rembg

Project Overview

rembg is a tool for removing image backgrounds, separating the foreground subject from the backdrop with built-in AI models such as U2Net. It runs as a CLI, a Python library, an HTTP server, or a Docker container, and inference happens locally so images never need to leave your machine for a third-party service.

Local AI background removal — one command lifts the subject out of the frame.

Key Features

• One-click removal: feed it any image and it detects the subject, returning a transparent-background result
• Multiple modes: command line, Python library, HTTP server, and Docker container
• Model choice: ships with u2net, isnet, and other models that can be swapped per scene, people or objects
• CPU and GPU: CPU inference by default, with onnxruntime-gpu acceleration for NVIDIA cards
• Alpha matting: optional edge refinement makes fine details like hair blend naturally
• Batch processing: the CLI can process a whole directory of images

Target Audience

• E-commerce and asset production staff
• Content teams batch-processing images
• App developers embedding background removal
• Privacy-conscious designers who need local processing

Quick Start

1. Install the backend that matches your hardware

pip install 'rembg[cpu,cli]'

2. Process a single image from the command line

rembg i input.png output.png

3. Or call it in code

from rembg import remove
from PIL import Image

inp = Image.open('input.png')
remove(inp).save('output.png')

Core Advantages

• Local inference: models run on your machine, images stay inside the network
• Flexible modes: one capability serves personal scripts through production deployments
• Ready out of the box: models download automatically on first run, no manual setup

System Requirements

Notes

• Models download on first run: the model files are large; pre-place them manually if your network is slow
• Results vary by image: for subjects close in color to the background or with complex edges, try a different model or enable alpha matting