inference.ipynb audio output silent

[abaddon-moriarty]

Hi, I'm trying to run the inference.ipynb with the published models, neither model got errors. Also there are no problems with the plot results: But when I try to play the audio output, there is no voice, my speakers are definitely detecting a sound but it's just white noise, and yet again there are no explicit errors. I'm very new to voice synthesis, there might be something I overlooked, but I followed the indications to the t, so I'm not sure what could have gone wrong.

I have looked through other issues but I haven't seen any problem similar to this, if I have missed it, thank you for redirecting me to those issues. any help would be appreciated.

[holynova-SD]

I also have this problem, have you solved it？

[abaddon-moriarty]

@holynova-SD No, not yet, sorry. I access it by remote server. I was thinking maybe it had an impact on the ability to get the output back? (not sure why it would though, but I don't see what else could be the problem)

Are you running notebook locally?

[41WhiteElephants]

Probably waveglow is producing ndarray of NaNs.

to debug try: print(audio[0].data.cpu().numpy())

Change loading model to this, it worked for me: waveglow = torch.hub.load('nvidia/DeepLearningExamples:torchhub', 'nvidia_waveglow')

[mohamad-hasan-sohan-ajini]

Hi Avoid to transfer waveglow weights to fp16 will solve the issue.

regards

[anaivebird]

just remove .half() twice on ipynb will solve this problem

[wsywsywsywsywsy979]

just remove .half() twice on ipynb will solve this problem

Hi，inferece code still generates empty outpt(0.00s) after I remove the .half() twice ,could you give me other advice? (I run the code in docker,I don't know whether it's the main reason.)

[sun-rabbit]

hi,I run inference.ipynb directly on colab，but，But I made this mistake. How can I solve it

[Nicksooooo]

just remove .half() twice on ipynb will solve this problem

Thank you very much.i have succeeded

[wsywsywsywsywsy979]

just remove .half() twice on ipynb will solve this problem

Thank you very much.i have succeeded

Thank you, I'm also succeeded. ：）