Your voice. Your workflow.

VoiceStudio is an open-source application for voice cloning, voice design, video dubbing, dictation, transcription, and audiobook creation. It supports 646 languages and is a fully-local alternative to services like ElevenLabs. You can run it entirely on your own hardware for complete privacy and control.

VoiceStudio is available on macOS, Linux, and Windows through the Electron desktop application. For mobile, there are preview iOS and Android companion apps. You can also run it in a Docker container or integrate it into your applications via the local API.

NVIDIA GPU (Windows/Linux): Supports CUDA acceleration for best performance.
Apple Silicon: Supports Metal (MPS) acceleration.
No dedicated GPU: Fully usable on CPU, though slower. The setup installs the small CPU build of PyTorch (requires about 5 GB of free disk space).
Windows on ARM: Experimental support with a native ARM64 app and x64 Python backend under emulation, CPU only only.
Intel Macs: The app UI runs locally, but you'll need to connect to a remote backend for processing.

The easiest way is with our one-command installer. On macOS or Linux, run: curl -fsSL https://voicestudio.sh/install | sh. On Windows, run irm https://voicestudio.sh/install | iex in PowerShell. You can also download pre-built packages from the Releases page for your platform. See our installation documentation for detailed guides.

VoiceStudio comes with OmniVoice as the default engine, but supports a growing catalog of other speech engines. You can choose your preferred engine from the application settings. The available engines offer different trade-offs in terms of quality, speed, and language support. Check our engine catalog for details.

Yes, VoiceStudio is fully open-source under the GNU Affero General Public License v3.0 (AGPL-3.0). This means you can use, modify, and distribute the software freely, even for commercial purposes, as long as you comply with the terms of the license. The source code is available on our GitHub repository.

Yes, the VoiceStudio application itself can be used for commercial purposes under the AGPL-3.0 license. However, please be aware that the license terms for the model weights, tokenizers, voices, and generated output are separate from the application license. A paid application license does not automatically grant you rights to the underlying models. Always review the specific licenses for the models you use.

Voice cloning works by training a model on a short audio sample of a person's voice (typically 1-10 seconds). The model learns the unique characteristics of that voice, including pitch, tone, and timbre. Once trained, you can input any text and the model will generate speech in the cloned voice. For best results, use a clean recording with minimal background noise.

The Model Context Protocol (MCP) is a standard for connecting AI models to external tools and data sources. VoiceStudio supports MCP, allowing it to be integrated into workflows with AI agents and other automation tools. This enables you to build sophisticated voice-based applications and systems that can leverage VoiceStudio's capabilities as a component.

We welcome contributions from the community! You can contribute by reporting bugs, suggesting features, translating the interface, or contributing code. Please read our contribution guidelines and make sure to sign our Contributor License Agreement (CLA) for your first pull request. Join our Discord server to connect with other contributors.

If you encounter issues, please first check our troubleshooting guide. If you can't find a solution, you can open an issue on GitHub. For more interactive help, join our Discord server where you can chat with the developers and other users.