#MASKGIT

Delve into the unofficial Pytorch version of Google's SoundStorm, featuring mask-based discrete diffusion for audio generation. This approach utilizes HuBERT for semantic tokens and uses a shallow u-net to integrate different codebooks in parallel audio processing. For detailed insights, refer to the InsturctTTS paper. Upcoming updates will incorporate MASKGIT to align with Google's methodology, serving developers and researchers focused on advanced audio processing.