VoiceStudio

VoiceStudio is a desktop application and local API for voice cloning, voice design, video dubbing, dictation, and long-form audio production. It runs on your own machine using one of 16 TTS engines and 11 ASR engines, with no account, API key, or usage meter required for the local workflow. It's built for developers, podcasters, audiobook producers, and anyone who wants ElevenLabs-style voice tools without sending audio to a third party.

The app ships for macOS (Apple Silicon), Windows, and Linux, plus Docker images for server or headless setups. Compute routes automatically across CUDA, Apple Silicon MPS/MLX, ROCm, or CPU, with remote worker support if you want to offload generation to another machine. Everything you produce (voices, projects, settings, outputs) stays local by default; network features are explicit opt-ins.

Key features

• Voice cloning: zero-shot synthesis from a short reference clip, three seconds minimum, 5-15 seconds recommended for better results.
• Voice design: build a voice from age, accent, pitch, and delivery instructions instead of a sample.
• Video dubbing: transcribe, translate, preserve speaker identity, synthesize, and export dubbed video with a waveform/transcript editor and glossary support.
• Audiobook production: multi-voice scripts, EPUB/PDF import, chapter rendering, and .m4b export.
• Dictation widget: system-wide shortcut for live transcription with optional local-LLM cleanup.
• Vocal isolation and diarization: Demucs for speech/background separation, Pyannote and WhisperX for speaker assignment.
• Batch queue: process large sets of audio/video jobs or watch a folder for new files automatically.
• Model catalogue: install, remove, and route TTS, ASR, and LLM models per engine, including remote worker installs.
• GPU auto-detect: routes work across CUDA, MPS, ROCm, or CPU with per-engine checks.
• AI watermarking: AudioSeal embedding and detection for generated audio.
• MCP server: exposes synthesis and transcription as tools for MCP clients.
• Interfaces: desktop app, local REST/SSE/WebSocket API, an OpenAI-compatible audio API, and the MCP server.

Ideal use cases

VoiceStudio fits teams and individuals who need voice cloning or dubbing but can't or don't want to send audio and text to a hosted API, for privacy, cost, or offline reasons. It's a good match for audiobook producers doing chapter-length narration, video creators dubbing content into multiple languages while keeping speaker timing, and developers who want a local OpenAI-compatible TTS/STT API to build against without per-request billing. The dictation widget and batch queue also make it useful for anyone doing high-volume transcription or voice generation on their own hardware.

It's less of a fit if you need managed scaling, zero local setup, or guaranteed uptime without maintaining hardware. Intel Macs can't run the local Python backend (a remote backend is required instead), and heavier engines need real VRAM or unified memory to perform well. If you just need occasional short clips and don't want to manage models, disk space, or updates, a hosted service is simpler.

Installation

Download a packaged build from the latest GitHub release, or run it via Docker:

docker run -d -p 127.0.0.1:3900:3900 -v omnivoice-data:/app/omnivoice_data --name voicestudio palashdeb/omnivoice-studio:stable

Note: published Docker images are linux/amd64 only. On Apple Silicon, use the native macOS app for GPU acceleration.

Platform packages:

• macOS 13.3+ (Apple Silicon): DMG installer, needs a one-time right-click + Open approval on first launch.
• Windows 10/11 x64: MSI installer (a current-user build is available for non-admin installs).
• Linux x86_64: AppImage, requires glibc 2.39+.

First launch creates a managed Python environment and downloads the default model; later launches reuse both.

To run from source, install Node 20+/Bun and Python 3.11+, then:

git clone https://github.com/debpalash/VoiceStudio.git
cd VoiceStudio
bun install
bun run desktop

The desktop launcher configures Python dependencies via uv automatically. Use bun run dev for the browser UI. A Google Colab notebook is also available for trying it without a local install.

For your first voice: launch the app, open Voice Cloning, add a clean reference sample (3-15 seconds), enter text, choose a language, and generate.