Grad-TTS: A Diffusion Probabilistic Model for Text-to-Speech

       Vadim Popov * 1 Ivan Vovk * 1 2 Vladimir Gogoryan 1 2 Tasnima Sadekova 1 Mikhail Kudinov 1

               Abstract              2014) and Normalizing Flows (Rezende & Mohamed, 2015)
     Recently, denoising diffusion probabilistic mod-   were used in the design of TTS engines for a parallel gener-
     els and generative score matching have shown     ation with comparable quality of the synthesized speech.
     high potential  ...