UNSUPERVISED TTS SPEAKER ADAPTATION USING X-VECTOR CONTROLLED PARAMETERS, Zhihang Xu, Bo Chen, Kai Yu (submitted to Interspeech2020) demo page
each line show 3 test audios from splited testset
Text |
登录各大影院 |
在奥迪与蓝天工作室的密切沟通下 |
带着炫酷的环球冒险 |
original-vocode |
|
|
|
unsupervised-xvec-adaptation |
|
|
|
supervised-adaptation |
|
|
|
xvec-controlled-xvec-adaptation(proposed) |
|
|
|
Lets' see more examples which are out of text domain
original-vocode |
|
|
|
Text |
均不与国信证券存在关联关系,或者代持 |
场外市场不是以融资为目的市场 |
比亚马逊当前的市值要高出百分之三十六 |
unsupervised-xvec-adaptation |
|
|
|
supervised-adaptation |
|
|
|
xvec-controlled-xvec-adaptation(proposed) |
|
|
|