What you’ll have at the end
A desktop app that clones and designs voices, dubs videos, narrates audiobooks and transcribes audio, all on your own computer. There’s no monthly plan and no upload to someone else’s server.
Written from VoiceStudio’s GitHub README and its Windows and Mac install docs, read on 17 Nov 2026.
Clone your own voice, or a voice whose owner gave you written permission. VoiceStudio’s own terms prohibit unauthorised cloning, impersonation and fraud. The app’s licence (AGPL-3.0) allows commercial use, but each voice model has its own licence, and some don’t allow commercial use. Check it in the Model Manager before using a voice in an ad or client project.
What you need
- Windows 10 or 11 (64-bit), a Mac, or Linux. There’s also an Android app.
- About 10 GB of free disk space for the app, its Python setup and voice models. Without an NVIDIA graphics card, it needs about 5 GB.
- A graphics card is optional. With an NVIDIA GPU it’s fast. Without one, it runs on the processor, just slower.
Setup in 5 steps
Download the installer
Open the repo’s Releases page and download the file for your computer:
- Most Windows PCs (Intel / AMD):
VoiceStudio-<version>-win-x64.exe - Windows on ARM (Snapdragon): the
arm64installer (experimental) - Mac: the DMG, or use Homebrew (see the repo’s macOS guide)
Or, on Windows, let the official install script pick the right file and check it for you. Run this in a normal (not admin) PowerShell window:
irm https://voicestudio.sh/install.ps1 | iex- Most Windows PCs (Intel / AMD):
Run the setup wizard
It installs for your user account only, so you don’t need admin rights. You can choose the folder.
Low on C: drive space?Voice engines can take several GB each. To keep them on another drive, run this once in PowerShell before you open the app, then restart it:
[Environment]::SetEnvironmentVariable('OMNIVOICE_ENGINES_DIR', 'D:\VoiceStudio\engines', 'User')Let the first launch finish
The first time you open it, VoiceStudio downloads its own Python, FFmpeg and model files. You don’t need to install anything else. It detects whether you have an NVIDIA GPU and picks the right version.
Pick a voice engine
The default engine is VoiceStudio, powered by OmniVoice. Others are in the Model Manager. Read the licence shown next to each model.
On a laptop without an NVIDIA GPU, choose a light engine (KittenTTS, Supertonic-3 or PocketTTS) and a small Whisper model. The setup screen lists these as its CPU preset.
Make your first voice
Record a clean sample of your own voice in a quiet room, create a voice from it, then type a script and generate audio. For dubbing, load a video and choose the target language.
Good first projects
- A voice-over for a reel in your own voice, without recording every take.
- An English video dubbed into Malayalam, or the other way round. Check the result with a native speaker before you post it.
- Turning a blog post into audio narration.
If something doesn’t work
| Problem | Fix |
|---|---|
| Windows blocks part of the setup | Smart App Control can block third-party Python files that the app downloads. See the troubleshooting doc in the repo. |
| Very slow | Switch to a light CPU engine (step 4). Close other heavy apps. |
| Disk full | Move engines off C: (step 2), or remove engines you don’t use. |
| Installer seems stuck | Wait a few minutes, then check Task Manager. The repo’s Windows guide lists what to send the developers. |
Removing it
The uninstaller asks whether to also delete the downloaded runtime and models. Your voices and projects are kept either way.
Back to the full list: 4 open-source alternatives to paid AI tools