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.
Share this post