A streaming silent-speech neuroprosthesis for naturalistic voice restoration

This repository includes training data to replicate "A streaming silent-speech neuroprosthesis for naturalistic voice restoration." The code to reproduce the figures, train and deploy the models, and the online pre-trained model and synthesizer are in code.tar.gz (2GB). The code is placed here for the editorial team and the referees, but if the manuscript is accepted, it will be made publicly available on GitHub upon publication. After downloading and unzipping the code, please follow the instructions in README.md to reproduce the paper's main results. Given the sensitive nature and speech-BCI privacy considerations of a personalized brain-to-voice model, we kindly ask the reviewers to take additional precautions to keep the personalized synthesizer private and remove it from their devices upon the conclusion of the revision process. All neural data are in d_000000~000033.tar.gz (75GB). If the manuscript is accepted, we will provide a non-personalized brain-to-voice model to share with researchers from other institutions.