soundfile.LibsndfileError: Error opening 'xx/xx/xx.wav': File contains data in an unknown format.

[QHCV]

I find a problem when training my model:

soundfile.LibsndfileError: Error opening 'xx/xx/xx.wav': File contains data in an unknown format.

I run this in Ubuntu

I have tried many methods. What can I do about this problem? Think you!

[Plutoisme]

ffmpeg is all you need, i meet the same question in other projects, you can try conda install ffmpeg.

[QHCV]

ffmpeg is all you need, i meet the same question in other projects, you can try conda install ffmpeg.

The problem is the same.

[MonkeyDNoppu]

ffmpeg cannot solve my problem (fyi, I'm trying to asr a mp4)

[QHCV]

I'm very sorry, but it turned out that the audio files in the dataset were corrupted, which caused this error. The dataset I used is https://github.com/microsoft/DNS-Challenge/tree/interspeech2020/master. I switched to other data sets and there was no error, thank you for your help.

[Plutoisme]

when you used the DNS-Challenge dataset, did you use the shell file 'download-dns-challenge-4.sh ' to download the data?

[QHCV]

when you used the DNS-Challenge dataset, did you use the shell file 'download-dns-challenge-4.sh ' to download the data?

No, I use the DCN 2020 dataset.

[Plutoisme]

when you used the DNS-Challenge dataset, did you use the shell file 'download-dns-challenge-4.sh ' to download the data?

No, I use the DCN 2020 dataset.

Sometimes the similar bug is relate to the encoding format，I usually use the same function to deal with different .wav files, sometimes show bug, ffmpeg may solve a part of the problem.

[QHCV]

Sometimes the similar bug is relate to the encoding format，I usually use the same function to deal with different .wav files, sometimes show bug, ffmpeg may solve a part of the problem.

Thank you for sharing your experience.