AI-Song-Cover-RVC icon indicating copy to clipboard operation
AI-Song-Cover-RVC copied to clipboard

Published 20 hours ago • verified publisher icon ardha27

the de-reverb does not work.

Open mohmmad-okasha opened this issue 2 years ago • 13 comments

in All in One Inference by SociallyIneptWeeb the de-reverb does not work, the voice still with reverb so the output not good.

mohmmad-okasha avatar Sep 25 '23 08:09 mohmmad-okasha

Then just turn it off. image Also, I think -60 is good enough, too much might error'ed out the code.

MinatoIsuki avatar Sep 25 '23 10:09 MinatoIsuki

Then just turn it off. image Also, I think -60 is good enough, too much might error'ed out the code.

thanks, but not working with me, I tried -60, -999

mohmmad-okasha avatar Sep 25 '23 12:09 mohmmad-okasha

Can you give me your song? Also, can you try messing with the Reverb Control settings to see if it works? image

thanks, but not working with me, I tried -60, -999

MinatoIsuki avatar Sep 25 '23 12:09 MinatoIsuki

Can you give me your song? Also, can you try messing with the Reverb Control settings to see if it works? image

thanks, but not working with me, I tried -60, -999

voice : https://youtu.be/31kElE83sCU?si=EKnRqK5BJGoGm0x4

model : https://drive.usercontent.google.com/download?id=1IyMdatoJbZKzxTDIFQo7IWK9iWokSYiM&export=download&authuser=2&confirm=t&uuid=14925b56-02a4-4624-a0e9-c784f0ab6045&at=APZUnTV4h6pZUchgQpQXyWAKdqYr:1695316131825

my reverb settings: image

Please help me with the best possible settings

mohmmad-okasha avatar Sep 25 '23 12:09 mohmmad-okasha

Hey, I'm here, still messing with the value. I'm going to send and edit all changes to all variable that I have tried here.

  • index_rate = 1 | Failed
  • reverb_size = 0, reverb_wetness = 0 | Failed
  • remix_mix_rate = 1 | Failed
  • backup_vol = -60, reverb_dryness = 1 (as solution on AICoverGen issue) | Failed
  • reverb_damping = 1 | Failed Looks like it is. It could be because of the song itself or the DeReverb. Because the SociallyIneptWeeb repo includes DeReverb for song smoothness and it's mainly for music to voice conversion.

You should use a male voice model for a male song, female voice model for a female song, since it could match each other sometimes. Try looking up cover version for your song.

You can also use Mangio-RVC-Fork (use your PC) or Retrieval-based-Voice-Conversion-WebUI (use Colab, but the notebook is in Chinese) if you wish.

Make sure to close this issue when you are done.

MinatoIsuki avatar Sep 25 '23 12:09 MinatoIsuki

Hey, I'm here, still messing with the value. I'm going to send and edit all changes to all variable that I have tried here.

  • index_rate = 1 | Failed
  • reverb_size = 0, reverb_wetness = 0 | Failed
  • remix_mix_rate = 1 | Failed
  • backup_vol = -60, reverb_dryness = 1 (as solution on AICoverGen issue) | Failed Looks like the problem goes down to the song itself, or I haven't found that value yet?

can you try with this please ? https://youtu.be/4gQmVyLTvRw?si=tJ6NkSOva_bWO2sS

mohmmad-okasha avatar Sep 25 '23 13:09 mohmmad-okasha

Hey, I'm here, still messing with the value. I'm going to send and edit all changes to all variable that I have tried here.

  • index_rate = 1 | Failed
  • reverb_size = 0, reverb_wetness = 0 | Failed
  • remix_mix_rate = 1 | Failed
  • backup_vol = -60, reverb_dryness = 1 (as solution on AICoverGen issue) | Failed
  • reverb_damping = 1 | Failed Looks like it is. It could be because of the song itself or the DeReverb. Because the SociallyIneptWeeb repo includes DeReverb for song smoothness and it's mainly for music to voice conversion.

You should use a male voice model for a male song, female voice model for a female song, since it could match each other sometimes. Try looking up cover version for your song.

You can also use Mangio-RVC-Fork (use your PC) or Retrieval-based-Voice-Conversion-WebUI (use Colab, but the notebook is in Chinese) if you wish.

Make sure to close this issue when you are done.

the voice for male and the model for male

mohmmad-okasha avatar Sep 25 '23 13:09 mohmmad-okasha

Hey, I'm here, still messing with the value. I'm going to send and edit all changes to all variable that I have tried here.

  • index_rate = 1 | Failed
  • reverb_size = 0, reverb_wetness = 0 | Failed
  • remix_mix_rate = 1 | Failed
  • backup_vol = -60, reverb_dryness = 1 (as solution on AICoverGen issue) | Failed
  • reverb_damping = 1 | Failed Looks like it is. It could be because of the song itself or the DeReverb. Because the SociallyIneptWeeb repo includes DeReverb for song smoothness and it's mainly for music to voice conversion.

You should use a male voice model for a male song, female voice model for a female song, since it could match each other sometimes. Try looking up cover version for your song. You can also use Mangio-RVC-Fork (use your PC) or Retrieval-based-Voice-Conversion-WebUI (use Colab, but the notebook is in Chinese) if you wish. Make sure to close this issue when you are done.

the voice for male and the model for male

This repo is mostly for music cover, not really for voice only. Both your sample only contain voice, and that is why the audio sometime feels weird. Please try out the repo that I have mentioned.

Also, what is that first sample video (https://youtu.be/31kElE83sCU?si=EKnRqK5BJGoGm0x4), looks like it's your country tradition of something? I've seen it on TikTok but I have no idea what is that called.

MinatoIsuki avatar Sep 25 '23 13:09 MinatoIsuki

Hey, I'm here, still messing with the value. I'm going to send and edit all changes to all variable that I have tried here.

  • index_rate = 1 | Failed
  • reverb_size = 0, reverb_wetness = 0 | Failed
  • remix_mix_rate = 1 | Failed
  • backup_vol = -60, reverb_dryness = 1 (as solution on AICoverGen issue) | Failed
  • reverb_damping = 1 | Failed Looks like it is. It could be because of the song itself or the DeReverb. Because the SociallyIneptWeeb repo includes DeReverb for song smoothness and it's mainly for music to voice conversion.

You should use a male voice model for a male song, female voice model for a female song, since it could match each other sometimes. Try looking up cover version for your song. You can also use Mangio-RVC-Fork (use your PC) or Retrieval-based-Voice-Conversion-WebUI (use Colab, but the notebook is in Chinese) if you wish. Make sure to close this issue when you are done.

the voice for male and the model for male

This repo is mostly for music cover, not really for voice only. Both your sample only contain voice, and that is why the audio sometime feels weird. Please try out the repo that I have mentioned.

Also, what is that first sample video (https://youtu.be/31kElE83sCU?si=EKnRqK5BJGoGm0x4), looks like it's your country tradition of something? I've seen it on TikTok but I have no idea what is that called.

He reads the Qur’an, “the Book of God revealed to the people,” and this records from Muslim prayers

you can read in English : https://quranenc.com/en/browse/english_saheeh

mohmmad-okasha avatar Sep 25 '23 13:09 mohmmad-okasha

@MinatoIsuki how can I contact you?

mohmmad-okasha avatar Sep 26 '23 13:09 mohmmad-okasha

@MinatoIsuki how can I contact you?

How about Discord? Also, why?

MinatoIsuki avatar Sep 26 '23 14:09 MinatoIsuki

@MinatoIsuki

how to use UVR-DeEcho-DeReverb instead of Reverb_HQ_By_FoxJoy.onnx ?

https://huggingface.co/seanghay/uvr_models/blob/main/UVR-DeEcho-DeReverb.pth

mohmmad-okasha avatar Oct 01 '23 11:10 mohmmad-okasha

@MinatoIsuki

how to use UVR-DeEcho-DeReverb instead of Reverb_HQ_By_FoxJoy.onnx ?

https://huggingface.co/seanghay/uvr_models/blob/main/UVR-DeEcho-DeReverb.pth

I'm seeing some compatibility issue here. I don't really think you can use pth for onnx. Also, please contact SociallyIneptWeeb as his/her repo is the one that Ardha uses.

https://github.com/SociallyIneptWeeb/AICoverGen

MinatoIsuki avatar Oct 01 '23 11:10 MinatoIsuki