Amazon researchers detail BASE TTS, the largest text-to-speech model yet, which they claim exhibits "emergent" qualities improving its natural speaking ability (Devin Coldewey/Ztoog)