GuangChen2016

Multi-GPU training could not work normally?

1

RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one. This error indicates th at your module has parameters that were not used in producing...

severe metallic sound

9

Hi, thanks for your nice jobs. I used your codes for ny own datasets and the synthesized voices seems not that normal at 160K steps now. Though we could still...

Merge multi-band linear prediction into LPCNet

15

As reported in [https://arxiv.org/pdf/2005.05551.pdf](url), the multi-band method could enable the vocoder to generate several speech samples in parallel at one step, thus significantly improve the efficiency of speech synthesis. I...

some of the synthesized results (about 3% utterances）has some artifacts (noise). In details, the mel-spectrum in corresponding ares discontinuous, shown as follows：  Any suggestions to improve the this?

some qustions?

1

Hi, I have some questions as belows: 1. Are the speaker encoder models of the base tts models and tone color converter model be the same model structure? Is there...

Hi，MIWoo， Did you tried the new version LPCNet which employ quanlization for GRU? [https://github.com/xiph/LPCNet](url).

hi @MaxMax2016 , CLONE的稳定性和效果好像比VITS好，有计划实现下这个框架么 https://arxiv.org/pdf/2207.06088.pdf

hifigan结构设计？

1

Hi，想问一下关于hifigan结构设计上，用到了原始hifigan结构中的transpose_upsamples外，还叠加了nn.Upsample的出发点是什么？这么做的好处是结果更稳定么？ 感觉把这两个结构堆叠在一起增加了计算量？

flow matching训练的pipeline？

3

hi，想问一下 flow matching训练的pipeline在哪呢？ 看了run.sh里只支持llm的训练，但是README里说是支持的，能否告知一下呢，谢谢啦