LoadImageParamLite

Load an image you generated with this pack and LoadImageParamLite does something stock LoadImage won't: it reads the recipe that was baked into the file. Alongside the usual image and mask outputs, it hands you width, height, and parameters_text - the A1111-style parameter string that SaveImageWithParamTextLite wrote when the image was saved.

Mechanically it's simple. It opens the PNG (or WebP), looks for the parameters text chunk (or the matching EXIF tag in WebP), and strips the parameters: prefix so you get the clean recipe text. If the file has no embedded recipe - a random image off the internet, say - you get an empty parameters_text and the other outputs still work fine, so it degrades gracefully into a normal image loader.

The genuinely useful part is that this completes a round trip. Generate with a preset, save with SaveImageWithParamTextLite, and the file is now a portable settings card: load it later, grab parameters_text, and you can see exactly what produced it - or paste that text into a prompt field when you want to recreate the look. It's the same idea behind A1111's PNG Info tab, applied to this pack's files.

Two implementation details worth knowing. It watches the file, not just the widget: the node hashes the image contents and re-runs when the file changes, so swapping a new image into the input folder updates the outputs without you touching the node. And it handles multi-frame images - animated PNGs and similar sequences come through as a batch with a mask per frame, same as the core loader.

The honest caveat: parameters_text is a string, not magic. Nothing in this pack auto-applies the recipe to a new generation - you still feed the prompt back into the preset hub by hand. But as a way to pull settings out of your own archive without hunting through old workflow files, it's exactly the piece the pack's save side was missing.

Install via ComfyUI Manager (search "ComfyUI_mittimiLoadPresetLite") or git clone https://github.com/mittimi/ComfyUI_mittimiLoadPresetLite into custom_nodes, then restart. Only dependency is toml; no models to download.