Sonar 8.5 Producer Edition: Best Noise Removal Plugins Without "Underwater" Effects?

[Annabelle]

I'm making this virtual instrument that I like to call "Way-Cool Bell Choir", and I'm working on splitting the original session files into individual notes. But before I do, I need to have a noise removal plugin, since there's this horrible roaring noise with a screech, which might be the studio lights or the studio air conditioners. I wonder if there are professional noise removal plugins, where you can teach the plugin what is noise, what is vocals, and what is music. Especially ones that are able to remove the noise without that crazy "underwater" effect. I know Waves makes Clarity VX, but as I've been informed, there's not a version of Clairty VX that works with Windows 7. Unfortunately, Clarity VX also seems to get rid of the sound of the bells ringing, and just leaves the spoken voices behind.

[reginaldStjohn]

Izotope RX has noise removal abilities.
https://www.izotope.com/en/products/rx.html?srsltid=AfmBOoohDmH3qiPl5xhNzzZytP6MKDBfj13sk9IglYTuLgYOCZEb9eww

[Annabelle]

57 minutes ago, reginaldStjohn said:

Izotope RX has noise removal abilities.
https://www.izotope.com/en/products/rx.html?srsltid=AfmBOoohDmH3qiPl5xhNzzZytP6MKDBfj13sk9IglYTuLgYOCZEb9eww

But from what I hear when I listen to it on demonstration videos, I think there's that "underwater" effect left over.

[Chaps]

7 minutes ago, Annabelle said:

But from what I hear when I listen to it on demonstration videos, I think there's that "underwater" effect left over.

I don't have that program but I have used the noise reduction feature in Adobe Audition 1.5. Those audio artifacts are created when the noise reduction is extreme. Using a lower setting might give you results that, while not perfect, are acceptable. Adobe Audition will also, like you mentioned, allow you to select what is 'noise' in the audio. A very important feature, for sure.

[treesha]

If you upload it somewhere and post a link maybe someone here with newer capabilities could try to clean it up for you. With both screeching and roaring it seems like it could be challenging. Simple stems can separate out different parts but the results aren’t always very clean. Like you say bells and voices are similar too.  Anyways maybe someone here could help you out. 

[mettelus]

5 minutes ago, treesha said:

If you upload it somewhere and post a link maybe someone here with newer capabilities could try to clean it up for you. With both screeching and roaring it seems like it could be challenging. Simple stems can separate out different parts but the results aren’t always very clean. Like you say bells and voices are similar too.  Anyways maybe someone here could help you out. 

@treesha beat me to it, but this sounds like a one-off scenario. In addition to a Noise Print Capture/Removal, it sounds like you also have transient material issues with that file. Both of these are more visually focused with the spectrogram, so even having such an app may cause you more grief than good. If you are willing to post a link (or even send that link in a PM so it is not public), folks here including myself can take a look at that for you and send the results back to you. I typically only sit still in late evening hours, but being able to actually process that file for you would be significantly faster than you trying to learn a new app and do it yourself.

[Annabelle]

2 hours ago, mettelus said:

@treesha beat me to it, but this sounds like a one-off scenario. In addition to a Noise Print Capture/Removal, it sounds like you also have transient material issues with that file. Both of these are more visually focused with the spectrogram, so even having such an app may cause you more grief than good. If you are willing to post a link (or even send that link in a PM so it is not public), folks here including myself can take a look at that for you and send the results back to you. I typically only sit still in late evening hours, but being able to actually process that file for you would be significantly faster than you trying to learn a new app and do it yourself.

Wouldn't that be amazing if someone made an app that could tell what parts are what, such as individual instruments like guitar, without changing the "format" of the audio? I think that's what it's called. Or is it "formant"? I'm confused on that one!

[mettelus]

1 hour ago, Annabelle said:

Wouldn't that be amazing if someone made an app that could tell what parts are what, such as individual instruments like guitar, without changing the "format" of the audio? I think that's what it's called. Or is it "formant"? I'm confused on that one!

Noise removal/stem separation tools are rarely something that are "fire and forget" unfortunately. Pretty much all instruments overlap the midrange, so the timbre is often what differentiates them. Even with advances in AI algorithms, many instruments (voice included) have frequency ranges that get separated into the wrong bucket, so there can be a lot of surgical manipulation/recombination required to get the actual instrument isolated. The higher end tools (e.g., iZotope RX, Steinberg SpectraLayers, etc.) actually have drawing tools for the spectrogram to assist in that surgery. My ultimate concern is that most audio apps deal with visual representations of audio, so unless it is a "fire and forget" operation, you could be delving into something you weren't anticipating.

Again, if you are willing to post that audio for others to look at (either here or via private messages), I think you might be surprised by those willing to help you out. It also gives folks an immediate understanding of the complexity of your undertaking, or even alternative methods to achieve your goals.

[sjoens]

Your best bet is to record the samples "clean" in a "quiet" room or else only record during the quiet cycle of equipment like refrigerators.

Noise separation is hit and miss for most software.  Normally they rely on a constant sound that can only be isolated if that sound is already singled out somewhere on the recording.  But since some sound frequencies overlap, removing them can still cause some degradation to the remaining sound - which can give the underwater effect you mention.

Melodyne can sometimes separate such sounds into blobs that can be individually deleted.  It might work with Sonar 8.5 as a VST but may be better used standalone.

[Michael Elliott]

16 hours ago, Annabelle said:

  But before I do, I need to have a noise removal plugin, since there's this horrible roaring noise with a screech, which might be the studio lights or the studio air conditioners. I wonder if there are professional noise removal plugins, where you can teach the plugin what is noise, what is vocals, and what is music. Especially ones that are able to remove the noise without that crazy "underwater" effect. 

The only things I have used are FFT types like the ancient CoolEdit program used (I think Audacity and other programs also have this available, but Audacity's usage of it is harder to do the way I need to).  If your sound file has a portion that does not have any signal in it, just the noise, and this is the same noise that is in the signal portion, you can select the noise to teach the FFT what to remove, then apply that to whatever portion of the file you need to.   It's not perfect, but it's good, as long as it is the *same* noise (like a droning fan, etc).  If it is a bunch of random sounds  that need to be removed, then it doesn't work very well.    

 

A while back I checked for "AI" sound filter webpages, and found some, but I don't think I got around to trying them out for whatever reason. 

 

 

[Annabelle]

1 hour ago, sjoens said:

Your best bet is to record the samples "clean" in a "quiet" room or else only record during the quiet cycle of equipment like refrigerators.

Noise separation is hit and miss for most software.  Normally they rely on a constant sound that can only be isolated if that sound is already singled out somewhere on the recording.  But since some sound frequencies overlap, removing them can still cause some degradation to the remaining sound - which can give the underwater effect you mention.

Melodyne can sometimes separate such sounds into blobs that can be individually deleted.  It might work with Sonar 8.5 as a VST but may be better used standalone.

1. The recordings were made by me at Synergy Studios in Seattle in 2008.

2. I'd love to try Melodyne. Is there a version that works with Windows 7?

3. Better yet, is it screenreader accessible?

[Max Arwood]

I have almost every noise reducing and reverb reducing etc software made including VX clarity. I would be glad to listen to it too.  PM me would work. 

[Max Arwood]

7 hours ago, Annabelle said:

Wouldn't that be amazing if someone made an app that could tell what parts are what, such as individual instruments like guitar, without changing the "format" of the audio? I think that's what it's called. Or is it "formant"?

With all the new AI apps this is getting closer to a reality. In the last year or so there have been some great AI advances. 
like Clarity Vx Pro by Waves. I am expecting many more in the next few years. 

[sjoens]

9 hours ago, Annabelle said:

2. I'd love to try Melodyne. Is there a version that works with Windows 7?

3. Better yet, is it screenreader accessible?

Looks like v4 was the last one to work with Win7 but you might ask Celemony about those points.  info@celemony.com

Edited Sunday at 12:33 PM by sjoens

[Annabelle]

38 minutes ago, sjoens said:

Looks like v4 was the last one to work with Win7 but you might ask Celemony about those points.  info@celemony.com

I wonder if version 4 is still available.

[treesha]

I’ve used melodyne a lot for many reasons both corrective and creatively, and I have tried to clean up things with it. Sometimes it does well, other times a blob I delete that I believe to be noise or something unappealing- when I delete that blob it dramatically and undesirably changes the sound of what I’m trying to improve. It works best for simpler things, but say a guitar with reverb the array of blobs is widespread and deleting one can make unwelcome changes. Rich material produces complex blob material that may or may not be helpful to use for cleaning up tasks. 

[Bristol_Jonesey]

1 hour ago, Annabelle said:

I wonder if version 4 is still available.

As an alternative, you do have V-Vocal with Sonar 8.5

[sjoens]

Been a while since I've used that.  For me, V-Vocal can only be used in 32 bit DAWs as it crashes every 64 bit SONAR/CbB I have, so it's unusable here.

It also doesn't seem as intuitive for separating out different sounds if it can at all.

Edited Sunday at 09:36 PM by sjoens

[Byron Dickens]

From what I have read and experienced, noise removal is best accomplished in multiple subtle passes instead of trying to do it all at once. Even better is doing it with several different tools. Also, it is always better to do too little than too much. 

[John Vere]

Ah get out the turd polish! But it’s hard to make a turd into something else other than a turd. 
 

The bottom line is we can easily fix minor flaws in audio recordings but only a miracle by someone who is very knowledgeable and skilled can save audio recordings that are very flawed. It’s wishful thinking that AI can replace a human that has that ability. For now anyway. 
 

Most manipulative audio software does result in various degrees of artifacts. Even Melodyne requires a very light touch. And I doubt it would be ideal for you as its whole design is based on visual editing of audio. 
It works like a midi piano roll but the blobs are tiny pieces of audio. You have to have the Studio version to get the level of editing you would need for noise removal. I think it’s like $700. 
 

I would definitely take up the offers for free help with this. 

But the only real solution is to re record audio recordings that had issues. 

 

Edited yesterday at 02:54 AM by John Vere