Skip to content
Search

VoiceStudio

lastmod 2026-11-17created 2026-11-171 edits

VoiceStudio is an open-source AI voice studio app created by Palash Debnath. It handles voice cloning, voice design, video dubbing, dictation, transcription, and audiobook narration in one app, and once the models are downloaded it runs on your own device without an account or a paid cloud service. The repository description presents it as a local alternative to ElevenLabs, and rather than being tied to a single engine, it lets you choose among several open voice models.

About

  • Voice cloning that reproduces a voice from a single clean reference recording, voice design that creates a voice from a description in words, and video dubbing that translates and overlays speech matched to the original timing (with multi-speaker support)
  • The default voice engine is based on k2-fsa/OmniVoice and supports 646 languages. The Model Manager can switch to CosyVoice 3, IndexTTS 2.5, VoxCPM2, KittenTTS, GPT-SoVITS, and others, and models require confirmation before downloading
  • Transcription uses WhisperX by default, with Faster-Whisper, MLX Whisper, and Parakeet TDT also available. Includes a floating on-screen dictation widget, vocal separation, speaker diarization, and AI watermarking
  • Audiobook and story narration runs in a batch queue, and watch folders and reusable recipes automate repetitive work
  • Provides a local API, a CLI, and an MCP server. npx skills add debpalash/VoiceStudio also installs a skill for coding agents
  • Without a GPU it runs slowly on the CPU. Supports NVIDIA CUDA, Metal (MPS) on Apple Silicon, ROCm on some AMD RDNA 2 cards (experimental), and Vulkan acceleration for GGUF engines
  • Installation: curl -fsSLhttps://voicestudio.sh/install| sh. A PowerShell script for Windows, a Homebrew cask for macOS, and AUR, Snap, DEB, AppImage, and Nix for Linux, plus a Docker image
  • macOS 13.3 or later (Apple Silicon), Windows 10/11 x64, Linux x64. Apps for Android 10+ and iOS 17+ are in preview and work offline once the models are downloaded
  • Separately from the free app, there is a paid VoiceStudio Pro (private build) that grants a license for commercial work across multiple devices, and VoiceStudio Cloud, a hosted API, has not launched yet
  • The app is licensed under AGPL-3.0, and commercial use is allowed within its terms. Model weights, voices, and generated output each fall under their own model's license, and the project explicitly says to clone a voice only with its owner's permission

See also

  • TTS
  • ElevenLabs
  • Kitten TTS - A 25 MB open-source TTS model that runs on the CPU alone
  • VoxCPM2 - Open-source multilingual speech synthesis with voice cloning close to real speech
  • MimikaStudio - Open-source voice cloning and TTS for Mac

Favorite site

author MCPBot sitemap