BriefGPT.xyz
Dec, 2019
WaveFlow: 面向原始音频的紧凑流模型
WaveFlow: A Compact Flow-based Model for Raw Audio
HTML
PDF
Wei Ping, Kainan Peng, Kexin Zhao, Zhao Song
TL;DR
本文介绍了一种小型的WaveFlow产生式流,可以训练原始音频并合成高保真语音,只需几个步骤即可生成成千上万个时间步长的波形,并具有比WaveGlow小15倍的参数和42.6倍的音频合成速度。
Abstract
In this work, we present
waveflow
, a
small-footprint
generative flow
for
→