VoiceStudio
Fully local, open-source ElevenLabs alternative for voice cloning, voice design, video dubbing, and transcription in 646 languages
Overview
Description
VoiceStudio (formerly OmniVoice Studio) is an open-source AI voice studio that runs entirely on your own machine. It clones voices from a short sample, designs new voices from a text description, dubs videos into other languages, and handles dictation, transcription, and audiobook creation across up to 646 languages. No account, no cloud upload, and no per-character fees: your audio and models stay local.
Features
- Voice Cloning: Create a reusable voice from a few seconds of clean speech and use it for any script.
- Voice Design: Describe a voice in plain language and generate a brand-new synthetic speaker.
- Video Dubbing: Transcribe, translate, and re-voice video files while keeping timing aligned.
- Transcription & Dictation: Local speech-to-text powered by whisper.cpp and other ASR engines.
- Audiobooks & Stories: Build chaptered audiobooks and multi-voice scenes from long-form text.
- Model Manager: Pick from dozens of TTS/ASR engines (OmniVoice, CosyVoice, GGUF builds) and download only what you need.
Technology Stack
A Python backend built on FastAPI, PyTorch, and torchaudio runs the speech engines, with native GGUF engines accelerated via CUDA, ROCm, Metal (MPS), or Vulkan. The desktop app is built with Electron and TypeScript, alongside mobile clients and an MCP server for agent integrations.
Requirements
Installers are available for macOS (Apple Silicon), Windows 10/11, and Linux (AppImage, deb, Snap, Nix), plus Docker images on GHCR. An NVIDIA, AMD, or Apple Silicon GPU is recommended for speed, but the app is fully usable on CPU.
