Anyone know of a plugin to remove noise?

This indicates an error in the ttl file for the plugin. The LV2 library we use is now more strict about these errors.

see http://lv2plug.in/pages/validating-lv2-data.html

For the case at hand probably just a missing @prefix at the top of the file:

@prefix pprop: <http://lv2plug.in/ns/ext/port-props#>

I’m sorry for all this trouble caused. I believe that it’s fixed now. At least sord_validator_lv2 is not giving me anymore errors. pprops was already there I just forgot to put ‘.’ at the end. I’m testing this on Ardour 4 so I don’t know if this it’s fixed in 5. Let me know if is the case or file a bug report in the github site. Again sorry for being a noob in all this.

Hi lucianodato,
Yes, it builds and installs and shows up in Ardour5 now.
Thanks a million for this. I will find some noisy audio files and try it out!

It only needs a few small changes to compile on OSX:

The Makefile checks for “fftw3” presence (but really “fftw3f” is used). You need pkg-config, lv2-headers and fftw3f (all of which are easily accessible on OSX, ideally use a static lib of fftw3f).
The LDFLAGS are wrong (simply remove the -Wl,–as-needed -Wl,-Bstatic -Wl,-Bdynamic). Here just make LDFLAGS="-shared" pkg-config --libs fftw3f did the trick.

For good measure you should add -arch i386 -arch x86_64 to the CFLAGS for multi-arch binaries and you’ll end up with http://robin.linuxaudio.org/tmp/noise-repellent-osx-git8edcfd7e6.zip (untested, unzip in ~/.lv2 to get ~/.lv2/nrepel.lv2)

Windows will be a bit harder, but not much :slight_smile:

@calimerox That’s not the intended behavior. Noise Whitening is just a post process option (It tries to make the residual noise more like white noise, it will sound brighter). I will check what I’m doing wrong maybe it’s a bug. The plugin is mono for now and I didn’t try yet but it seems that with plugin connections that could be resolved using 2 instances of the plugin. I wouldn’t say it’s way better than audacity denoiser maybe just a bit in certain cases. The thing it’s that this one is “real time”, so it’s much more useful for tweaking the settings. This project is a part of my engineer thesis, I saw the opportunity to bring something useful for the community and to force myself to be involved with it. Besides I need this too! so it’s a win win

this is great! yes that is my problem with audacity, until you find the right settings you have to go through x iterations of tryouts … i ll test it again on my studio desktop computer and will let you know if i find some bugs…

@calimerox Great! Thanks! Just make sure to get the latest master

@lucianodato

Haven’t said thanks for working on this yet, but thanks:) This is something sorely needed in the open source community. If I can get things to settle down (Just finished working a 15 hour day for a festival with long stories, and a week of 12+ hours prepping for it) I am looking forward to trying it and comparing it to other options out there.

Is there anything that would prevent it from working cross platform?

     Seablade

@seablade My makefile is pretty basic, I suppose I can look at other projects that do support compiling for other platforms and see how to pull that off. Be aware that this is nowhere near close to professional stuff like izotope or acon digital or others. But it’s something. Eventually it will get better and who knows maybe someone get interested and collaborate to improve it.

@lucianodato

The build system itself isn’t usually to difficult to make build crossplatform, it is more a question of libraries etc. you used. I can take a look at it, as I mentioned when I get time, but if there isn’t anything very linux specific you know of in your code that is a good sign. It is usually easier to port linux software to other platforms because so many libs are cross platform, than the other way around where people might use apple or MS specific libs.

      Seablade

@ lucianodato no reason to apologize, it is just great you develop this plugin! ! I m was able now to compile and it shows up in ardour5!!

I can press the learn button and it captures some noise (as i can see in the plugin analysis window) Although it seems i dont hear any reduction or difference to the source file. maybe i have to check the plugin connections again. but it usually should work immediately after pressing “noise listen” and releasing, changing the reduction amount should be audible already, right? Thanks again for your effort and work!

a little update: I tried on a monotrack and it works!! I also had overseen that i have to select “noise whitening” to have the reduction work . I tried it on a noisy dialog track in a car and just as a fast try it does a very good job! (and yes way better than the standard audacity noise reduction)

Heh I see x42 is ahead of me like usual:)

   Seablade

Noise-repellent plugin woks grate for mi
I was struggling with a background sound in 1 of my recording, until now, I’ve deal with it using a combo of gate and narrows equalisation of the problematical frequencies, but it was un-natural.
Now with this plug-in, it work out of the box… capturing the parasite sound and adjusting strenght and smoothing. Sound natural, way beter than Audacity solution, and adjustable in real time.
Thanks so much.

Hi there,
I must say that I was quite doubtfully about noise-reduction with this open-source plugin, but I finally found Noise-repellent very efficient ! I’m so enthusiastic now after having tested it. It’s simple to use, well documented, and it sound really not so bad when you find the right settings.
Thanks so much lucianodato for this very nice tool ! Well done !
Cheers

@lucianodato: I just cloned the repository and built this plugin. Before I spend much time working on it, would you expect the plugin to help with noise that is much higher than typical power related hum frequencies? I have a recording of an interview that I want to clean up, and some type of alarm began sounding part way through the interview.
I used Sonic Visualiser to look at the spectrum, and it seems to be a tone at around 1800 Hz, with some smaller amount of energy at
600Hz, 3000Hz, and 4200 Hz (plus or minus few hundred Hz) for a few seconds, then changing to a strong tone at 870Hz with harmonics at around 1800 Hz and 2700 Hz for a few seconds, then a tone at 650Hz with harmonics at 1970Hz and lesser higher harmonics. It is quite irritating to listen to, which I suppose is the point of the alarm, but it was quite inconvenient to have the alarm sound during an interview.

@ccaudle for pathological cases you want to check out Audacity’s spectral editing: https://youtu.be/pGku4-CJnzc

@ccaudle

Typically I might look into hum/buzz removal rather than broadband noise reduction for your case. Two different tools for two different uses. A typical hum/buzz removal tool will let you specify the fundamental and the strength of the harmonics to remove. As x42 mentioned spectral editing tools also allow similar end goals as well though require a destructive workflow (Not a bad thing, in fact I find destructive editing better for noise removal, though I may still utilize the hum/buzz tools first and spectral editing if that didn’t work).

Seablade

Robin is right spectral editing is far better for these cases. Noise-repellent algorithm works for stationary noises.This type of noise do not change in frequency and in time, at least not that much.