This is the node that turns a directory of frames into "one wire" - and with Numerical sorting and start_index, it's a proper batch loader, not just a folder reader.