Waves Restoration vs. Izotope RX

[G-lay]

Hey guys & gals,

Any opinions on these two products?
I've used the Izotope RX, and I was very impressed, but maybe there are some of you that worked with both, and migt be able to share some advice, and opinions?

cheers,

G-Lay

[eoats]

I find I get much better results with the iZotope Denoiser than with Wave's x-Noise (Haven'y used the waves Z-Noise).

The extremely good Declipper which is for me far & away the best thing for clearing distortion I've ever seen used. I never got the Waves to achive anything close.

[Rick Sanchez]

Originally Posted by eoats

I find I get much better results with the iZotope Denoiser than with Wave's x-Noise (Haven'y used the waves Z-Noise).

The extremely good Declipper which is for me far & away the best thing for clearing distortion I've ever seen used. I never got the Waves to achive anything close.

+1 for Izotope. A great tool and always improving it.

[danijel]

For broadband noise, check out this and this.

If I had to go for one or the other, I'd go for Waves, because X-Crackle is irreplaceable to me for turning lav rustle into smooth "cloth movement", while Z-Noise is very close to RX denoiser. Then, there is the spectral repair which is unique in RX thus far.

[Lipflap]

Whereas I often use iZotope for broadband denoise (along with Waves Noise Suppressor), I find X-Crackle and X-Click unbeatable for radio mic troubles, clicks and the like.
With iZotope Spectral Repair, you can perform miracles, if you have the time.

Personally, I like having both iZotope RX and the Waves Restoration Bundle, since they work in somewhat different manners and there are times when one or the other tool, or a combination, is just what I need.

Full disclosure: I occasionally write manuals, training courses, etc. for Waves.

[airbornesound]

I am a big fan of iZotope, and also Z-Noise

The de-clipper is stunning. Decrackle and dehiss are also good. Spectral repair and hum removal I found were OK.

Waves X-Noise I found to be pretty weak. The rest of the restoration package I found lacking as well -- I found it just couldn't work as aggressively as the iZotope package.

Waves Z-Noise I find is superb. Better than the iZotope DeHiss in my opinion, although the DeHiss is still good. Z-Noise I find is far superior to X-Noise.

[quadraphonics]

Danijel,

I looked at the links that you posted and I have a general concern with the overall technique applied to get the results. It sounds like the person doing the restoration was trying to take broadband noise and hums out in one pass. This is usually done in multiple passes with an EQ before the broadband noise reduction.

Personally, I have used the Cedar DNS2000, Sonic NoNoise , WaveArts and iZotope for broadband noise reduction.

The Cedar is my favorite, but needs a very gentle hand. The Sonic I feel can be pushed harder than most of the software based systems before it aretfacts. The iZotope is nice, but I find that I get artefacting when it handles more than 3-4dB of noise reduction at a time.

Randall

Originally Posted by danijel

For broadband noise, check out this and this.

If I had to go for one or the other, I'd go for Waves, because X-Crackle is irreplaceable to me for turning lav rustle into smooth "cloth movement", while Z-Noise is very close to RX denoiser. Then, there is the spectral repair which is unique in RX thus far.

[Lipflap]

Originally Posted by quadraphonics

It sounds like the person doing the restoration was trying to take broadband noise and hums out in one pass. This is usually done in multiple passes with an EQ before the broadband noise reduction.

Yes. During a normal restoration job, one should first deal with harmonic noises (with some flavor of multi-band EQ, after figuring out the harmonic sequence with an FFT or equivalent), then deal with transients that can be removed with interpolation processors (all the processors that end in "-click" and "-crackle"), then do broadband processing, and finally perhaps a tiny touch of EQ to deal with residual harmonics. And, as when you paint a window, several light passes beat the hell out of one killer pass.

My all-time favorite restoration package is still the old Sonic Solutions USP NoNOISE system (discontinued in about 1998). You need an advanced degree in physics to figure it out, but once mastered, you can get remarkable results.

[danijel]

Originally Posted by quadraphonics

=
I looked at the links that you posted and I have a general concern with the overall technique applied to get the results.

Aleksey's methods are scientific - his goal was to reduce noise by 12dB in one pass AFAIR (although I don't understand how he measured that), not to make the best NR for a specific line of dialog in a specific scene. So, take it just for what it is - not a definitive test to rule them all Real-life comparisons would depend on many other factors, including the operator skill and other sounds in the scene - later in the thread, Branko makes a quick multi-pass notch+Cedar attempt..... that's probably as good as it gets - and I couldn't really say iZotope's one-pass was any worse - only different.... (IMHO, of course)

[Alexey Lukin]

I agree that my test is quite narrow. My primary goal was actually to demonstrate possible weaknesses of noise reduction systems.

If there's any interest in a more complete and thorough testing — I'd be willing to participate and help with samples from several algorithms.

[Lipflap]

Originally Posted by Alexey Lukin

I agree that my test is quite narrow. My primary goal was actually to demonstrate possible weaknesses of noise reduction systems.

From my experience, the primary weakness in any noise reduction process or processor is the operator. We easily lull ourselves into unrealistic expectations, or - for all sorts of legitimate reasons - get impatient and cut corners. Given the time (which is often not ours to give), you can get pretty good results from any number of systems.

[akinla]

Just began using izotope rx in tandem with my old school analogue noise reduction tools, and have found it very useful, it helped some nasty production sound on "Dinner for Schmucks" and "Burlesque" Andy Koyama, Sony Pictures Entertainment

[ggegan]

Originally Posted by akinla

Just began using izotope rx in tandem with my old school analogue noise reduction tools, and have found it very useful, it helped some nasty production sound on "Dinner for Schmucks" and "Burlesque" Andy Koyama, Sony Pictures Entertainment

Hi Andy, welcome to the forum! Good to see another veteran here.

Even though I primarily mix FX, RX is a very useful tool for me. I have been playing with it in order to reduce the background noise of some of my field recordings. It works pretty well, but requires a very gentle touch and reasonable expectations. It isn't a cure all for getting rid of traffic roar, but it certainly helps a lot. I also like the McDSP NF575 noise filter a lot, but it doesn't help broad band noise much except as a high pass to get rid of rumble.

[evanb]

For those of you who talk about the stunning declipping, can you elaborate? I have very little luck with it.

And even without declipping, I think RX is far superior to Waves.

And one more question...is the advanced RX worth the extra &??

-Evan Benjamin

[ggegan]

Originally Posted by evanb

And one more question...is the advanced RX worth the extra &??

I was wondering the same thing. 3X the cost. I know it has dithering and sample rate conversion, both of which Pro Tools can handle, but I also noticed that it had some extra goodies in the Denoiser module that supposedly reduce artifacts. If it can significantly improve the creation of artifacts associated with denoising, then that would be a big plus. Has anyone determined that it is that much better?

[pmx]

the dither and src technology is licensed to AE's Sample Manager, so if you want that pick it up. it saves you a lot of money.

[Jfriah]

Neither is "superior". My experience has been: you use what you have (gentle passes are better than 'going for the fence'). If you have multiple tools (ie: I flip between iZo and Waves X-Noise), you try them both. Sometimes you just get a 'hunch' which will work best for _______ based on experience.

My experience (and I'm constantly using the tools...believe me) indicates that, for me, iZo is worth its weight in gold to me and I use it about 80-20 compared to X-Noise.

But there are times when I 'need' to go to X-Noise (or my CEDAR DNS-2000) because iZo isn't cutting it.

Some noise signatures are different...

Good luck.
Jeff

[muzeon]

Izo RX, Denoiser better than X-Noise, IMHO. Like using de-Crackle though, and the WNS is pretty cool too, for a fraction of the cost of Cedar. Yes, be careful not to sit on it too much, may end up taking 'life' out of the track. Sometimes use RX or X-noise as sound design tools... unexpected fun. Good luck with your research.

[Player1]

Originally Posted by Lipflap

My all-time favorite restoration package is still the old Sonic Solutions USP NoNOISE system (discontinued in about 1998). You need an advanced degree in physics to figure it out, but once mastered, you can get remarkable results.

Agreed, I used Sonic NoNoise for about 7 years and while I've had Waves restore and Z noise for years I've never been able to get the same results.
I also like the PQ CD writing program, I wish Pro Tools had something like it.

[pechnatunk]

Originally Posted by Lipflap

Yes. During a normal restoration job, one should first deal with harmonic noises (with some flavor of multi-band EQ, after figuring out the harmonic sequence with an FFT or equivalent), then deal with transients that can be removed with interpolation processors (all the processors that end in "-click" and "-crackle"), then do broadband processing, and finally perhaps a tiny touch of EQ to deal with residual harmonics. And, as when you paint a window, several light passes beat the hell out of one killer pass.

This is probably THE most important thing I have read on GS.


I'm pretty sure I know the answer to this - but just to double check.
With all this talk of multiple passes - am I right in thinking that with each pass you re-train the denoiser?

[Lipflap]

Originally Posted by pechnatunk

I'm pretty sure I know the answer to this - but just to double check.
With all this talk of multiple passes - am I right in thinking that with each pass you re-train the denoiser?

If you're using a multi-pass, multi-pprocessor technique like this, you won't touch the broadband processor until after you've dealt with harmonic noise, using some sort of a multi-band EQ or "humm filter." If that's the case, you don't need to retrain the broadband device, since you've not yet used it. Take a noise sample after EQ processing. However, if you do some sort of EQ after broadband cleaning, you'll need to grab a new noise sample if you plan another pass of broadband.

Also, if you rely on some sort of graphic display to figure out where lie the problems (FFT, histogram, etc.), you'll need to update the picture after each processing step.
Good luck

[ggegan]

I've been experimenting with a technique in RX Denoiser where I divide the spectrum up prior to sampling the noise. For example, when dealing with traffic rumble where the most important audio that you want to keep is above 100Hz, like voices, I do something like this:

1) Apply a gently sloped high pass filter below 100Hz
2) Select all frequencies above about 150-200Hz and remove them using the Gain module at -inf
3) Sample and train Denoiser with the noise that is left
4) Go back to the pass prior to the Gain pass, recovering the removed frequencies and apply a light Denoiser pass or two. (I use the envelope to ramp the processing into the lower frequencies)
5) Do the same process for the high frequencies above the speech range.
6) After the top and bottom have been cleaned up a bit, sample noise from the entire frequency spectrum and do some light passes of processing, resampling the noise after each pass.
7) If there are any short stray artifacts, I try to remove them using the Spectral Repair module, selecting only the areas and frequencies where the artifact lives.

I'm still trying to perfect the process, but it seems to be working pretty well. It sounds pretty natural and seems to reduce the creation of artifacts.

The only tool I wish RX had is a copy and paste that can be restricted by frequency. For instance, if after all the processing, there are some artifacts in the lower frequencies from a truck by, I wish I could copy some audio in that frequency range from somewhere else and paste it over the problem area without affecting any frequencies above or below. Spectral Repair does this to a degree, but the 4 second restriction is a pain and I would like to be able to copy the audio from a non adjacent area.

[Jfriah]

As to the Advanced vs Normal Rx-- perhaps MATTI will read this and chime in; I just have the regular version and Matti and I have had shootouts and I give the cup to him with the Advanced DeNoiser.

True, Gary, it has a few features found in other places (but it is also marketed as a standalone, so that's not surprising).

And I've definitely heard a difference when using A mode vs C (offline) mode on iZo (C being better but taking wayyyyyyyyyy longer to process).

Jeff
(and this all reminds me I'm going to start a new thread today)