Premium Membership - All Voxengo Plugins at a Fixed One-Time Fee - Click Here For More Info
Summer 2018 Discount at Voxengo
Forums     Discussions     Announces, Releases and Discussions Voxengo Redunoise - a new PC noise reduction system (alpha)





After a month of development, Redunoise enters the beta stage.

You can download v1.0f beta here:

The package contains everything a release version will have.  If you are interested in an in-depth knowledge you should read the HTML help file, it contains almost everything you may need to know.

For everyone else I suggest to try Redunoise out for its Makeup equalizer since it sounds uniquely, and to me it's simply a golden equalizer (sorry for some hype).

I've also updated MP3 demos so you can evaluate the NR algorithm right away:




I've spent many, many hours working with all sorts of noise reduction systems and the thing that I immediately miss in your new one is a 'difference' button to let me hear precisely what is being removed.  Without this, everything is guesswork.  Any chance of adding it?

Why does the frequency response run from just under 1KHz to 20kHz?  Can the plug-in not be used for low-frequency noises?

Thanks for a very interesting and new approach to noise reduction.

- Peter

A few comments -- excellent sound quality (you've seen my comments on KVR): I agree with Peter that it would be nice to have a "difference" button; Why do you show frequency on a linear rather than logarithmic scale -- there'd be less need for zooming if it was the latter?; A "control point" (perhaps just outside spectrum plot to the side of each curve which can be moved up and down to move the *whole* curve -- perhaps all curves could be moved without needing to select them first using these "control points"); The ability to capture the noise spectrum and get the threshold curves to match it -- this would save loads of time!; The ability to copy the shape of one curve to another.

These are mostly UI enhacements that would make the program which, whatever you might think, is pretty complex :-) a little easier/quicker to set up.  The manual also needs more explanation as to how best to set up the curves -- I imagine a lot of people would be completely lost trying to figure out how to set up this program to achieve best results.

Peter, thanks for pointing this out.  Because of the new approach I've used it is simply not an obvious thing to implement a residual output option.  The other thing why possibly I have not devoted much thought to this option is that it is not totally useful in my opinion (I've spent much time with NR in the past, too).  Since there is a bypass button, difference can be immediately heard.  OK, if I can figure out how to implement this option, it will be available right away.

You can also zoom-in the control surface using the ALT key.

I really don't see any real problem in having linear scale only.  But I think I'll add the 'log' option in the next version.

This is true that this program transformed into a pretty complex on, but otherwise it's quality was not any different from the existing NR programs.

In the version I'm going to release soon you can select all control points with the right mouse button click, so that there are no problems with moving the whole curve.

'Noise print' capture is on my 'todo' list.

Thanks for your comments!

Thanks for the response.  I really do miss the 'difference' function, because it lets you hear immediately if you're removing any of the music signal as well as the noise, so if this can be added as you suggest it would be a bonus.

I agree with Andrew's comments on usability and learning.  I got very confused over which parameter I was adjusting, especially when setting the threshold levels (and I still don't understand fully the idea behind them).

- Peter

Some comments on the usability of the beta.  All are made positively with the aim of making it better.  If they read like criticism, that is a mis-reading.

1) Terminology.  The Help uses the terms 'program' and 'bank' without defining what thay mean.

2) The y-axis.  The key to understanding how to control the plug-in is to notice that the y-axis changes to correspond to the parameter being adjusted.  Thus, when setting Attack and Release, the y-axis shows Seconds; when setting other values it shows in dB; on History it shows the length of the history buffer.

3) Calibration of the y-axis in some cases is bizarre.  On History, for example, the tick marks are approx. every 1.2 (!); on the Threshold values it's approx. every 10.7dB; and so on.

4) Same is true of the x-axis.

5) Range of the y-axis.  Threshold seems to assume we will pass it signals that can peak up to +54dB (!!).  Reduction seems to allow values up to 95.8dB of reduction (pretty quiet, the result should be!) and is calibrated with - values whereas they should be +.  We reduce by +10dB, thus making it quieter; if we reduced by -10dB as shown in the beta we would be making it louder.  All axis ranges should be re-thought to display 'sensible' ranges for the function.

6) Purpose.  It seems that one could set multiple attack, release, history, etc values based on frequency.  What for?  This looks like a function of the program that is looking for a reason to exist.  I'd suggest these values would be better set on a single control knob , unless Alexey can explain why one would want, say, a history value of seven from 0-1kHz and then sloping to a value of 12 at 20kHz... and the user can determine the best setting to use.

7) Please allow us to read the current value of a control point by hovering the mouse over the point.

8 ) I am still getting very confused as to which parameter I'm changing.  I hate this 'blue line' that represents the active parameter, with some non-active ones shown in red, and others not at all.  How about different colours, at least?

9) The Reset button with no Undo available is a guaranteed way to make the user kick the computer, cat, or any other nearby animate object. (Without the ability to Undo I suggest it be renamed the Ohshit button.)

10) Using CurveEQ, the Monitor button, when clicked, brings up a little menu of the possible choices.  In Redunoise, the equivalent button has no legend but simply toggles through the available options.  Similarly in CurveEQ there is a gain control slider: in Redunoise there are two control knobs labelled IN and OUT (why two?) that seem to control overall gain.  Slider?  Knobs?  The same function in two of your plug-ins should be controlled the same way.  Please be consistent in your UI!

Well, that's ten ideas for improvement.  Enough for the moment, I think.

- Peter Harrison

11) Please, please, please - a Difference button!

Thanks for ideas and suggestions.  I'll consider some of these.

However, from what I've read I really think that you should simply experiment a little more, and get accustomed to Redunoise.  Sorry if this sounds forceful.

BTW, *all* parameters available are actually relative values.  It is simply impossible to define attack and release in seconds just because processing is performed in blocks.  Thus both Attack and Release parameters are relative and in any case you have to learn how they affect the process.

Spectrum analyzer is tuned in a way so that it can reach 60 dB mark, and again this is relative.  Of course, it is possible to shift it to 0 dB, but this is not totally useful, especially when you *visually* define the Threshold parameter curve.

ALL parameters are defined in dB, only the History parameter has a 'titleless' numeric definition.  I will include a note about this in the manual.

What kind of colours can you offer?

Aleksey, thanks for the response.  Now I'm confused.

About Attack and Release, the Help says:

"These controls basically work like the attack and the release of some compressor.  So, if you are familiar with compressors, understanding these controls should not cause any problem to you."

Yes, I am familiar with many types of compressor, both hardware and software.  In all of them, I think, Attack and Release values are time-based.  I know how to use those values to get the results I want.

But in your reply to my earlier message you wrote:

"ALL parameters are defined in dB..."

and again:

"*all* parameters available are actually relative values.  It is simply impossible to define attack and release in seconds..."

So do I deduce correctly that in Redunoise the Attack and Release values are calibrated in RELATIVE DB?  If so, then these parameters are behaving like no other compressor that I have ever encountered and I don't know what the settings mean or what they control.

Please can you help with guidance on whether I have understood this correctly, and what values would typically be used.  And why would the user want to have (as I asked earlier) different values of attack and release at different frequencies?

Thanks for your help.

- Pete

Yes, I understand your confusion.  I possibly have to find a way to clarify this a bit.

Typical values would be -0.5 dB for Attack and -1 dB for Release.

The meaning of these figures is pretty straightforward: with -0.5 dB attack internal filter reach the necessary reduction value later, because its added difference is being damped by -0.5 dB every step (not seconds here).  The release works in the same manner: filters reach 'no change' position later because their added difference is being damped by -1 dB every step.

Frequency-dependent definition is actually a very useful option for every parameter.  E.g. if `swishes' go out of control in some frequency range only, you can increase the Attack and the Release in that range, only.

This topic was last updated 180 days ago, and thus it was archived.  Replying is disabled for this topic.