debpalash/OmniVoice-Studio

What it solves

OmniVoice Studio provides a completely local, open-source alternative to cloud-based voice AI services like ElevenLabs. It eliminates the need for monthly subscriptions, API keys, and cloud accounts while ensuring that sensitive voice data never leaves the user's machine.

How it works

The project uses a Tauri v2 desktop shell (Rust) with a React UI and a bundled Python/FastAPI backend. It integrates a wide array of AI models for speech synthesis and recognition, supporting 14 different TTS (Text-to-Speech) engines and 11 ASR (Automatic Speech Recognition) engines. It can automatically detect hardware (CUDA, Apple Silicon MPS, ROCm, or CPU) to optimize performance and offload tasks to the CPU if VRAM is limited.

Who it’s for

It is designed for users who want professional-grade voice cloning, dictation, and dubbing tools without relying on cloud services, as well as developers who want an extensible framework to add their own TTS engines.

Highlights

• Zero-Shot Voice Cloning: Mirror any voice using a 3-second audio clip across 646 languages.
• Video Dubbing: An end-to-end pipeline that transcribes, translates, re-voices, and exports video as MP4.
• Voice Design: Create new voices from scratch by specifying gender, age, accent, pitch, and emotion.
• Local-First Tooling: Includes a dictation widget for any app, an audiobook editor (EPUB/PDF to .m4b), and a multi-voice script editor for stories.
• Advanced Audio Processing: Features Demucs-powered vocal isolation and Pyannote/WhisperX speaker diarization.
• MCP Server: Integration for use with AI clients like Claude and Cursor.