MaxMax2016 comments

Results 243 comments of


                                            MaxMax2016

training dataset

https://wenet.org.cn/opencpop/download/

full song with dnn pitch

@Lukelluke 简单的讲，就是使用帧级别的phone,midi通过transformer预测帧级别的pitch，相关pitch预测可以参考 https://github.com/MoonInTheRiver/DiffSinger/blob/master/modules/fastspeech/pe.py https://github.com/yl4579/PitchExtractor

Benchmark for Multiband-Hifigan

https://github.com/xcmyz/FastVocoder/tree/main/model/generator :multiband_hifigan.py @BridgetteSong @v-nhandt21 may help you

pretrain model for test

look forward to

Unable to run deepFIlter form CLI

i meet the same error, which python should use

Chinese data DB1 support time?

另外PortaSpeech是不是比FastSpeech2需要更多的训练语料呢？因为要处理字级别的信息。

Chinese data DB1 support time?

ok，i will try.

accuracy compares with VITS?

i think PortaSpeech + MB MelGAN can work on cpu very well, VITS need gpu to work.

Recipe for opencpop database

waiting for you