GitHub - debpalash/VoiceStudio: VoiceStudio is the open-source, fully-local ElevenLabs alternative — voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.

VoiceStudio is an open-source desktop application that handles voice cloning, video dubbing, text-to-speech dictation, and transcription across 646 languages. You can clone existing voices with a reference recording, design custom voices from scratch, or dub videos with synchronized speech. The app runs locally on your machine—no cloud dependency required—though remote services are optional if you want them. It supports multiple speech engines, with OmniVoice as the default, and lets you build reusable recipes and watch folders for batch jobs. The project also exposes a local API and MCP (Model Context Protocol) for integration with coding agents like Claude, Cursor, and Codex. Installation is straightforward via curl on macOS and Linux, or PowerShell on Windows; you can also grab pre-built binaries from the releases page. Hardware support covers NVIDIA GPUs with CUDA, Apple Silicon with Metal acceleration, and CPU-only setups (slower but functional). The app requires about 5 GB of free disk space for the CPU build of PyTorch. It's built with Electron and switched from Tauri as of version 0.5.3, so existing Tauri users need to reinstall. Development setup needs Git, Node.js 22+, Bun, and Rust/Cargo if you're building from source. The license is AGPL-3.0, meaning commercial use is allowed under those conditions, but voice cloning requires explicit permission from whoever owns the original voice. Model weights, tokenizers, and generated output have separate licensing terms. The Model Manager now tracks versioned model sources and evidence links, though it doesn't enforce or guarantee commercial rights—that's on you to verify. You can support development through Ko-fi, PayPal, or sponsorship, and the project welcomes contributions and paid partnership placements.