Noise Reduction Software

[sjaudio]

I need to get some noise reduction software for my DAW. I run Sonar 5 so it has to be PC/VST or Direct X compatible, and I don't want or have the money to buy Waves. Any suggestions?

[mrufino1]

Soundsoap

[cdbabel]

Soundsoap

Sound Soap is the most affordable solution and the results can be fantastic. Bias, the company that produces SoundSoap, makes very stable and easy to use products (like PEAK).

I recently used SoundSoap to remaster an old french tape for my college. The voice on the cassett was almost inaudible when I started and its now crisp. However, due to the level of noise reduction I wanted to achieve, SoundSoap did cause a significant amount of artifacting, as if I had compressed the audio file into a low fi MP3.

As far as noise reduction on music goes, use it sparingly. I wouldn't use SoundSoap to get "studio Silence", but it is useful to use it lightly on tracks with things like computer hum or florencent lights. I've found it almost indispenable for live shows.

SoundSoap also is very good at learning noises. It needs only about 2 seconds of the noise that you want removed in order to remove it. I suspect it samples the noise and plays back the inverse wave of the noise to cancle out the noise (this was an old world war 2 trick, where they would use white noise at certain intervals to make a message inaudible, but then on the recieving end they'd play back the inverse wave of the white noise to restore the origional message., Its my best guess on how SoundSoap preservse most everything but the noise, but I have no proof for this.)

I've only used the non-pro version of SoundSoap 2, but I definately can recommend it for ease of use and sound quality (though all noise reduction comes at a price.)

SoundSoap is VST compatable.

[JGriffin]

However, due to the level of noise reduction I wanted to achieve, SoundSoap did cause a significant amount of artifacting, as if I had compressed the audio file into a low fi MP3.

That's a common by-product of over-using most any noise reduction software, not just SoundSoap (my definition of "over-use" is based in a personal aesthetic; I'd rather leave a little noise in than push the sound so far that the "bad mp3" artifacting results). From your description, it seems to work the same way as other noise-reduction softwares, by taking a sample of what's "bad" and filtering that part of the waveform out. I doubt they simply invert the wave, though, I believe even back in the early 1990s with Sonic Solutions' NoNoise it was a bit more complicated than that.

[cdbabel]

However, due to the level of noise reduction I wanted to achieve, SoundSoap did cause a significant amount of artifacting, as if I had compressed the audio file into a low fi MP3.

That's a common by-product of over-using most any noise reduction software, not just SoundSoap (my definition of "over-use" is based in a personal aesthetic; I'd rather leave a little noise in than push the sound so far that the "bad mp3" artifacting results). From your description, it seems to work the same way as other noise-reduction softwares, by taking a sample of what's "bad" and filtering that part of the waveform out. I doubt they simply invert the wave, though, I believe even back in the early 1990s with Sonic Solutions' NoNoise it was a bit more complicated than that.

Yah, the problem is really interesting to me. The question of how to filter specific sounds out seems to dictate the need to use an inverted waveform to me, just from a physics stand point, but I'd love to hear other ideas about how it might work. This might be worthy of another forum topic! I'm excited.

[Mark Alan Miller]

Noise reduction of the kind described here is not "cancelling" anything out.
In fact, the only way to cancel out a waveform by inverting it's polarity is by using the exact waveform on itself - that is, the snippet used as the noise sample, exactly aligned but polarity inverted with itself, would cancel.
As noise is random, there would be no way to invert this sample over any other area of the program material and have it do anything other than add more noise!

What these programs do is create a template of frequency vs amplitude of the noise sample, and then use that as a threshold for a multi-multi-multi-band expander, effectively. Often 1024 or 2048 individual bands of expansion. When you push these algorithms too hard, what you get is essentially different parts of the frequency spectrum opening up in unnatural and un-musically related ways, creating that 'mp3' sound... which, by the way, can be a cool effect on individual tracks, or even freaked-out remixes (I used it on the end of a remix of mine one time, for example.)

[cdbabel]

Noise reduction of the kind described here is not "cancelling" anything out.
In fact, the only way to cancel out a waveform by inverting it's polarity is by using the exact waveform on itself - that is, the snippet used as the noise sample, exactly aligned but polarity inverted with itself, would cancel.
As noise is random, there would be no way to invert this sample over any other area of the program material and have it do anything other than add more noise!

What these programs do is create a template of frequency vs amplitude of the noise sample, and then use that as a threshold for a multi-multi-multi-band expander, effectively. Often 1024 or 2048 individual bands of expansion. When you push these algorithms too hard, what you get is essentially different parts of the frequency spectrum opening up in unnatural and un-musically related ways, creating that 'mp3' sound... which, by the way, can be a cool effect on individual tracks, or even freaked-out remixes (I used it on the end of a remix of mine one time, for example.)

Wow, thanks for the explination! Quick question, I thought that when you inverted the polarity of a wave form, but then decreased its amplitude and then played it along with the origional, it would decrease the amplitude of the origional by the same factor. Isn't that the case, or am I swinging in the dark?

I see your point about noise being random though, even if that case above is true, it wouldn't work on noise.

The mp3 sound can work nicely on vocals, particuraly when you want to make fun of Cher. (Ever notice that Auto-Tune makes similar artifacts...ofcourse, Cher would never use AutoTune...never...)

[Mark Alan Miller]

Wow, thanks for the explination! Quick question, I thought that when you inverted the polarity of a wave form, but then decreased its amplitude and then played it along with the origional, it would decrease the amplitude of the origional by the same factor. Isn't that the case, or am I swinging in the dark?

I see your point about noise being random though, even if that case above is true, it wouldn't work on noise.

The mp3 sound can work nicely on vocals, particuraly when you want to make fun of Cher. (Ever notice that Auto-Tune makes similar artifacts...ofcourse, Cher would never use AutoTune...never...)

If you take two identical waveforms, invert the polarity of one and then sum them at unity gain, you get complete cancellation, as you probably know. Changing the level of one off of unity gain will make the cancellation increasingly less effective as the difference in level between the two is increased.

I liken "Auto-Tune abuse" more to a kind of glitchy vocoding-type sound, but I know what you're sayin'. And, despite "evidence" to the contrary, I maintain that while there is vocoding on the track "Believe" there is also "Auto-Tune abuse" on the tag lyric. But that has been hotly debated on this forum before, if I recall...

[cdbabel]

Wow, thanks for the explination! Quick question, I thought that when you inverted the polarity of a wave form, but then decreased its amplitude and then played it along with the origional, it would decrease the amplitude of the origional by the same factor. Isn't that the case, or am I swinging in the dark?

I see your point about noise being random though, even if that case above is true, it wouldn't work on noise.

The mp3 sound can work nicely on vocals, particuraly when you want to make fun of Cher. (Ever notice that Auto-Tune makes similar artifacts...ofcourse, Cher would never use AutoTune...never...)

If you take two identical waveforms, invert the polarity of one and then sum them at unity gain, you get complete cancellation, as you probably know. Changing the level of one off of unity gain will make the cancellation increasingly less effective as the difference in level between the two is increased.

I liken "Auto-Tune abuse" more to a kind of glitchy vocoding-type sound, but I know what you're sayin'. And, despite "evidence" to the contrary, I maintain that while there is vocoding on the track "Believe" there is also "Auto-Tune abuse" on the tag lyric. But that has been hotly debated on this forum before, if I recall...

Ha, hottly debating Cher on this forum...somehow thats hard for me to imagine.

[kayagum]

For $100 less than the pro version of Soundsoap (although the Soundsoap 2 is still the value champ), you can buy a complete version of Adobe Audition... the noise reduction plugin that comes with Audition is worth the $300 by itself.

I saved a whole CD project last year thanks to that 1 plugin. It is very effective, and there are many parameters you can tweak if you need to control the amount of reduction and preventing artifacts.

[JGriffin]

I liken "Auto-Tune abuse" more to a kind of glitchy vocoding-type sound, but I know what you're sayin'. And, despite "evidence" to the contrary, I maintain that while there is vocoding on the track "Believe" there is also "Auto-Tune abuse" on the tag lyric. But that has been hotly debated on this forum before, if I recall...

Ha, hottly debating Cher on this forum...somehow thats hard for me to imagine.[/quote]

No, it's true. I'm on Mark's side, I think the frozen-note sound on that track is actually a vocoder. Others will argue that it's an extreme Auto-Tune setting. Mark's probably right that there's both things on there.

[Mark Alan Miller]

No, it's true. I'm on Mark's side, I think the frozen-note sound on that track is actually a vocoder. Others will argue that it's an extreme Auto-Tune setting. Mark's probably right that there's both things on there.

Well, I think that the glitchy "do you believe in life after love" line is Auto-Tune, and then it's clearly followed by a vocoded line... is that what you mean?
I'm pretty confident in this, as I had been glitching vocals and things with my ATR-1 for over a year prior to that single and would recognise that particular sound anywhere.

Uh-oh. the debate re-emerges! Run for the hills!

Or at least run for being back on-topic.

[JGriffin]

Well, I think that the glitchy "do you believe in life after love" line is Auto-Tune, and then it's clearly followed by a vocoded line... is that what you mean?
I'm pretty confident in this, as I had been glitching vocals and things with my ATR-1 for over a year prior to that single and would recognise that particular sound anywhere.

Uh-oh. the debate re-emerges! Run for the hills!

Or at least run for being back on-topic.

I'd have to go back and listen to remind myself, and I gotta be honest, I'm in no rush. As I said, you're probably right.

[Mark Alan Miller]

and I gotta be honest, I'm in no rush.

Can't say I blame ya.

[Johnny B]

For $100 less than the pro version of Soundsoap (although the Soundsoap 2 is still the value champ), you can buy a complete version of Adobe Audition... the noise reduction plugin that comes with Audition is worth the $300 by itself.

I saved a whole CD project last year thanks to that 1 plugin. It is very effective, and there are many parameters you can tweak if you need to control the amount of reduction and preventing artifacts.

I agree 100%. I've done loads of cleanup with it. The stuff that's come out the best has been old cassette 4-track stuff that I transferred and remixed, because I could do small amounts on each track just to keep the tape noise buildup to a minimum. But I've done program cleanup as well, from both vinyl and cassette with excellent results. If you're trying to clean up really bad hissy cassettes, you'll hit problems, of course, but otherwise if you play around with it, you can get good results on a number of different "problem" areas (hiss, hum, crackles).