# voicebox

> Local voice studio for cloning voices, generating speech and dictating into any app

- Repository: https://github.com/jamiepine/voicebox
- Category: Image, video & audio
- Stars: 58,272 · Forks: 7,315 · Language: Python · License: MIT
- Experience: Beginner friendly
- Page: https://ir.adyog.com/repos/jamiepine/voicebox/

Voicebox puts both halves of voice work in one app. Output means turning text into speech, including in a cloned voice. Input means speaking and getting text back. The README presents it as a free, open-source alternative to ElevenLabs for output and WisprFlow for dictation.

You give it a short audio sample, and it clones the voice using one of seven bundled TTS engines, such as Qwen3-TTS, Chatterbox, Kokoro and HumeAI TADA. Each engine has different strengths in language coverage, speed and size, and you can switch between them per generation. Dictation uses a global hotkey and OpenAI's Whisper. An optional bundled local LLM can tidy up the transcript before it is pasted.

It also has a REST API and a built-in MCP server. An agent such as Claude Code, Cursor or Cline can call a single tool, `voicebox_speak`, and talk to you in a voice you chose. The app is built with Tauri rather than Electron. It runs on Apple Silicon (MLX), Windows (CUDA), AMD ROCm, Intel Arc and Docker. Linux binaries are not yet available, so Linux users have to build from source.

## What it does

- **Multiple TTS engines and voice cloning**: Seven engines are switchable per generation. Cloning works from a short reference sample, and 50+ preset voices need no sample. The engines cover 23 languages in total.
- **Global dictation**: A global hotkey supports push-to-talk and toggle modes. On macOS the transcript is pasted into the focused field. Whisper handles transcription, and an optional local LLM removes filler words.
- **Agent voice output**: Agents call `voicebox_speak` over MCP, or `POST /speak` for anything that doesn't speak MCP. They speak in a voice profile you have cloned.
- **Long-form generation and Stories editor**: Long text is split at sentence boundaries, generated in chunks and crossfaded, up to 50,000 characters. A queue runs jobs one at a time. A multi-track Stories editor builds podcasts and conversations.
- **Effects, versions and captures**: Eight effects built on Spotify's pedalboard library, with four presets. Each generation keeps its original and its effect versions. Captures store every recording alongside its transcript.

## What you could build

- Clone your own voice to narrate scripts, articles or book chapters.
- Dictate into any app with a hotkey, with audio and transcripts kept on your own machine.
- Give a coding agent a spoken voice so it announces finished tasks and asks questions aloud.
- Produce a multi-voice podcast or dialogue in the Stories timeline editor.
- Call the local REST API from your own app or script to add speech generation or transcription.

## Who it's for

Suits creators, developers and privacy-minded users who want voice tools offline. The desktop app needs no coding, but the API and MCP features assume some developer familiarity, and GPU hardware helps.

## Get started

Download the installer for your platform from voicebox.sh or the GitHub releases page: a DMG for macOS or an MSI for Windows. You can also run the Docker command. Linux has no pre-built binaries yet, so follow the build-from-source guide on voicebox.sh/linux-install.

```
docker compose up
```

Tags: voice-cloning, text-to-speech, whisper, mcp, local-first