🔗 • Getting started • Installation • Tips, Tricks & FAQ

Welcome to MetaVoice Live, our real-time AI voice changer! Live converts your voice while preserving your intonations, emotions, and accent.

We are open-sourcing our code to give the community the freedom to make their own improvements. This repository contains source code for:

• Our ML model inference on Windows & Nvidia GPUs, and,
• User-facing desktop app

Please use Windows as the development environment. We recommend using Cmder as the terminal of choice.

1. Run conda create -n mvc python=3.10 -c conda-forge and conda activate mvc
2. Copy Makefile.variable.sample, rename to Makefile.variable & update the vars with their appropriate value. Make sure the site-packages directory exists, or adjust it.
3. Run make setup
4. Run make install-cuda
5. Add the windows env variable set METAVOICELIVE_ROOT=%cd%
6. Setup Git LFS

   • Install Git LFS for Windows here
   • Initialise Git LFS within the repository by running: git lfs install
   • Pull the model weights via: git lfs pull

(Optional) You might also want to copy .env.sample into .env and fill those values if you can.

• ai/ -> ML model weights & inferencing pipeline
• services/desktop_app -> electron application, see its README.md

We welcome PRs & issues. If you have questions, please mention @sidroopdaska or @towc

Some ideas for first PRs:

• Port inference_rt.py -> C++
• Streamline the Electron app build & release process
• Add support to package Live for Mac

🙏 If you come across something sensitive, e.g. vulnerabilities or access keys, please let us know privately on 📧 [email protected] first & give us 2 weeks to resolve it.

• Twitter

MetaVoice Live is licensed under the GPL-3.0 license.

Please contact us at 📧 [email protected] to request access to a larger version of the model.

MetaVoice does not take responsibility for any output generated. Please use responsibly.