ComfyUI-RMBG Project Overview

Core facts

Read by a script from the official API, not rewritten by a model.

Repository: https://github.com/1038lab/ComfyUI-RMBG
License: GPL-3.0
Languages: Python
Maintainer: 1038lab
Homepage: https://1038lab.github.io/ComfyUI-RMBG/
Verified: 2026-10-31

What it is

ComfyUI-RMBG is a ComfyUI custom node for image background removal and segmentation of objects, faces, clothing, and fashion elements. It requires ComfyUI. It requires Python 3.10 or higher.

The required packages are installed automatically. It is licensed under the GPL-3.0 License.

Main features

The RMBG node supports the models RMBG-2.0, INSPYRENET, BEN, and BEN2 for background removal.
The RMBG node supports batch processing.
The Segment node provides text-prompted object detection.
The Segment node supports both tag-style and natural language inputs.
SAM2 Segmentation provides text-prompted segmentation with SAM2 models (Tiny, Small, Base+, Large).
SAM2 Segmentation downloads models automatically on first use, with a manual download option.
SAM3 Multiplex Segmentation offers merged or separate multi-segment output modes, confidence thresholds, and segment picking.

Getting started

One installation method is to search for the node in ComfyUI-Manager and install it.
Another installation method is to clone the repository to the ComfyUI custom_nodes folder.
Another installation method is to install via Comfy CLI.
Most models can be downloaded automatically on first use, while VRMBG-3.0 requires manual download due to official BRIA license terms.
In basic usage, the RMBG (Remove Background) node is loaded from the AILab RMBG category.
In basic usage, an image is connected to the input.

From the README · bash

  ./ComfyUI/python_embeded/python -m pip install -r requirements.txt

From the README · bash

  cd ComfyUI/custom_nodes
  git clone https://github.com/1038lab/ComfyUI-RMBG

From the README · bash

  ./ComfyUI/python_embeded/python -m pip install -r requirements.txt

From the README · bash

  comfy node install ComfyUI-RMBG

From the README · bash

  ./ComfyUI/python_embeded/python -m pip install -r requirements.txt

Who it is for

In basic usage, a model is selected from the dropdown menu. The sensitivity parameter controls the background removal sensitivity, with a range of 0.0-1.0. The process_res parameter sets the processing resolution, from 512-2048 in steps of 128.

The mask_blur parameter sets the blur amount for the mask, from 0-64. The mask_offset parameter adjusts mask edges, from -20 to 20. The background parameter chooses the output background color.

The invert_output parameter flips the mask and image output. The RMBG node outputs a MASK, which is a binary mask of the foreground. ComfyUI-RMBG was updated to v3.3.0 on 2026/10/20.

Version v3.3.0 added native FP16 acceleration for RMBG-2.0 and BiRefNet with automatic intermediate VRAM clearing. Version v3.2.0 added native Apple Silicon (MPS) GPU acceleration for macOS.

Related tools listed on this site.