voicebox: From Voice Cloning to Agent Conversations, All Locally Without the Cloud

It unifies voice output and input features previously handled separately by ElevenLabs and WisprFlow. Clone a voice with just a few seconds of audio samples or choose from over 50 presets to generate natural-sounding speech in 23 languages. All processing is done locally, ensuring voice data is never sent to external servers.

1 / 3

Switch between 7 TTS engines, including Qwen3-TTS, Kokoro, and Chatterbox, depending on the situation. Selecting Chatterbox Turbo recognizes emotion tags like [laugh] and [sigh] to produce expressive dialogue. With 8 audio effects based on Spotify's pedalboard library and automatic chunking, even long scripts are synthesized seamlessly without interruptions.

Press a global hotkey to dictate text into any app, with Whisper-based STT converting it instantly. It includes a built-in MCP server, allowing AI agents like Claude Code or Cursor to respond using a specified voice. Built with Tauri and Rust, it delivers lightweight native performance compared to Electron.