OmniVoice v0.2.1 – Clone Your Voice in 600+ Languages! Complete Installation Guide (2026)

Summary

The video introduces OmniVoice v0.2.1, an open-source text-to-speech and voice-generation tool. The presenter demonstrates sample speech and describes the following capabilities:

• Voice cloning: Generate speech from an uploaded reference recording, with support for more than 600 languages and a choice of accent.
• Voice design: Create a new voice by selecting a language, gender, and other characteristics.
• Speech controls: Enter text and adjust options such as accent, speed, and inference steps. Adding a transcript of the reference audio may improve cloning results.

The presenter describes the generated voices as natural-sounding and plays sample outputs during the introduction and demonstration.

Installation Guide

The tutorial explains how to install and run OmniVoice locally on Windows:

1. Check that the system has a graphics card with at least 3 GB of VRAM, CUDA Toolkit 11.8–12.8, Python 3.10.11, FFmpeg, Visual Studio, and Git for Windows.
2. Open the OmniVoice GitHub page, copy its Git URL, and clone the repository from a command window opened in the desired folder.
3. Create and activate a Python virtual environment.
4. Install PyTorch using the command for the installed CUDA version. The presenter notes that the CUDA version in the command may need to be changed and that torch and torchaudio should be included.
5. Install the project requirements, then run the OmniVoice launch command. Once the models have downloaded, open the displayed local URL in a browser.
6. Optionally, create a batch file in the project folder to launch OmniVoice by double-clicking it in the future.

The tutorial then demonstrates generating cloned speech from text and reference audio, followed by creating a new voice with the Voice Design option.

Speaker and Audio

The presenter and narrator is from the AiThatSpeaks channel. The audio examples are generated voice samples.