some clip in tail

[deyituo]

As it needs to pad some zeros to fit the downsample ratio, I find that it may cause the mel to have large values in the last frames. But I cannot find the reason. And if I pad more self.downsample_ratio frames, it will be ok. emm...

code: https://github.com/haoheliu/voicefixer/blob/main/voicefixer/restorer/model_kqq_bn.py#L70

pad_len += self.downsample_ratio

[johannahom]

I also have this issues with some audios, but cannot seem to fix it. I added more padding but it did not help.

[sciai-ai]

any update on this?

[haoheliu]

Sorry for the late update. I'm trying to look into it recently. Will inform this thread once it's fixed.

[Afiyetolsun]

Hey @haoheliu! How is it going? Do u need help with fix?

[Afiyetolsun]

Error still exist, i think length can be cutted to by %SampRate of restore_inmem fft