We are moving away from maintaining this repository.

This repository includes

• Production model code for Tacotron2, so-ViTS-svc, zero-shot RADTTS, HiFi-GAN, and RVC.
• Production training code for Tacotron2, RADTTS, HiFi-GAN, and RVC.
• Fill-populating inference, additional languages, and Torchmoji emotional encoding.

Notebooks are available here, and models to fine-tune from are available here here.

If you want to install on your own machine, create a virtual environment and install like

conda create -n 'test-env' python=3.10 -y
source activate test-env
pip install git+https://github.com/uberduck-ai/uberduck-ml-dev

Train a radtts on LJ Speech as follows

cd tutorials/radtts
bash download.sh
bash train.sh

You will need to adjust the paths and potentially other training settings in tutorials/radtts/demo_config.json. This code has been tested on a single T4 as well as 2 A6000s.

For processing of new datasets, see tutorials/radtts/radtts_data_processing.ipynb.

We love contributions!

To install in development mode, run

pip install pre-commit black # format your code on commit by installing black!
git clone git@github.com:uberduck-ai/uberduck-ml-dev.git
cd uberduck-ml-dev
pre-commit install # Install required Git hooks
python setup.py develop # Install the library

In an environment or image with uberduck-ml-dev installed, run

cd uberduck-ml-dev
python -m pytest