Load input/output image

Load input/output image is the pack's take on ComfyUI's stock Load Image, with one addition that matters more than it sounds like: it can read from the output folder, not just the input folder. Output files show up in the dropdown tagged with [output], so a workflow can pick up an image your own previous pass just saved.

That's the whole reason it exists, and it only makes sense in the automation context the pack is built for. The README's flagship flow is a multi-pass high-res pipeline: generate at ~1MP, HiRes-fix it, then upscale - and each pass saves to output/ for the next one to consume. Stock Load Image is blind to all of that; you'd have to manually copy files into input/ between passes, which is the exact chore automation is supposed to kill. This node closes the loop.

How it works

It behaves like stock ComfyUI Load Image under the hood - EXIF transpose, alpha channel converted to a mask, a blank 64×64 mask when there's no alpha, multi-frame GIF support - and then adds two things:

• A image combo input that merges input and output filenames, newest first.
• A refresh button that calls the pack's own /ig1api/images API endpoint (registered on ComfyUI startup) to re-scan both folders without restarting anything.

Outputs are output_image (IMAGE) and output_mask (MASK), same as the built-in node. The node is flagged experimental in the source, so treat it as "works fine, may change shape."

Where people get tripped up

• The refresh button needs the pack's API route alive. The endpoint is registered by the pack when it loads, so if the pack failed to install or register, the refresh does nothing and you won't see [output] files at all. First diagnostic: did the IG1 Tools category appear? No → the pack isn't loaded, nothing else matters.
• The image input is a plain combo, not a search box. There's no tooltip on it, and the list can get long in a busy output folder - the [output] tag is your friend when scanning.
• It's for round-tripping your own saves. If your workflow never reads back from output/, the stock Load Image does the identical job and you don't need this node. Don't install the pack for this one alone.

If you're building the pack's advertised generate → HiRes → upscale automation, though, this is the node that lets pass two and three pick up where the previous pass stopped, no file juggling in between. It slots in wherever you'd otherwise fight the stock loader's input-folder-only blind spot.

Install

Same as the rest of the pack - it's one repo: ComfyUI Manager → "Install via Git URL" → https://github.com/iguanesolutions/comfyui-ig1-tools.git (or search "Flux Resolution"), then restart ComfyUI. No model files, no heavy dependencies - the requirements (requests, aiohttp, torch, numpy, Pillow) are already satisfied by a stock install. Find it under IG1 Tools after the restart.