Unofficial implementation of paper "NANSY++: Unified Voice Synthesis with Neural Analysis and Synthesis" in Pytorch Lightning following guidelines from lightning-hydra-template and drawing inspiration from revsic unofficial implementation.
We focus on 2 subtasks:
- voice conversion: documentation, R&D observations
- text-to-speech: documentation, R&D obeservations
We also provide pretrained checkpoints for both models (please refer to the documentation), and compare our results with those reported in the article on this Notion page.
git clone --recurse-submodules https://github.com/MWM-io/nansypp.git
cd nansypp
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt