PD_Load Images

PD_Load Images (Load_Images_V1) loads every image in a folder - including subfolders - into a batch, and crucially hands you the file paths along with the pixels. That last bit is the differentiator. Core ComfyUI's Load Image is one image at a time from its own input directory; this node is for when you want to walk a whole folder and still know which file each frame came from.

Why you'd reach for it

Batch processing, basically. You point it at a directory, and out comes a stack of images, their masks, and the source file paths. That path output is gold for anything where the filename matters downstream - pairing images with their .txt caption files, matching by naming convention, or just knowing what you're looking at in a batch preview. It's the natural front end for the "process a folder" workflows this pack is full of, and it pairs well with a LoRA-dataset-prep flow where you load a directory and need names to line up with captions.

How it works

It walks the directory recursively (os.walk), collects image files, and sorts them. The sort_method dropdown matters more than it looks: numeric sorts by embedded numbers (so img2 comes before img10), alphabetic is plain string order, and natural does human-style ordering. Getting this right is the difference between your batch running in a sensible sequence and a nonsense one. Results are cached by default; load_always forces a re-read of the folder, which you want if the folder changes while a workflow is open. image_load_cap (0 = unlimited) and start_index let you grab a slice rather than everything. The seed input suggests the V1 supports some shuffling/selection - it's there if you want reproducibility across runs.

Outputs

• images - the batch of loaded IMAGE tensors.
• masks - the matching alpha masks.
• file_paths - full paths for every loaded image. This is the output the core Load Image doesn't give you.

Installing

This ships in 7BEII/Comfyui_PDuse. ComfyUI Manager → search "Comfyui_PDuse" → install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/7BEII/Comfyui_PDuse
cd Comfyui_PDuse
pip install -r requirements.txt

Restart after. No model downloads; dependencies are Pillow, numpy, opencv-python and the pack's usual list, with torch from ComfyUI.

Gotchas

The folder path is typed by hand, so it's absolute-path land - there's no file browser like core Load Image has. And because results are cached, a folder you edit while ComfyUI is open can look stale until you flip load_always. If you want the names (not just full paths) or per-file numbering as separate outputs, that's what the Advance variant in the same pack adds; V1 keeps it to three outputs and a seed.