ScarlotSoft Load Image

Every img2img workflow starts the same way: get a picture into the graph. ComfyUI's core LoadImage does this fine, and ScarlotSoft Load Image does it with a custom upload button, drag-and-drop, and the suite's dark theme. Functionally it's the same node - dropdown of files in your ComfyUI input folder, plus an upload control - but the JS here replaces the plain file picker with a styled "Upload Image" button and a hover state that makes it obvious where to drop a file. Small UI polish, same job.

What's actually worth reading about is what it does with your file, because it's the good kind of load node. It pulls the image, applies EXIF orientation correction (that one matters when you drag in a phone photo), converts to RGB, and scales to a 0–1 float tensor. Then the interesting part: if the image has an alpha channel, it extracts that as a MASK output - inverted to ComfyUI's polarity, where a mask value of 1 means "paint here." So a PNG with transparency gives you a ready-to-use cutout mask for free. No alpha channel? You get an empty mask instead of a crash, and the pipeline keeps running.

It also handles multi-frame files - GIFs and animated images iterate over every frame and come out as a full batch on the IMAGE output, with matching masks. That's a genuinely useful behavior for frame-stacking or video-ish inputs, and it's why the outputs are image and mask, in that order, both batch-aware.

Two honest notes. First, this reads from the standard input directory like the core node - drop files into ComfyUI/input/ or use the upload button; there's no extra folder to configure. Second, it's a loader, not a metadata reader - if you want the workflow text out of a PNG, the suite's separate MetadataReader is that node. This one is pure pixels-plus-alpha.

Install with the pack - ComfyUI Manager → search ScarlotSoft → install → restart:

cd ComfyUI/custom_nodes
git clone https://github.com/scarlotsoft/ComfyUI_ScarlotSoft

No dependencies, nothing to download. Suite theme applies; hard-refresh after a UI-mode switch if the button looks off.

Should you switch from the core loader? If you're building ScarlotSoft graphs, yes - consistency and the nicer upload affordance. If you're not, the core LoadImage does everything this does except the custom button and the GIF batch handling. The GIF→batch behavior is the one real reason to reach for this over the core node: the core loader is single-image, and if you regularly feed animations into your graphs, the frame batching here is a quiet superpower.