zhangyongmao/VISinger2

Issues

• Could the f0 prediction be further improved？

  #23 opened 3 months ago by lhc991025
  1

• How to extract pitch and melody?

  #22 opened 4 months ago by xingaocs
  0

• About the data preparation

  #20 opened 9 months ago by SeiniTetsuko
  1

• About the license ?

  #21 opened 8 months ago by juntaosun
  0

• question on make proper dateset

  #19 opened a year ago by li-henan
  0

• Does SVS work in english lyrics?

  #18 opened a year ago by AbdullahMustafa040
  0

• decoder condition和论文中不一致

  #16 opened a year ago by iehppp2010
  2

• STFT question in multi-frequency discriminator

  #15 opened a year ago by jerryuhoo
  1

• The impact of flow on performance

  #12 opened a year ago by jerryuhoo
  1

• how to train a small model

  #14 opened a year ago by imiskolee
  1

• train issue on 3090 device

  #13 opened a year ago by imiskolee
  4

• pretrained model load error

  #10 opened a year ago by Liujingxiu23
  1

• meaningless audio envelope in sp

  #11 opened a year ago by Liujingxiu23
  0

• necessity of upsampling f0 during data processing

  #9 opened 2 years ago by cnlinxi
  2

• 为什么论文里vocoder用了LSTM但是代码里注释掉了没用？

  #8 opened 2 years ago by splinter21
  1

• 对dsp_wav做频谱loss有什么用？直接对final result做GAN、feature和L1 loss不就行了吗？

  #7 opened 2 years ago by splinter21
  1

• 使用了FFT block代替原版VITS里面的transformer里面的注意力编码器和flow，这2点有无ab实验验证确实有提升？

  #2 opened 2 years ago by splinter21
  1

• 关于gtdur的累计误差问题

  #5 opened 2 years ago by innnky
  2

• Assume Single Node Multi GPUs Training Only

  #6 opened 2 years ago by ConClas
  1

• VISinger2的標記格式

  #4 opened 2 years ago by Archivoice
  1

• 请问是否有计划放出预训练模型呢

  #3 opened 2 years ago by innnky
  0

• Will the code be open-sourced?

  #1 opened 2 years ago by 980202006
  32