Load Image MXD

Core Load Image gives you an image and a mask. What it doesn't give you is the prompt that made that image - so when you want to refine, upscale, or img2img your own output, you either retype the prompt or go digging in the workflow metadata. Load Image MXD is the same loader with the metadata plumbing added: pick an image from your input folder and it returns the image, a mask made from alpha, and the positive and negative prompts that were embedded when the image was generated.

What you get

• image - the loader's dropdown, which also accepts uploads (it has the standard image-upload handling).
• IMAGE - the loaded image (handles animated formats too - GIF/webp frames come through as a batch).
• MASK - 1 - alpha from the alpha channel when present; empty mask otherwise. Same inversion convention as the pack's batch loader, so MASK is ready for inpainting use.
• positive and negative - extracted from the file's embedded ComfyUI prompt metadata. Generated an image in ComfyUI, saved it (or dragged it into inputs), load it here, and your prompts come right back out as strings you can wire into your text encoder.

The workflow it enables

The refine loop, made honest: generate at low res, load the result back with this node, and the prompt that produced it is sitting right there in your graph - no copy-paste, no metadata spelunking. Run the same prompt through a sampler with the image as img2img input and you've got an automatic self-refining pass. Pair it with the pack's batch loader if you want the whole folder at once; this is the single-image version.

Honest notes

• The mask only means something when the image has an alpha channel. A JPEG (or a flattened PNG) gives you an empty mask, which is fine - you just don't get an inpainting region for free.
• Prompts are only as good as the metadata. If the file was saved without workflow/prompt metadata (some tools strip it), positive/negative come back empty. ComfyUI's own saves include it, so for self-generated images this basically always works.
• It reads prompts from the first frame for animated formats - the loader is explicit about that in its code - so multi-frame files get one prompt set for the whole batch.

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 image in the node menu. It's a small upgrade to a node you already use every day, and the payoff is never retyping a prompt again.