Izwi

Local first audio inference engine for TTS, ASR, and audio chat workflows.

Rust native runtime with OpenAI compatible API

Generate natural speech from text with support for custom voices and voice design.

Transcribe audio to text with state-of-the-art ASR models running locally.

Chat with multimodal audio models for interactive voice conversations.

terminal

$ izwi serve
  Server running at http://localhost:8080

$ izwi tts "Hello from Izwi" --output hello.wav
  Generated hello.wav