AQ_LoadImageBase64

ComfyUI's normal Load Image reads files. This one reads a base64-encoded image string - which is how images come back from web APIs, HTTP endpoints, and other nodes that hand you data instead of paths. It's the glue for API-driven workflows: something out there produces an image as text, and this node makes it a real tensor inside your graph.

How it works

It base64-decodes the string, opens it with PIL, and converts it to the same float tensor format the core loader produces. The mask output is where it gets interesting: if the image has an alpha channel, the alpha becomes a MASK - 1 where the pixel is fully transparent, 0 where it's opaque, matching ComfyUI's convention that 1 = region to affect. If there's no alpha channel, it returns a dummy 64×64 zero mask instead.

Inputs

• image - a base64 string, single line.

Outputs

IMAGE, and MASK (derived from alpha, if present).

Install

Ships in AQnodes:

cd ComfyUI/custom_nodes
git clone https://github.com/2frames/ComfyUI-AQnodes
cd ComfyUI-AQnodes
pip install -r requirements.txt

or search "AQnodes" in ComfyUI Manager and restart.

Gotchas

The mask output is only meaningful for images that actually carry alpha. Feed it a JPEG and you get a 64×64 zero mask - which is not the same size as the image, so wiring it into inpainting will produce something broken or confusing. If you need a real mask, feed an RGBA PNG. And there's no validation: a malformed base64 string just throws a decode error, so make sure whatever you're piping in is actually valid base64 of a decodable image. For the API round-trip use case, this pairs naturally with the pack's save/return side - encode on the way out, decode back in here.