Youtube Autonomous Audio Narration Coqui Voice Module

The Audio narration Coqui voice module.

Please, check the 'pyproject.toml' file to see the dependencies.

General information

This project is based on coqui-tts (needs a license for commercial use), which is working with different models that are handled in a different way. Some models are older and others are more recent. There is, apparently, a sweet spot of dependencies versions to make all of them work together. I leave it here, but maybe we should prioritize the most modern models.

This project is, by now, using the xtts_v2 model only.

Stack estable universal

• python==3.11.0.final.0
• torch==2.5.1
• torchaudio==2.5.1
• TTS==0.25.x
• transformers==4.38.2

Which is allowing VITS, XTTS v2, CSS10, Tacotron2 and other multilingual models.

About the models

The models are downloaded into the cache if needed. In windows, this means here (C:/Users/dania/AppData/Local/tts).

You must set the TTS_HOME environment variable to choose where your models are (mine is D:/coqui-tts). If you don't set it, the fallback will use the local data folder, but I made it to raise an Exception if this happens to be able to control it.

Downloading the model needs to day yes to an input to accept the non-commercial agreement.

Instructions

To make this coqui-tts work we will need the extra = ['languages'] and the espeak-ng installed to be able to narrate.

Espeak-NG

1. Go to the releases page and download the latest version (or one that fits your needs), the .msi if Windows:

• https://github.com/espeak-ng/espeak-ng/releases

2. Register the folder in which it has been installed in your system PATH.

You will probably need this also:

• https://github.com/erew123/alltalk_tts/wiki/Install-%E2%80%90-WINDOWS-%E2%80%90-Python-C-&-SDK-Requirements