Repolode
Submit
jamiepine

jamiepine/voicebox

Local-first open-source AI voice studio for cloning voices, generating speech, dictating into any app, and giving AI agents a voice.

Open-source alternative to:ElevenLabsWisprFlow

View on GitHubWebsite
Stars
57,983
Forks
7,269
Open issues
515
Last commit
12 days ago

What is it?

Voicebox is a desktop voice studio that runs on your own machine. It clones a voice from a short audio sample, generates speech with one of eight TTS engines across 23 languages, and transcribes speech with Whisper. A global hotkey lets you dictate into the focused text field, with optional local LLM cleanup before the text is pasted.

The app also includes a stories timeline editor for multi-voice conversations, post-processing effects such as pitch shift, reverb and compression, and a captures tab that stores each recording alongside its transcript. Voice profiles can carry a personality that a bundled local Qwen3 model uses to rewrite text in that character's voice.

Voicebox exposes a REST API and a built-in MCP server, so MCP-aware agents such as Claude Code or Cursor can speak, transcribe and list profiles through tool calls. It runs on macOS with MLX, on Windows, Linux and Docker with CUDA, ROCm, Intel XPU or CPU backends.

Key features

  • 01Voice cloning from a short reference sample, plus preset voices from Kokoro and Qwen CustomVoice
  • 02Global push-to-talk dictation with Whisper transcription and optional LLM refinement
  • 03Eight TTS engines switchable per generation, with paralinguistic tags in Chatterbox Turbo
  • 04Eight audio effects via Pedalboard, with reusable presets
  • 05Long-form generation with auto-chunking and crossfade
  • 06REST API and built-in MCP server for agent integration

Use cases

  • Giving a coding agent a cloned voice that reads out task completions and questions
  • Dictating text into any desktop application with a hotkey
  • Producing narration for articles, chapters or podcasts with a cloned or preset voice
  • Building voice features into your own app through the local REST API

Topics

#voice-ai#text-to-speech#speech-to-text#voice-clone#whisper#local-first#mcp#tauri#ai#cuda

FAQ

What is voicebox?
Local-first open-source AI voice studio for cloning voices, generating speech, dictating into any app, and giving AI agents a voice. Voicebox is a desktop voice studio that runs on your own machine. It clones a voice from a short audio sample, generates speech with one of eight TTS engines across 23 languages, and transcribes speech with Whisper. A global hotkey lets you dictate into the focused text field, with optional local LLM cleanup before the text is pasted.
Is voicebox free to use? What license does it have?
voicebox is open source under the MIT license. Check the license terms before using it in commercial products.
Is voicebox still maintained?
Actively maintained. The last commit to voicebox was on Oct 7, 2026.
What can I use voicebox for?
Giving a coding agent a cloned voice that reads out task completions and questions Dictating text into any desktop application with a hotkey Producing narration for articles, chapters or podcasts with a cloned or preset voice Building voice features into your own app through the local REST API
Where can I get voicebox?
The source code and installation instructions are on GitHub: https://github.com/jamiepine/voicebox

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects