One node pack to cut them all: background removal for ComfyUI

What it does ComfyUI-RMBG wraps a growing list of background-removal and segmentation models—RMBG-2.0, BiRefNet, BEN2, SAM variants, GroundingDINO, and others—into a single set of ComfyUI custom nodes. It started as a simple background remover but has expanded into a general segmentation workbench with face parsing (19 categories), clothes segmentation (18 categories), fashion segmentation, text-prompted object detection, and a pile of utility nodes for stitching, cropping, masking, and compositing. The README reads like a changelog on fast-forward, which tells you exactly how quickly new models get bolted on.

The interesting bit The project treats model diversity as a feature, not a bug. Rather than betting on one architecture, it gives you a buffet of matting and segmentation approaches—matting models for hair-level detail, SAM for promptable masks, BiRefNet-HR for 2048×2048 resolution—then layers utility nodes on top so you can actually composite the results without leaving ComfyUI. It is essentially glue code, but it is glue code that saves you from managing ten separate repositories.

Key highlights

• Bundles 10+ segmentation and matting models under one interface, including recent additions like SAM3 and SDMatte.
• Specialized nodes for face parsing, clothing, and fashion segmentation with multi-category selection.
• Utility nodes for image stitching, mask overlay, object removal, and batch processing.
• Text-prompted segmentation using tag-style prompts (“cat, dog”) via SAM and GroundingDINO.
• High-resolution support up to 2048×2048 through the BiRefNet-HR model.

Caveats

• The documentation is almost entirely a changelog; architectural details and usage guidance are sparse.
• Recent releases show a pattern of rapid bug fixes for compatibility (transformers v4.49+, i18n errors, SAM3 segmentation bugs, background color handling issues).
• The scope has crept well beyond background removal into general image utilities, so the name is now slightly misleading.

Verdict Worth a look if you do heavy inpainting, compositing, or fashion/face editing inside ComfyUI and want one node pack instead of a scattered plugin collection. Skip it if you need a lightweight, focused background remover or if you prefer calling segmentation APIs directly.