Voicebox
Voicebox is a local-first, open-source voice studio that clones voices, generates speech across seven TTS engines, transcribes dictation into any app, and gives MCP-aware AI agents a voice of the user's choosing.
It positions itself as a self-hosted alternative to ElevenLabs and WisprFlow combined, covering both voice output and voice input in one app. Supported capabilities include zero-shot voice cloning from a short reference sample, more than fifty preset voices, twenty-three languages, post-processing effects such as pitch shift and reverb, expressive delivery tags, unlimited-length generation with automatic chunking, and a multi-track editor for assembling conversations or narrated pieces.
A global dictation hotkey and an in-app microphone provide speech-to-text in any text field, and a REST API plus a built-in MCP server let an agent such as Claude Code, Cursor or Cline call a voice-output tool or a bundled local LLM for persona-based rewriting of what it says.
Built with Tauri (Rust); runs on macOS, Windows, Linux and Docker. TypeScript, MIT licensed.