jamiepine/voicebox

Voicebox solves the fragmentation of voice AI tools by integrating voice cloning, TTS, dictation, and agent voice output into one local app, eliminating reliance on cloud services for privacy. It also bridges the gap between input (dictation) and output (speech) with a unified API and MCP server for AI agents.

Developers can build applications that clone voices from seconds of audio, generate speech in 23 languages across 7 TTS engines, and enable dictation into any text field. Real-world use cases include creating custom voice assistants, generating audiobooks or podcasts with multi-track editing, and giving AI agents (like Claude Code or Cursor) a personalized voice via MCP. The ceiling includes expressive speech with paralinguistic tags, unlimited-length audio generation, and full voice I/O integration into any app through REST and MCP APIs.