RNNoise 0.2 Released With AVX2 Optimizations For Neural Network Noise Suppression

[phoronix]

Phoronix: RNNoise 0.2 Released With AVX2 Optimizations For Neural Network Noise Suppression

Xiph.Org has released RNNoise 0.2 as the recurrent neural network for audio noise reduction. This library leverages a neural network model for enhancing real-time noise suppression...

https://www.phoronix.com/news/RNNoise-0.2-Released

[Anux]

Not sure what I should think about this, the best was always without any noise suppression.

[Quackdoc]

I highly reccomend people also try https://github.com/Rikorose/DeepFilterNet. I got results comparable with RTXnoise with it. and it's for sure my favourite one right now, Ill be testing 0.2 rnnoise against it for sure

[slalomsk8er]

I highly reccomend people also try https://github.com/Rikorose/DeepFilterNet. I got results comparable with RTXnoise with it. and it's for sure my favourite one right now, Ill be testing 0.2 rnnoise against it for sure

Please, let us know how it went.

[Anux]

I highly reccomend people also try https://github.com/Rikorose/DeepFilterNet. I got results comparable with RTXnoise with it. and it's for sure my favourite one right now, Ill be testing 0.2 rnnoise against it for sure

It is a bit better then rnnoise but still worse than no suppression (judged from the github sample video).

[Quackdoc]

It is a bit better then rnnoise but still worse than no suppression (judged from the github sample video).

You can't judge based on one video, you need to judge your usecase, below are clips tested from an original I recorded. let me know if I need to upload them else where. but here deepfilternet is massively preferable for me then no suppression

EDIT: reuploaded

original

https://files.catbox.moe/l93v64.flac

rnnoise

https://files.catbox.moe/61n812.flac

rtx voice

https://files.catbox.moe/rgwfqa.flac

deepfilternet

https://files.catbox.moe/wrmcx4.flac

[slalomsk8er]

You can't judge based on one video, you need to judge your usecase, below are clips tested from an original I recorded. let me know if I need to upload them else where. but here deepfilternet is massively preferable for me then no suppression

deepfilternet
https://files.catbox.moe/wrmcx4.flac

Thanks, a lot. I agree, it depends heavily on the usecase.

Last link has some garbage attached but worked after I removed it.

[Quackdoc]

Thanks, a lot. I agree, it depends heavily on the usecase.

Last link has some garbage attached but worked after I removed it.

weird, anyways this isn't rnnoise 0.2 but 0.1, I havent installed 0.2 yet. ill ping you when I do

[Quackdoc]

slalomsk8er

https://files.catbox.moe/l8pqv3.flac

It's not too bad, For sure an upgrade but I still prefer deepfilternet

[Anux]

deepfilternet

THX, deepfilter is clearly the best but your original sample was already super easy to understand, no suppression needed. The samples from the rnnoise web page at 0 db with street noise were much more problematic maybe I try deepfilter with that.