Before choosing an install path, check the Runtime Support Matrix for the current backend contract. In particular:
  • GitHub Release artifacts are not the same thing as CUDA source builds.
  • Linux and Windows GitHub Release artifacts are CPU-only and do not bundle CUDA runtime libraries.
  • CUDA acceleration requires a compatible NVIDIA driver and GPU; source builds still require the CUDA toolkit.
  • The Docker CUDA profile is the CUDA distribution path for NVIDIA Linux hosts and may require CUDA_COMPUTE_CAP when built on a machine without nvidia-smi.

​
Platforms

PlatformGuide
macOSmacOS Installation
LinuxLinux Installation
WindowsWindows Installation
From SourceBuild from Source

​
Quick Install

​
macOS

# Download and install the .dmg from GitHub Releases
open https://github.com/izwi-ai/izwi/releases

​
Linux (Debian/Ubuntu)

# Download the .deb package
wget https://github.com/izwi-ai/izwi/releases/latest/download/izwi_amd64.deb
sudo dpkg -i izwi_amd64.deb

​
Windows

Download and run the installer from GitHub Releases.

​
Verify Installation

After installation, verify everything is working: 
izwi version --full
You should see the compiled backend list for the installed CLI. After you start the server with izwi serve, use izwi status --detailed to verify runtime backend selection. If you specifically need NVIDIA CUDA support, see:

​
Optional: espeak-ng for Kokoro-82M

If you want to use the Kokoro-82M TTS model, you also need espeak-ng installed separately. See your platform guide for commands:

​
Next Steps

macOS Installation

Install Izwi on macOS with the DMG, optional CLI tools, and Apple Silicon acceleration.

Linux Installation

Install Izwi on Linux with release packages, source builds, services, and CUDA guidance.

Windows Installation

Install Izwi on Windows with the installer, optional eSpeak NG, CUDA notes, and source build setup.

Build from Source

Build Izwi from source for development, custom backends, and release packages.