Getting Started
Get Izwi running in under 5 minutes.
Step 1: Install Izwi
macOS
Download the latest .dmg from GitHub Releases, then:
- Open the downloaded
.dmgfile - Drag Izwi.app to your Applications folder
- Launch Izwi from Applications
On first launch, Izwi will set up the izwi command-line tool automatically.
Linux
Download the .deb package from GitHub Releases:
sudo dpkg -i izwi_*.debWindows
Download and run the .exe installer from GitHub Releases.
See Installation for detailed platform-specific instructions.
Step 2: Start the Server
Open a terminal and run:
izwi serveYou should see:
Izwi server running at http://localhost:8080Desktop mode (opens the native app):
izwi serve --mode desktopWeb mode (opens in your browser):
izwi serve --mode webStep 3: Download a Model
Izwi needs AI models to work. Download your first model:
izwi pull qwen3-tts-0.6b-baseThis downloads a compact text-to-speech model (~1.2 GB).
View all available models:
izwi listView your downloaded models:
izwi list --localStep 4: Try It Out
Generate Speech
izwi tts "Hello! Welcome to Izwi." --output hello.wavPlay the generated audio:
izwi tts "Hello! Welcome to Izwi." --playTranscribe Audio
First, download an ASR model:
izwi pull qwen3-asr-0.6bThen transcribe:
izwi transcribe your-audio.wavStep 5: Open the Web UI
With the server running, open your browser to:
http://localhost:8080The web interface provides:
- Voice — Real-time voice conversations
- Chat — Text-based AI chat
- Transcription — Upload and transcribe audio files
- Text to Speech — Generate speech from text
- Voice Cloning — Clone voices from samples
- Models — Manage your downloaded models