AI BriefingIzwi 0.1.0 releases fully local voice AI stack for Mac, Linux, and Windows
·2026.10.22 03:17
Key point
The open-source desktop app supports real-time voice conversations, TTS, and ASR using models like Qwen3-TTS and Parakeet TDT v3 without requiring Python.
Details
- Izwi 0.1.0 is the first stable release of a free, Apache-2.0 licensed desktop application for local voice AI, available for Mac, Linux, and Windows.
- The app handles real-time voice conversations (ASR to LLM to TTS), text-to-speech, voice cloning, and long-form audio generation via a Studio interface.
- Inference runs natively in Rust on Candle with custom Metal and CUDA support, eliminating the need for a Python environment.
- An OpenAI-compatible /v1 API is provided for chat completions, audio speech, and transcriptions, allowing existing code to point at localhost.
- Supported models include Qwen3-TTS, Kokoro-82M, Fish Audio S2 Pro, VibeVoice, Voxtral TTS, LFM2.5 Audio, Parakeet TDT v3, Whisper large-v3-turbo, Qwen3-ASR, Granite Speech 4.1, Nemotron 3.5 ASR, Voxtral Mini, Sortformer, Nemotron 3 diarization, and Qwen3.5, 3.6, 3.8, and LFM2.5 for chat.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Respan releases open-source Rust AI gateway with plugin-based architecture
ArtCraft releases open-source IDE for AI image and video creation with 62-model catalog