Izwi

Izwi is a local voice AI app and runtime for people and teams processing speech on their own hardware or inside infrastructure they control. It brings audio tools and voice conversations into a desktop app, browser interface, CLI and inference server.

• Transcription and speakers: Transcribe audio, identify separate speakers with diarization, and align text with speech using forced alignment.
• Speech production: Generate speech with text-to-speech models, clone or design voices, save voices for reuse, and work on longer Studio projects.
• Voice conversations: Combine local speech recognition, chat and speech generation for real-time spoken exchanges.
• Application APIs: Connect applications through OpenAI-compatible APIs for chat, transcription and speech generation.
• Model management: Download and manage models locally, including Whisper and Parakeet for recognition, Qwen3-TTS and Kokoro-82M for speech, Sortformer for diarization, and Qwen3 and Gemma for chat.

The runtime runs on macOS, Windows and Linux without cloud services or API keys, and inference data stays local. Optional desktop analytics are off by default and exclude prompts, transcripts and audio payloads. The runtime is open source under MIT; model weights and bundled assets can have separate licenses or access terms.

Izwi also provides paid private LLM deployment and managed operations services. Those deployments run in a customer's cloud account or data center, separately from the runtime you run on your machine.