STYLETTS 2

Summary from the TLDR AI newsletter

StyleTTS 2 is a text-to-speech model designed to achieve human-level speech synthesis by combining style diffusion and adversarial training with large speech language models.