Load Image (App Mode)

This is a plain ComfyUI LoadImage in a trench coat. The entire point of Load Image (App Mode) is that it inherits ComfyUI's native image uploader - upload, input-directory picker, multi-frame reads, alpha mask, cache checks, all of it - and then adds one behavior the stock node doesn't have: when the node is Bypassed, its panel input folds itself away.

That matters once you package a workflow into an "app" where non-technical users flip nodes Active / Bypass / Mute. With a normal LoadImage, a bypassed node still shows an upload widget and users click it wondering why nothing changes. This node's front-end script (web/app_mode_bypass.js) watches the node's mode and hides the input, then restores the original panel state when you flip back to Active. If you never build app-mode workflows, there's nothing to see here - the stock LoadImage does the same job for free.

How it works

The class literally subclasses ComfyUI's core CoreLoadImage and overrides only the category, description, and app-mode plumbing. That's why every behavior matches the native node exactly: the image input is the same enum of files in your input folder plus the uploader, and the outputs are the same two you already know.

• IMAGE - the loaded tensor, B,C,H,W. Goes to a VAE Encode, an image-to-image sampler, an img2img conditioning chain, whatever takes an IMAGE.
• MASK - the alpha channel as a mask. This is what you want for inpaint workflows or any place you need the image's transparency as a selection.

The input that matters

Only one, and it's the one you set everywhere: image - pick a file from the dropdown or hit upload. That's the whole input surface. The Mask output appears only when your source actually has an alpha channel, same as stock LoadImage.

How to install it

It ships in the DAELab pack, so install the pack once and this node is in your palette. ComfyUI Manager: search "ComfyUI-DAELab-Custom-Nodes-Library". Or the manual route:

cd ComfyUI/custom_nodes
git clone https://github.com/TeutonicMepheles/ComfyUI-DAELab-Custom-Nodes-Library

then restart ComfyUI. No models, no requirements file - this node needs nothing beyond ComfyUI itself. Two gotchas worth knowing up front: it's a legacy-style node (it subclasses ComfyUI's native LoadImage), so it works on any reasonably current ComfyUI, though the pack also ships schema-based nodes that need the newer comfy_api API - if your log says No module named 'comfy_api', update ComfyUI Core rather than pip-installing a random package. And clone the whole repo - the front-end scripts live in web/, and the pack loads them via WEB_DIRECTORY.

Common issues & troubleshooting

The node doesn't show up. If you previously had any of these nodes installed from an older, scattered location, the pack skips registration on duplicate node IDs. Remove the old copies and restart.

The Bypass behavior does nothing. App-mode input folding only kicks in when the workflow is being treated as an app (the node is selected as an app input). On a normal canvas, Bypass behaves exactly like stock ComfyUI - which is the intended design, not a bug.

No MASK output. Your image has no alpha channel. Use a PNG with transparency or one of the background-removal routes in the KB if you need an alpha.