search

wwmm / easyeffects

Limiter, compressor, convolver, equalizer and auto volume and many other plugins for PipeWire applications
GNU General Public License v3.0
6.5k stars 269 forks source link

ladsp plugins #2402

Closed chiddekel closed 1 year ago

chiddekel commented 1 year ago

EasyEffects Version

7.0.4

What package are you using?

Fedora

Distribution

Fedora 38

Describe the bug

Wish list - add suport for this:

ladspa-realtime-noise-suppression-plugin

copr repo - lkiesow/noise-suppression-for-voice

Expected Behavior

No response

Debug Log

Debug Log
``` Paste your log here ```

Additional Information

No response

wwmm commented 1 year ago

What does it do that can not be done with our noise reduction and speech processor plugins?. Our speech processor plugin has voice detection and a general purpose noise reduction while the noise reduction plugin based on rnnoise works well for noises that match its training.

At this moment we do not have any ladspa wrapper. One would have to be written from scratch and would be used only by this plugin. And in its source code I see it also uses the rnnoise library. So it does not seem there is any advantage in adding support for it.

chiddekel commented 1 year ago

How accurate this plugin is comparing to Nvidia noises cancellation?

wwmm commented 1 year ago

How accurate this plugin is comparing to Nvidia noises cancellation?

As the last Windows I used was Windows 98 I have no idea about how they compare. Considering that Nvidia is a big company they kinda have the obligation of being better. But I do not know.

One weakness of any plugin based on rnnoise is that its neural network was trained for specific types of noise. When the noise you want to filter matches the neural network training it works really well. But not so much on the other cases. There are ways to create a custom training. And you can load the file in EasyEffects. But I have never tried to create one. The rnnoise page on github have some pointers in this regard if I am not mistaken.

chiddekel commented 1 year ago

How about using the noise suppression api for extend model capable to more advnaced use case with start on esayeffect or direc lasdsp plugin use? https://ts2.space/en/a-comprehensive-guide-to-google-cloud-speech-to-text-apis-noise-suppression-techniques/

wwmm commented 1 year ago

How about using the noise suppression api for extend model capable to more advnaced use cas

It doesn't seem to be suitable for realtime audio processing. And being a cloud solution there is also the problem of depending on the internet connection. It makes sense on a web browser. But it does not seem a good match for our case.

wwmm commented 1 year ago

How about using the noise suppression api for extend model capable to more advnaced use cas

It doesn't seem to be suitable for realtime audio processing. And being a cloud solution there is also the problem of depending on the internet connection. It makes sense on a web browser. But it does not seem a good match for our case.

chiddekel commented 1 year ago

How about using the noise suppression api for extend model capable to more advnaced use cas

It doesn't seem to be suitable for realtime audio processing. And being a cloud solution there is also the problem of depending on the internet connection. It makes sense on a web browser. But it does not seem a good match for our case.

If any of this interest - There is container options for mentioned api.

https://cloud.google.com/speech-to-text/priv/docs

wwmm commented 1 year ago

At this moment there are 3 noise reduction plugins in EasyEffects. It should be enough. Let's close this issue.

chiddekel commented 1 year ago

I'm out of Computer 🖥️ - can nam it?

wwmm commented 1 year ago

I'm out of Computer 🖥️ - can nam it?

Recently an external contributor added a plugin based on the DeepFilterNet library. As all noise filters is not perfect. But it works quite well in my personal experience. It will help to cover the cases where the other 2 fail.