LoadImageList☀

ComfyUI's built-in LoadImage picks one file from a dropdown. When your job is "process all of these," you need something that reads a folder and hands you a batch - which is what LoadImageList does. It's the image twin of the pack's LoadTextList, and the two are meant to work together for folder-to-pipeline work.

How it works

Give it a folder, and it globs *.png, *.jpg, *.jpeg, *.gif, and *.bmp, sorts them by your chosen mode, slices them with start_index and max_output, and loads each into an IMAGE tensor. You get two list outputs: images (one tensor per file) and file_name (just the basenames). It skips files that fail to load instead of aborting, and logs a [DaShuai] message for each one it drops.

The sort modes are the real selling point. sort_mode gives you None, Alphabetical (ASC/DESC), Numerical (ASC/DESC), and Datetime (ASC/DESC, by file mtime). If your dataset has numbered frames or a meaningful file order, setting this is the difference between "in order" and "whatever the filesystem felt like that day."

Inputs

• folder_path - where the images live. Must exist, or the node raises.
• max_output - cap on how many to load; 0 means all of them.
• start_index - skip the first N files.
• sort_mode - the 7-way dropdown above.
• always_reload - same story as LoadTextList: accepted, but this node never caches, so it always re-reads the folder fresh. The toggle is effectively a no-op in the current code.

Outputs are list-typed IMAGE and STRING, so they feed nodes that take image lists (for example, feeding a batch through an upscaler, a captioner, or an image-to-image sweep).

Gotchas

• Numerical sort extracts digits from the filename. Files without any digits sort as zero, which can scramble a mixed naming scheme.
• No VAE-encoded latents, no masks. This loads plain RGB tensors - if you need latent batches or alpha channels, look elsewhere.
• This is a folder node, not a file-picker. It can't see ComfyUI's input folder dropdown list; it takes a raw path string. For absolute paths on a network share, that's actually the point.

Install

Same pack as the rest of DashuaiTools:

cd ComfyUI/custom_nodes
git clone https://github.com/Hasasasa/ComfyUI_DashuaiTools

then restart ComfyUI, or use ComfyUI Manager and search "DashuaiTools". It depends only on Pillow/numpy/torch, all already present in ComfyUI.

If you're batch-processing a folder of images - upscaling, captioning, img2img runs, before/after comparisons - this is the loader that makes the batch a one-click thing.