LoadImageParam

Stock LoadImage gives you an image, a mask, and a shrug. If that image was saved with an A1111-style parameter block - by the pack's own SaveImageWithParamText, or by A1111/Forge itself - LoadImageParam (class LoadImageParamMittimi) hands you the recipe too: width, height, and the full parameters text pulled straight out of the file's metadata.

What it is

This is the pack's fifth node, and the author never got around to putting it in the README - the docs list four nodes and this one just ships. It's essentially the stock ComfyUI LoadImage with extra outputs bolted on. Same image-upload widget, same image input, same batch handling and EXIF orientation fixing. The difference is the reading: for PNGs it pulls the parameters text entry from the metadata chunk; for WEBP it reads EXIF tag 270. Everything it outputs is derived from the file on disk, so there's nothing else to configure - just pick an image.

The outputs

• image and mask - identical in behavior to stock LoadImage, alpha channel and all.
• width and height - the pixel dimensions as INTs. This is the sneaky-useful part: wire these into a resolution switcher or a conditioning/size node and you can drive an img2img or upscale pass straight from whatever image you loaded, no manual resizing.
• parameters_text - the whole parameter string as text, or empty if the file has none. For an image saved by this pack's sibling node, that's the full A1111-style block (prompts, steps, sampler, CFG, seed, model, VAE). You can display it, pipe it into a text node, or parse it however you like.

It returns a STRING, not a parsed structure - so it's a reader, not a settings-applier. You won't get the checkpoint and seed back as separate ports; you get the text block and you work with it from there.

Why you'd use it

Honestly, this node exists to close the round-trip: you save with SaveImageWithParamText, later you reload the image and the metadata comes back out. If you keep a folder of your own outputs, it turns each one back into a self-describing starting point for img2img or a remix. Since the metadata convention it reads is the same one A1111 and its forks write, it also works on images that never touched this pack - which is where it quietly earns its keep over stock LoadImage.

Install

Same pack, same route - ComfyUI Manager (search "ComfyUI_mittimiLoadPreset2") or:

cd ComfyUI/custom_nodes
git clone https://github.com/mittimi/ComfyUI_mittimiLoadPreset2

then restart ComfyUI. Dependency is just toml; no models to download. It loads from your standard input folder like any LoadImage, so drop files there or upload through the widget.

One honest caveat: if a file has no parameters metadata, parameters_text comes back as an empty string, not an error - so don't treat a blank output as a failure, it just means the image wasn't saved with a parameter block. And since the author maintains this one off-documentation, treat it as a "works but lightly maintained" node rather than a guaranteed-stable API.