We present different approaches to improve the streaming capabilities for the kNN-VC implementations, while also testing the limitations of each of the components of kNN-VC and comparing our solutions with a current method of naive chunking.
To test the effectiveness of the systems, we will compare the delay and the resulting Real-Time Factor (RTF) of each system. Additionally, we will deploy Whisper for calculating Word Error Rate (WER) and Character Error Rate (CER) values, and NISQA to provide us with Mean Opinion Score (MOS) values to compare the systems.
Steps to install your project:
git clone https://github.com/username/repo.git
cd repo
npm install
## Features
We provide 5 different approches and also the the naive chunking for comparison