Voicebox
Open-source desktop app to clone voices, generate speech, and dictate text locally.
- Stars
- 59.2k
- Forks
- 7.4k
- Licence
- MIT
- Language
- Python
- Last update
- 30 days ago
- Type
- App
What it is
Voicebox is a free, open-source app that runs voice tools on your own computer. It can copy a voice from a short audio sample and read text aloud in that voice, and it can turn your speech into text to type into other apps. It also offers a local API and an MCP server so AI agents can speak with a chosen voice.
- Clone a voice from a few seconds of reference audio and generate speech in it
- Dictate into any text field using a global hotkey with push-to-talk or toggle modes
- Build long scripts or articles with auto-chunking and crossfade
- Arrange conversations or podcasts on a multi-track Stories timeline
- Let an MCP-aware agent speak to you using the voicebox_speak tool
Great for
- People who want text-to-speech or voice cloning without sending audio to cloud services
- Creators making narration, podcasts, or audiobook-style chapters with long scripts
- Developers who want to add local voice input or output to their own apps through the REST API
- Users of MCP-aware agents such as Claude Code or Cursor who want the agent to speak aloud
Not ideal for
- People who need a simple Linux install today, since the README says pre-built Linux binaries are not yet available and points to build-from-source instructions
- Users with limited hardware, since several engines depend on GPU memory or capable CPUs and the README does not give full system requirements
- Anyone who needs only one narrow feature and does not want a multi-engine studio
How to get it
- 1
Download the app
You'll need: Apple Silicon or Intel Mac, or Windows (Linux requires build-from-source per README)
Run with Docker (as listed in the README download table)
docker compose up - 2
Run it
- Download the DMG for macOS or the MSI for Windows from the Download section, or use Docker
- Install and open Voicebox
- Choose or create a voice profile using a reference sample or a preset voice
- Enter text and generate speech, or set up the dictation hotkey
- 3
Get the most out of it
- Pick the TTS engine per generation: Qwen3-TTS for cloning, Kokoro for fast CPU use, Chatterbox Multilingual for broad language coverage
- Use Chatterbox Turbo if you want paralinguistic tags like [laugh] or [sigh] to be interpreted
- Attach a persona to a voice profile and use Compose, Rewrite, or Respond with the bundled local LLM
- Connect an MCP-aware agent to the built-in MCP server so it can call voicebox_speak
- Use the REST API to add voice input and output to your own applications
Downloads
Strengths
- Runs locally, and the README says models, voice data, and captures never leave the machine
- Offers several TTS engines and 23 languages in one app, switchable per generation
- Free and MIT-licensed, with a REST API and built-in MCP server
Trade-offs
- Setup depends on hardware: engines and GPU support vary by platform, and Linux has no pre-built binaries yet
- The README is truncated, so some details such as full system requirements and some engine specifics are not confirmed here
- Some features, like accessibility-verified auto-paste, are described as macOS-specific
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| voicebox (this) | 59.2k | App | beginner | 30 days ago | People who want text-to-speech or voice cloning without sending audio to cloud services |
| piper | 11.3k | App | intermediate | 1.2 years ago | Developers who want offline speech output in a local app or home automation setup |
| GPT-SoVITS | 64.6k | App | intermediate | 5 days ago | Hobbyists who want to experiment with cloning a voice they have permission to use |
| OpenVoice | 38.1k | AI model / agent | advanced | 1.5 years ago | Developers who want to build voice cloning or text-to-speech features into their own projects |
| meetily | 32.4k | App | beginner | 4 days ago | People who want meeting notes without sending recordings to a cloud service |
| buzz | 22.4k | App | beginner | 4 days ago | People who want to transcribe recordings without uploading them to an online service |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.