Abstract hard silhouette edge diagram for a local background remover mask notch.
Cutout-edge diagram: hard subject silhouette separated from a checker backdrop. Cedar Bind original geometry for factory-background-remover.

Install rembg 2.0.84

Windows

Paste into :

rembg-cli-cpu-installer.exe

Package: GitHub Releases rembg-cli-cpu-installer.exe (no winget package verified for this guide). Prefer the CPU installer; rembg-cli-gpu-installer.exe is the NVIDIA/CUDA alternate on the same tag..

Or download the file directly from the official GitHub release:

Download Windows CPU CLI installer rembg-cli-cpu-installer.exe · 114 MB · September 8, 2026

Also on this release: Windows GPU CLI installer

Windows 10 or 11, 64-bit. Prefer rembg-cli-cpu-installer.exe from the latest v* tag on danielgatis/rembg; use rembg-cli-gpu-installer.exe when you already have a working CUDA stack.

macOS

Paste into Terminal:

pip install "rembg[cpu,cli]"

Package: PyPI rembg extras (cpu,cli). No native macOS installer is attached to the GitHub release assets verified for this guide..

Or read the official setup notes:

Open the macOS install docs PyPI rembg[cpu,cli] install notes

Python >=3.11 and <3.14 per rembg README. Install the CLI extras with pip; there is no macOS .dmg on the release tag checked for this guide.

Linux

Paste into your terminal:

pip install "rembg[cpu,cli]"

Package: PyPI rembg extras (cpu,cli). GPU users may follow README rembg[gpu,cli] after onnxruntime-gpu is available; no native Linux installer is on the release assets verified here..

Other Linux commands
GPU CLI extras 
pip install "rembg[gpu,cli]"

Or read the official setup notes:

Open the Linux install docs PyPI rembg[cpu,cli] install notes

Python >=3.11 and <3.14 per rembg README. Prefer rembg[cpu,cli] unless you already meet onnxruntime-gpu requirements.

Windows PC

On Windows 10 or 11, download rembg-cli-cpu-installer.exe from GitHub Releases and run it. Use rembg-cli-gpu-installer.exe only when CUDA and onnxruntime-gpu already work on this box. Guide: Windows install.

macOS

There is no macOS installer on the release assets verified for this guide. Install Python in the supported range, then run pip install "rembg[cpu,cli]". Guide: macOS install.

Linux

Same pip path as macOS for everyday CPU work. GPU extras exist in the README after onnxruntime-gpu is ready. Guide: Linux install.

Locked-down machines

School and corporate images may block unknown exe files or pip installs. Stop rather than force a wrapper from a search ad. Use a machine you control when you need a clean background remover path, then copy outputs back.

Prove the machine

After install, run rembg --help and cut one small file. Document whether this computer used CPU installer, GPU installer, or pip so updates stay consistent. First hour: installed.

Related: download safe, releases, safety.

Disk and network expectations

Model weights need disk space. Plan for downloads on first use unless you prefetch with rembg d. A laptop on metered Wi-Fi should prefetch on office network before a shoot day. That planning is part of choosing a background remover for this computer, not an afterthought.

Battery-only sessions can still cut small files after models are cached. Large folder batches are happier on AC power.

Shared studio PCs

On a shared machine, install to a known path and leave a one-line text file that says CPU installer, GPU installer, or pip. The next editor should not have to guess. Pair that note with the Releases URL so updates stay on the same background remover door.

Avoid installing both the Windows exe and a conflicting pip rembg on the same PATH without documenting which command wins.

Quick decision tree

  • Windows and no CUDA story: CPU installer.
  • Windows with proven CUDA: GPU installer optional.
  • macOS or Linux: pip CPU CLI extras.
  • No admin rights: stop and use a machine you control.