Voicebox is a free, open-source AI voice studio for macOS, Windows and Linux that clones voices, generates speech and dictates into any app — entirely on your own machine. It has earned more than 58,000 stars on GitHub and passed a million downloads, with its latest updates in 2026. Think of it as a local, no-subscription alternative to cloud voice services like ElevenLabs and Wispr Flow.
Voicebox bundles seven text-to-speech engines — including Qwen3-TTS, Chatterbox and Kokoro — plus Whisper transcription, all running locally on your GPU or CPU. Clone a voice from as little as three seconds of audio uploaded, recorded or captured from system sound, then generate expressive speech, give the voice a written personality that can rewrite or compose lines in character, and dictate system-wide with a hotkey. An MCP endpoint lets AI agents like Claude Code and Cursor speak back in per-agent cloned voices, and a built-in REST API turns every engine into a local endpoint for your own apps.
Voicebox is completely free and open source under the MIT license. There are no subscriptions, credits or usage fees — you download the app and the models, and everything runs on your own hardware.
Voicebox is for creators, developers and accessibility users who want professional voice cloning and dictation without cloud bills or privacy trade-offs — and for agent builders who want their assistants to talk back out loud.