Open-source 'VoiceStudio' released with support for local voice cloning, dubbing, and transcription

The open-source studio VoiceStudio, which supports voice cloning, video dubbing, dictation, and long-form audio production on local hardware, has been released. It allows users to run private voice AI workflows without accounts, API keys, or subscriptions, and supports macOS, Windows, Linux, and Docker.

Key Features and Engines

• Diverse Engine Support: Supports 16 TTS engines, 11 ASR engines, and a catalog of 646 languages.
• Core Workflows: Offers zero-shot voice cloning, voice design, video dubbing (transcription/translation/synthesis), story/audiobook production, and a system-wide dictation widget.
• AI Watermarking: Integrates AudioSeal by default to identify the origin of synthesized audio.
• MCP Server: Includes a built-in MCP server that integrates with AI agents such as Claude Desktop and Cursor.

Technical Architecture and API

• OpenAI-Compatible API: Replaces the local backend (http://localhost:3900/v1) with an OpenAI API endpoint, ensuring compatibility with existing clients.
• Architecture: Follows a structure with a Tauri v2 (Rust) desktop shell, React + Vite UI, and FastAPI backend.
• Hardware Optimization: Automatically detects and routes to Apple Silicon (MPS/MLX), NVIDIA CUDA, Linux ROCm, or CPU.

License and Terms of Use

• License: The application follows the AGPL-3.0 license, while downloaded models retain their respective upstream licenses (e.g., OmniVoice weights are CC-BY-NC).
• Privacy: Data is stored only on the local disk by default, and remote features are opt-in only.