Mathemagic's ComfyUI Workflows [tools: background remover]
Catgirl because why not.
This post publishes my Tools: background remover workflow, which I occasionally use to extract characters from my gens.
This is a really simple workflow, using this ComfyUI node, which is based on the InSPyReNet technology. This workflow is mainly published for completeness, as well as to raise awareness for this excellent neural background remover, and the ComfyUI node that allows using it in Comfy. I'm not affiliated with either of these.
For other workflows, see:
Qwen workflows, including Qwen-Image and Qwen-Image-Edit
Z-Image Turbo workflows
Chroma1-HD workflows
Illustrious workflows
For more details, see my guide.
Notes
At least in my experience, this model often produces more accurate results than e.g. isnet or u2net (these are offered e.g. by the rembg extension for SD Forge).
The cover is the input image, created with Qwen-Image 2512, with inpainting. The txt2img seed is shown.
The text rendered correctly straight out of txt2img. In inpainting, I removed a spurious line of nonsense text, removed cat whiskers from the character, and then made the clothing details pop in the classical way (0.3 denoise, one region at a time).
This workflow takes in an image, and automatically saves two output images:
The black-and-white image is the foreground mask.
The weird-looking image is the result. It's actually RGBA, with the mask saved in the alpha channel. CivitAI previews seem to display only RGB. To see it correctly, click the image open in CivitAI, or load the PNG in an image viewer app.
The background remover workflow is embedded in both output images.