Audio samples of “FLOW-TTS: A NON-AUTOREGRESSIVE NETWORK FOR TEXT TO SPEECH BASED ON FLOW”

Abstract

In this work, we propose Flow-TTS, a non-autoregressive end-to-end neural TTS model based on generative flow. Unlike other non-autoregressive models, Flow-TTS can achieve high-quality speech generation by using a single feed-forward network. To our knowledge, Flow-TTS is the first TTS model utilizing flow in spectrogram generation network and the first non-autoregssive model which jointly learns the alignment and spectrogram generation through a single network. Experiments on LJSpeech show that the speech quality of Flow-TTS heavily approaches that of human and is even better than that of autoregressive model Tacotron 2 (outperforms Tacotron 2 with a gap of 0.09 in MOS). Meanwhile, the inference speed of Flow-TTS is about 23 times speed-up over Tacotron 2, which is comparable to FastSpeech.

We compare our method with Tacotron 2 (using pretrain model from https://github.com/NVIDIA/tacotron2) and FastSpeech (using pretrain model from https://github.com/espnet/espnet). All of the audio samples use WaveGlow as vocoder.

There seems to be no reason why ordinary paper should not be better made.

The various wards were all about eleven feet in height.

The moral welfare of the inmates was as closely looked after as the physical.

an incomplete and fallacious method of preventing contamination.

except the forgery of wills and powers of attorney.

by the blackmith in the usual way.