What’s new here
VoiceStudio fits the Model hosting and compute theme as a local app rather than a cloud compute layer. It runs voice cloning, video dubbing, dictation, transcription, and audiobook generation on your own machine, with remote services optional. It exposes a local API and an MCP server so agents can call it too. The default engine is k2-fsa/OmniVoice; other engines are selectable from a built-in catalog.
What it does
VoiceStudio is a desktop application for voice AI work that runs locally on Apple Silicon Macs, Windows, and Linux. On Intel Macs only the app UI runs, and it connects to a remote backend.
- Voice cloning: provide a reference recording, enter text, generate speech.
- Voice design: describe a voice in text and generate it from scratch.
- Video dubbing: replace audio tracks in video files with timed synthesized speech.
- Dictation: floating widget for speech-to-text input.
- Transcription and audiobook creation: batch jobs for converting audio or text files.
The app supports CUDA on NVIDIA GPUs, Metal (MPS) on Apple Silicon, and CPU-only mode for machines without a dedicated GPU. It ships with 42 releases as of v0.5.6 and covers 646 languages. A local API and MCP integration let coding agents drive audio workflows directly.
Who it’s for
Developers, content creators, and researchers who want ElevenLabs-style voice capabilities without sending audio to a third-party API. Also useful for anyone building agent workflows that need speech synthesis or transcription as a local service, since VoiceStudio exposes both a local API and an MCP endpoint.
Try it
macOS / Linux (one-command install):
curl -fsSL https://voicestudio.sh/install | shWindows (PowerShell):
irm https://voicestudio.sh/install | iexFrom source:
git clone https://github.com/debpalash/VoiceStudio.git
cd VoiceStudio
bun install
bun run setup:api
bun run devOr download a platform package (.dmg, .exe, .AppImage, .deb) from the Releases page.
How mature is it
57,035 stars, 6,399 forks, 82 contributors, 42 releases. Latest release is v0.5.6 (2026-09-23). At least 100 commits in the last 90 days (capped). Created April 2026; last push October 10, 2026. 59 open issues and pull requests. Licensed AGPL-3.0.