Why E2-TTS superior to nvidia PFlow-TTS?

lucidrains / e2-tts-pytorch

Implementation of E2-TTS, "Embarrassingly Easy Fully Non-Autoregressive Zero-Shot TTS", in Pytorch

MIT License

228 stars 21 forks source link

Why E2-TTS superior to nvidia PFlow-TTS? #3

Closed JohnHerry closed 1 month ago

JohnHerry commented 2 months ago

Hi, I am sorry for my misunderstanding if it is. But the they are all text-encoder, duration predictor and condition-flow-matching structure for PFlow-TTS and E2-TTS. why E2-TTS will be better? In my experiments before, all models with DP [duration predictor based on TextEncoder outputs] are not good at prosody and naturality.

lucidrains commented 2 months ago

don't think this paper ever made that claim