iSTFTNet-pytorch

iSTFTNet-pytorch copied to clipboard

Repgan

1

Pranjalya

A sample as good as HiFiGAN

2

tks very much!!!!  https://user-images.githubusercontent.com/16432329/177454772-43428023-5ed3-4c9e-b8af-fb68f1a11ff3.mp4

MaxMax2016

Single frequency line problem

14

Thanks for the implemention of ISTFT. It has better inference speed than hifigan v1.However, I found that there is a single frequency line which would cause little noise.I use 16KHZ...

mayfool

How about the audio quality?

6

Hi, thanks to the implement, the inference speed is impressive. How about the audio quality? And have you tried v2 config? Thanks in advance.

OnceJune

how about the quality of this net

4

Have you got good audio?

hdmjdp

As the issue https://github.com/rishikksh20/iSTFTNet-pytorch/issues/1, the line 164 in `stft.py` was changed to https://github.com/rishikksh20/iSTFTNet-pytorch/blob/e928a6b604033a3857757562af36241f9225adfc/stft.py#L164 But `inverse_transform.device()` will raise the exception mentioned in the title. So it can be changed to `inverse_transform.device`...

leminhnguyen

hi. does this repo implements tinyVITS?

1

https://arxiv.org/abs/2206.00208

lucasjinreal

 https://github.com/rishikksh20/iSTFTNet-pytorch/blob/ecbf0f635b36432bd3e432790326591bc86cadbc/models.py#L97 https://github.com/rishikksh20/iSTFTNet-pytorch/blob/ecbf0f635b36432bd3e432790326591bc86cadbc/config_v1.json#L16 Why is fs 16?

a897456

Predicted phase not in range [-pi .. pi], but in range [-1 .. 1]

2

The phase output of the generator currently can only range from -1 to 1, which is not enough as full phase in radians is expected later in `stft.inverse()` (either 0..2*pi...

kgoba

Has anyone tried to directly model the complex numbers instead of the phase and magnitude? What would be the problem if we model the real and imaginary parts directly?

yl4579