English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recently, GAN based speech synthesis methods, such as MelGAN, have becomevery popular. Compared to conventional autoregressive based methods, parallelstructures based generators make waveform generation process fast and stable.However, the quality of generated speech by autoregressive based neuralvocoders, such as WaveRNN, is still higher than GAN. To address this issue, wepropose a novel vocoder model: TFGAN, which is adversarially learned both intime and frequency domain. On one hand, we propose to discriminate ground-truthwaveform from synthetic one in frequency domain for offering more consistencyguarantees instead of only in time domain. On the other hand, in contrast tothe conventionally frequency-domain STFT loss approach or feature map loss bydiscriminator to learn waveform, we propose a set of time-domain loss thatencourage the generator to capture the waveform directly. TFGAN has nearly samesynthesis speed as MelGAN, but the fidelity is significantly improved by ournovel learning method. In our experiments, TFGAN shows the ability to achievecomparable mean opinion score (MOS) than autoregressive vocoder under speechsynthesis context.

TFGAN: Time and Frequency Domain Based Generative Adversarial Network for High-fidelity Speech Synthesis