jamiepine/voicebox at 26.6K Stars With +462 Today: Open-Source Voice Studio Ships 7 TTS Engines, Zero-Shot Cloning, and MCP Server for Agent Voice Output

GitHub·medium signalXBlueskyLinkedInCopy link

Voicebox is a Tauri desktop app combining voice cloning, 7 TTS engines (Qwen3-TTS, LuxTTS, Chatterbox, Kokoro, HumeAI TADA), speech-to-text via Whisper, and a post-processing chain (pitch shift, reverb, compression) in one local-first package. Key for the agent ecosystem: it ships an MCP server so any AI coding agent can generate speech output, plus 50+ preset voices with expressive paralinguistic tags. Supports MLX on Apple Silicon and PyTorch CUDA/ROCm. For builders adding voice to agent pipelines, this is the most complete open-source voice toolkit available — clone, dictate, and create without leaving your machine.