Amazon announces BASE TTS, a speech model with one billion parameters that supports voice-cloning and outperforms baseline TTS models. The model is trained on unlabeled speech audio scraped from the web, and its quality improves with increased data and model size.
4 Impressions