Load Image Batch (Inputs/Outputs) MXD

The most common real workflow isn't "generate one image" - it's "generate fifty, then upscale/refine/re-run them all." Core ComfyUI makes you load those one at a time. Load Image Batch (From Outputs) MXD loads a whole folder from your outputs directory in one shot and hands back the images as a batch, with two bonuses: a mask derived from each image's alpha channel, and the positive/negative prompts read from each file's embedded metadata.

What you get

• folder - a dropdown of folders under your outputs directory (pick the one with the batch).
• IMAGE (list) - every image in the folder, recursively, sorted.
• MASK (list) - per image: 1 - alpha from the alpha channel, so transparency becomes a proper inpainting-style mask. Images without alpha get an all-black (empty) mask.
• positive and negative (lists) - the prompts embedded in each PNG's ComfyUI metadata. Files generated by ComfyUI carry their workflow/prompt metadata; this node extracts positive and negative from it, so each item in the batch comes with its own prompts aligned by index.

All four outputs are lists, matched by position: images[3] pairs with masks[3], positive[3], negative[3]. That index alignment is the whole point - you can zip the batch together downstream and each image gets processed with the settings it was generated with.

Why this pattern is powerful

It's the batch-reprocess loop: generate a batch with one workflow, then drop Load Image Batch (From Outputs) MXD into a second workflow (upscale, refiner, img2img pass), set the folder, and process everything with per-image prompts restored from metadata. The KB's upscaling guidance keeps landing on "generate low, upscale after" - this node makes that after-pass a batch operation instead of fifty manual loads.

Gotchas

• The folder must exist and contain images, or the node raises a clear error (it does check). Empty folders are a failure mode, not a silent pass.
• Prompts come from embedded metadata. If your images were saved by a non-ComfyUI tool (or with metadata stripped), positive/negative come back as empty strings - the batch still loads, but the prompt pairing is gone.
• Masks come from alpha. An image with no transparency gets an empty mask, so don't expect a useful selection region unless the source had an alpha channel.

Install

ComfyUI Manager → search Maxed Out → install → restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Maxed-Out-99/ComfyUI-MaxedOut

No requirements.txt, no model downloads. Under MXD/Image. If batch refinement has been a manual grind, this node is the conveyor belt you've been missing.