Do you know any good Noise Remover VST plugin?

[LCF-AT]

Hi guys,

so I'am trying to find any good plugin I can use with any audio editor tools supporting VST input.I found some free VST here..,

https://www.reaper.fm/reaplugs/

...and in this set you can find the plugin called ReaFIR.Its working ok so far +/- but its already a old plugin and some similar internal plugin I have also in CoolEditPro 2.0.Now I thought that after 20 years there is any new plugin/s out there I could try out to get some better results now.On internet I just found some VST called "ERA Noise Remover" pretty offten but its not free and also found no download for it.Did anyone test this out and can tell whether its a good one (better as ReaFIR)?Otherwise I can just use some internal Noise Remover of any audio tools like audacity to get similar same results like with ReaFIR +/-.So my goal is it to remove noises of spoken audio.In most cases I get it work (more or less) but mostly the results sounds pretty aqua etc you know.So if you know some better VSTs / tools I could obtain better results then just tell me.Thank you.

greetz

[Teddy Rogers]

Last time I did anything with VST I used ReNoise (commercial) and Audacity (open source)...

Ted.

[Extreme Coders]

Have used iZotope RX 7 (commercial). Good stuff. Works great for noise removal.

[LCF-AT]

Hi guys,

thanks for the infos so far.The Audacity tool I tried already but I don't get pretty good results out (similar like with ReaFIR).Its ok but no highlight you know.I tried to find the tools you are talking about guys and found first that iZotope RX 7 one I tried to check out to remove noises and I'am really surprised.Right one first shot I got SUPER results.All noises are gone (95%) that the voice who is talking does still sound original (No aqua / flanger / under water sound).Thats HOT! 🌶️

But there is also a bad news about that tool.The reults I get out are pretty good BUT there is a big disadvantage.....I can not export the good results in a new audio file! Yes of course I can save / export any selection to MP3 / Wav and other formats but I don't get the same resuts out in the saved file.Seems to be any bug.Both selections (I have saved as new file and selection in the app) do sounds diffrent and I can also hear noises right at the beginning part.Thats really bad.Did you notice the same issue by using that app @Extreme Coders?

PS: I tried to export the good selection in the app with all formats and high quality but with all I get the same bad results out.

greetz

[LCF-AT]

PS: Ok I found the problem.On Windows itself I had enabled the volume equalization effect to hear more and iZotope does not use it (all other audio apps using it) that makes the diffrent in hearing.All fine now so far with iZotope and the export results.

Since I did update Windows to 20H2 the sound sounds diffrent as before.I also had to enable the volume equalization effect to hear more but also in enabled mode I have to increase the loudness control higher than before the update.Not sure what WIndows use for other audio drivers now.

greetz

EDIT: I see a problem trying to load AAC files.The app does do something / trying to load / but I don't get any wave or time to see = all empty.Nothing loaded.On the HP I can read this...

Supported file formats in RX

Supported Audio Import Formats:
WAV, Broadcast WAV, AIFF, MP3, WMA, AAC, SD2, BWF, OGG, FLAC

Supported Audio Export Formats:
WAV, Broadcast WAV, AIFF, OGG, FLAC

Supported Video Import Formats:
AVI, MPEG, Windows Media Video, MOV, M4V

....AAC in mp4 container dosent work to load.Avi and WMV also do not work to load so there I get a Bad Format Error message.Only thing I can do is to convert AAC to MP3.Uhhm!

Edited April 24, 2021 by LCF-AT
New Info added

[Extreme Coders]

Generally wav is the best format for import/export as it's lossless and almost all tools support it. So there won't be any quality loss when exporting it from one tool and importing it in another. Converting the final wav to a suitable format (mp3/aac/etc) is best done at the last stage when no further post-processing work is required. For inter-converting audio formats there's always ffmpeg, Handbrake, Avidmux.

[LCF-AT]

Hi again,

just trying to find still a solution with iZotope and audio input / output devices to get the same to hear as in all other tools who playing audio.

Example: Below a image of my sound device and what I have enabled.I have to because otherwise its just too quietly so I have to turn it on.The difference between on/off is hughe and in editor tools I need to have the same loudness as outside.On image below you can see its on.

So normaly to change anything in this setting above does change the whole global sound which get played from any app like  VLC player or CoolEdit or other app & tools but  unfortunately it dosen't work with iZotope and the sounds keep playing like in deactivated modus and I don't understand why and how to make it work same like with all other apps.So in iZotope I have a audio settings tab where I can set some devices what I tried already but its also not working...

...I really don't know how to get it work to hear the same loudness which get used for all other tools too.So did anyone else had the same problem with that iZotope tool and found out how to fix it etc?Its pretty bad that its not working yet for me because I need to edit in a gloabl loudness of course.Just asking because I found no solution for that and  maybe some of you have any clue what I could do etc.Thank you.

greetz

[CondZero]

I know this post is relatively old, but I have code in a music tool I developed in concert with already existing code that among other things can reduce noise. The problem as you know with any of these tools is that the higher frequencies many times will be diminished resulting in a need to attenuate what's left afterwards. Do I know they work, I was able to reduce a noise sample to the original sine wave that generated the noise. Audacity is a fine program, but it too has its limitations. PM me and I can forward you the code. It's written in C++.

[LCF-AT]

Hi CondZero,

sorry for my late answer.So you did wrote a own tool what can reduce noises?Or did you create a plugin etc?I also tried using Audacity but its not really good.So at the moment I get best results by using iZotope's Dialogue Isolate function to remove noises almost completely and keeping the voices sounding normaly (no Aqua or robotic style etc).So if you got some similar tool I should give it try then just show me. Below I did attach a example of a original audio snipped with much noises and the fixed one.Results are pretty good compared with the original one.Not sure whether its possible to make that better.

AudioNoiseExampleFiles.7z

greetz

[CondZero]

I can't seem to get to your download. To answer your question, my code is not a VST plug in. It relies on FFT so not sure how viable it would be as a real time noise suppressor. I adapted the code from my image processing program to work on music files. If you look at my attachment and study it, you can see what's going on. In essence, you are performing (2) different kinds of standard deviations to find the noise threshold within a given frame of the music file. We divide the file into frames for FFT purposes. Then we can use 1 of (4) different algos to process any music that exceeds our threshold. You will notice that I use type doubles as I've found that music files are very susceptible to minute differences in calculations with rounding errors skewing results. You will notice I've used OpenMP for speed.

The function (ar_parameters1) is the Reflection Coefficient and is less intrusive at reducing noise vs (ar_parameters) which uses std. deviation. You set the threshold lower for less noise or higher in value for more noise removal. It works very well but you run the risk of reducing the higher frequencies as you remove more noise. You will need the FFT library. In this case I'm using DCT vs FFT (real numbers vs complex numbers). Take a look at the code. 

The function (DCTdenoising) has (2) levels, 1 for each selected deviation you choose. As I've said, Level 1 (reflection coefficients), Level 2 (standard deviation) which is more intrusive at reducing noise.

Within each level, there are (4) algos which progressively reduce more noise. We're talking fractions of fractions of difference between them. If you use this function properly, you should be able to get the desired results.

Cheers

DCTdenoising.rar

[LCF-AT]

Hi,

so the download works here (AudioNoiseExampleFiles.7z).

Sounds all interesting of course but I'am not into C++ language yet.So my question is this case would be to ask you whether you made some tiny GUI for your functions or any CLI tool with commandline args to use it with any audio files?Just would like to test or hear any results you know.Otherwise try again to download my sample file package.Inside you can find 2 short audio files I made...

Original_audio_1.mp3
NoiseIsolate_audio_1.mp3

...you can hear the diffrents.Now use the Original_audio_1.mp3 file and use it with your functions to get the possible best results to remove the noises and compare it with the NoiseIsolate_audio_1.mp3 file.Are yours better?If you think so then attach your sample of it to hear it.

greetz

[CondZero]

OK, I had to login to Edge and not use Chrome (yeah, I know I shouldn't be using either) to get at your files. 

I'll test them out and get back to you. I just did a quick check. A couple of things. You are attempting to remove noise

from an audio track vs a music track. Different techniques are involved in so doing. I will run your example through both

and you can hear the slight differences. Along with ambient noise in the background is also a lower frequency humming noise.

This may entail a high pass filter to get rid of or we could use a notch filter to center on the voice frequency. These type things

are normally handled by a strictly audio type filter for voice. Singing with music is even more complicated if using an audio / voice

filter.

Edited August 17, 2021 by CondZero

[CondZero]

Unfortunately, I can't replicate your good denoised file to the level you have. I can get close but not as clean. Your clean file should make you happy. As an aside, as I've said, denoising music vs audio (voice) are 2 different animals. I am using RNNOISE available on Github written in C/C++ which I included in my audio program w/o the AI neural learning stuff. It works pretty well, but not as well as what you used for your file. It's pretty amazing stuff to see what is involved in denoising (or should I say Isolating) the voice from any and all surrounding interference. Good Luck to your success.

cheers

[LCF-AT]

Hi,

ah ok.Thanks for trying and comparing.Yes, the "Dialogue Isolate" function works pretty good and seems to be the best de-noiser I found & tried yet.I many cases I have to de-noise around sounds from talking sessions because some people don't do any sound checks before starting a record session and they also don't do any post processing which then results into horrible end-results.Most just don't care & share or sell. I really 'F' Hate that.

Would be nice if players like VLC would use some smart (AI) realtime de-noiser someday to prevent doing any own extern fixing anymore. But I think they are far away from that.

PS: Why has the Youtube player (all others too) no option to change the audio to MONO track?In diffrent cases I need to hear the mono mode if both audio channels are not same or only one has audio etc.Pretty strange right.Two channels / one is used / left or right side +/- and you got to hear the sound in a uncomfortable way.I never found any player on internet yet who has a Stereo/Mono switch button.I think that should be a must have function specially for people who having broken/damage/limited hearing.Just only a idea.

greetz

[CondZero]

Back in the old days, they used studios to isolate sound interference as well as provide acoustical dampening of sound. Why many recording have a very dry sound. It is plain stupid to record in any other environment if at all possible. As for the mono / stereo thing, this would be easy to implement. Keep in mind though, it's easy to implement using a wave file. You can change the header spec and alternate the wave data accordingly. Very easy stuff.