Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: VST which promises to fight the loudness war (Read 28647 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

VST which promises to fight the loudness war

Yesterday I took notice of Terry West's VST ReLife which claims to
Quote
bring back life to a heavy clipped audiofile - revolutionary way to recover transients and peaks.

It offers some adjustable options which are not explained in more detail, but it runs with foobar2000 with the help of George Yohng's VST wrapper.

Could somebody please provide a simple technical description how that VST exactly works, attended by an appraisal how this manipulations should be judged, first of all with regard to 'sound neutrality'? Any contribution would be highly appreciated.
This is HA. Not the Jerry Springer Show.

VST which promises to fight the loudness war

Reply #1
i can't give a technical description, but i can tell you that you can't recover clipped peaks, you can only make guesses.

VST which promises to fight the loudness war

Reply #2
Yesterday I took notice of Terry West's VST

(...)

Could somebody please provide a simple technical description how that VST exactly works, attended by an appraisal how this manipulations should be judged, first of all with regard to 'sound neutrality'? Any contribution would be highly appreciated.
Thank you for bringing the attention to this PlugIn! It really seems interesting - very similar to what iZotopes DeClipper offers. I really love the tool from iZotope, it can "restore" (don´t get this wrong, look below) around 60-70% of distorted peaks. If I got it right it does it so by reconstructing and interpolating. I don´t know exactly how it works, Alexey Lukin could give more detailed answers to that. But iZotope has one tiny disadvantage: it can only work with distorted peaks, not with tracks that were compressed without distortions. That´s the problem: newer releases don´t have distortions anymore. But hey, the DeClipper is not aimed at restoring botched mainstream releases.


i can't give a technical description, but i can tell you that you can't recover clipped peaks, you can only make guesses.
This is true. However, I tried it and compared it to a track that was declipped with iZotope. In preparation for this I created a distorted track myself by boosting the volume for +10 dB. I dithered it to 16 Bit and saved it in 16 Bit. Afterwards I declipped the track with the DeClipper and ReLife. Here are the results:

This is the undistorted original.


This is the declipped iZotope version.


This is the declipped ReLife version.

IMO the latter comes closer to the original. However, one can´t make conclusions by looking at spectograms. ReLife works with clipped audio material and with unclipped but dynamically compressed audio material. I don´t know how it does this, upon hearing the results I didn´t notice any strange pumping effects. It even works without being specified where to look for compressed dynamics / distorted peaks. And this is strange... how does it do this?

Maybe a person with more technical knowledge could shed some light into this.

EDIT: this is a strange plugin. I can´t figure out even remotely how it does what it does. I threw everything at it, Pop, Rock, Classical music, Electronic music... the results are simply stunning. I even tried material that is non-distorted and has a perfectly unprocessed dynamic. It does something to that also but far far less than to other material. Then it can even work with already processed material i.e. material that is lossy compressed, resampled, treated with reverb, anything. And it does so without a point of reference like the DeClipper from iZotope need (which also can´t work with processed material). Really, really strange.
marlene-d.blogspot.com


VST which promises to fight the loudness war

Reply #4
It really seems interesting - very similar to what iZotopes DeClipper offers. I really love the tool from iZotope, it can "restore" (don´t get this wrong, look below) around 60-70% of distorted peaks. If I got it right it does it so by reconstructing and interpolating.

You are right: iZotope Declipper throws away sections of the waveform above the user-specified level threshold and replaces them with the waveform interpolated from the surrounding material. It doesn't change the rest of the audio signal. Here's the graphical example of its work when the threshold is set to just below -6 dB FS (before and after):



It is not supposed to be used on non-clipped material, such as the output of a limiter, because this material can be "restored" much better by not throwing "waveform tops" away, but rather applying some gain envelope - a kind of "upward expansion" (apparently this is what ReLife does). However I've heard reports the people are still using Declipper that way and being happy with the results (maybe because the interpolation algorithm is pretty advanced).

VST which promises to fight the loudness war

Reply #5
Thank you for all the answers.

Cavaille:

Quote
I don´t know how it does this, upon hearing the results I didn´t notice any strange pumping effects.
...
I threw everything at it, Pop, Rock, Classical music, Electronic music... the results are simply stunning

Did your audio material include solo piano music (ideally both classical and Jazz) and are your statements cited here valid for this pieces of music also?
This is HA. Not the Jerry Springer Show.

VST which promises to fight the loudness war

Reply #6
It is not supposed to be used on non-clipped material, such as the output of a limiter, because this material can be "restored" much better by not throwing "waveform tops" away, but rather applying some gain envelope - a kind of "upward expansion" (apparently this is what ReLife does).
I´d think so too. It would be the only explanation I could come up with (given my limited technical knowledge and background). But the developer says his ReLife doesn´t use an expander. How could it be any different? He claims, he "discovered" it by accident when tinkering around with one of his EQ.
However I've heard reports the people are still using Declipper that way and being happy with the results (maybe because the interpolation algorithm is pretty advanced).
Many people are happy because sometimes the results on material you described are simply breathtaking. Though the outcome in such cases tends to be unreliable.
Did your audio material include solo piano music (ideally both classical and Jazz) and are your statements cited here valid for this pieces of music also?
No, it didn´t. The classical music was something with lots of brass, strings and percussion. The pop music was some stuff recent stuff from Madonna.
marlene-d.blogspot.com

VST which promises to fight the loudness war

Reply #7
I tried this on One Little Victory from Rush's Vapor Trails album. It doesn't make it a good recording but it seems takes the edge off and makes it less awful.

VST which promises to fight the loudness war

Reply #8
Vapour Trails does sound quite awful, but I wonder if it's out-fizzed by Number One Sons's Lessons (disclaimer: which I bought a long time ago, never liked much, and got rid of when I realised how terribly mastered it was ).

VST which promises to fight the loudness war

Reply #9
Thanks to all who contributed to this thread. I have learned that ReLife makes a good job from a technical point of view, but its useability depends on how well its processing algorithms meet the individual requirements of the source file; audible sound discolorations may occur. So I will decide occasionally whether to use this VST or to count my blessings.

Cavaille: the sites soundberg linked to above contain some useful information about ReLife. So if you want to keep the VST they may be worth reading for you too.
This is HA. Not the Jerry Springer Show.

VST which promises to fight the loudness war

Reply #10
I tried a brief ABX. Actually I didn't hear much of a difference. With a pop-track, the beats sounds slightly more soft (which I suppose are to be expected - Mastering Engineers are using compression to add "punch" as I understand it). I didn't really notice any coloration, but I may just not know what I should listen for.

I think I will use this as a DSP in my setup for a while. Hopefully it may even help reduce fatigue when I listen with headphones?
Can't wait for a HD-AAC encoder :P

VST which promises to fight the loudness war

Reply #11
Does this plugin work on replaygain-processed(precut) materials as well??

I wonder how does this plugin correctly 'estimate' the dynamics without the reference threshold level set.
Certainly a voodoo-magic! (Or is that what the upward expansion originally does?)

VST which promises to fight the loudness war

Reply #12
Does this plugin work on replaygain-processed(precut) materials as well??

Just tried that yesterday. It works just as well on processed material.
Can't wait for a HD-AAC encoder :P

VST which promises to fight the loudness war

Reply #13
I´ve tried it on several releases that could be described as a victim of the loudness war. This VST is able to work on every one of them - with differing results. On some albums it "restored" very much peaks, on others it restored less. It seems that it treats a lot of compression differently to less compression.

I have two releases of Madonna´s "Hard Candy": one on CD, one on Vinyl. The vinyl version is only a bit compressed with lots of dynamic peaks left while the CD version is a typical example of a crushed-to-death release. This VST restores the CD to the same kind of dynamic the vinyl version shows. They look very similar after treatment. I´ve also had superior results with a score I own: "Lara Croft: The Cradle of Life" from Alan Silvestri. This score has a lot of percussion samples in combination with choir and orchestra. When replaygained, it is approximately 6 dB too loud with dynamic peaks simply cut off. The DeClipper from iZotope can´t find distortions but the ReLife VST restored every peak. I first suspected that it only concentrates on lower frequencies but after this score I found that it seems to treat the whole frequency band. The results are very nice.

When one uses the third option inside the VST, it lowers the volume by precisely 4.21 dB - maybe for having enough headroom. This - 4.21 dB are very curious: why would anyone choose such a strange value? Furthermore, if I got it right the developer says that it doesn´t lower the volume. But maybe I read this wrong. I tried to find out if it changes the frequency response. Because the developer said he accidently found it by playing around with one of his self-programmed EQ. But it doesn´t: iZotope Ozone´s Matching EQ shows a flat line. Sometimes the tool changes the offset of certain peaks, moving them closer to the center without affecting the overall offset. It works with 32-Bit-Floating point and doesn´t seem to use any kind of filtering. But I´m not sure about that. The "Preclean" option should be avoided at any cost - it changes the frequency response a lot, I don´t know why this option even exists.

So far, this tool appears to be very valuable: I´m getting around 2-3 dB of peaks back on average, sometimes even up to 6 dB.

marlene-d.blogspot.com

VST which promises to fight the loudness war

Reply #14
Yesterday I took notice of Terry West's VST ReLife which claims to
Quote
bring back life to a heavy clipped audiofile - revolutionary way to recover transients and peaks.

It offers some adjustable options which are not explained in more detail,
[...]
Could somebody please provide a simple technical description how that VST exactly works

I tried it and it appears to be little more than just a high-pass filter.

VST which promises to fight the loudness war

Reply #15
I tried it and it appears to be little more than just a high-pass filter.
I´m sorry Greynol, but I disagree. I tested with the Matching EQ of iZotope Ozone if there are frequency differences between a "ReLived" file and the normal, replaygained versoin. I didn´t find any.

I suspect that you took sounds which were heavily distorted with real 0 dBFS clippings. If these distortions are removed it could sound like a high-pass filter was applied. I´ve had this with "Californication" from RHCP which I declipped with iZotope - after the clipping was removed and the lost peaks resynthesized it had measurably less high frequencies. Simply because the Matching EQ can´t decide whether the high frequencies consists of high frequency noise or pure high frequencies. Or did you use the "Preclean" option? Because this seems to apply a high-pass filter.

According to the "developer" of ReLife, it has problems with distortions (clipped signals) but works very good with everything else. I support this view. It works very good with brickwall limited material that hasn´t any distortions.
marlene-d.blogspot.com

VST which promises to fight the loudness war

Reply #16
I tested with the Matching EQ of iZotope Ozone if there are frequency differences between a "ReLived" file and the normal, replaygained versoin. I didn´t find any.
Do you know what a square wave looks like when it's sent through a high pass filter with an extremely low cut-off frequency?

I suspect that you took sounds which were heavily distorted with real 0 dBFS clippings. If these distortions are removed it could sound like a high-pass filter was applied.
This had nothing to do with what anything sounded like.  DC blocking doesn't usually make an audible difference. 

I´ve had this with "Californication" from RHCP which I declipped with iZotope
That's funny.  It was Californication that I used with both iZotope's de-clipper as well as the ReLife plugin (which is actual topic of this conversation).  ReLife is clearly performing DC blocking (read: high-pass) whereas iZotope is doing something different.

Or did you use the "Preclean" option?
It was not lit when foobar2000 did the conversion.  Perhaps it is using "preclean" anyway?

Feel free to search my posts on this topic, once concerning the vinyl version of Metallica's Death Magnetic and once concerning Pearl Jam's Vs.  Both of these discussions took place with Axon, with whom I trust implicitly with regards to this sort of thing, and you will see that he agrees with me.  Would you like me to show you some plots?  They will look just like what I provided in the Death Magnetic discussion: clearly a high-pass response.

If it turns out that preclean is in fact not being used then maybe you aren't interpreting your measurements correctly.

VST which promises to fight the loudness war

Reply #17
I did a quick test by running an IR through Relife:

(Red: right / Left: blue)

Algothm was set @ 3 & no preclean applied.
Also I noticed some phase skew as well:
http://j.imagehost.org/0026/relifez.png

VST which promises to fight the loudness war

Reply #18
Greynol hit the nail square on the head. When you apply the effect to a 1kHz sine, Type 3 "peak recovery", for example, is basically equivalent to a high-pass with 25 Hz cutoff.

The author is a funny guy, to write a whole plug-in and GUI around a simple finding like that. But he is already being worshiped with pictures like this:



I'm just wondering, though, why he had found this effect already built into his old double compressor plugin, as he wrote.

VST which promises to fight the loudness war

Reply #19
I tested with the Matching EQ of iZotope Ozone if there are frequency differences between a "ReLived" file and the normal, replaygained versoin. I didn´t find any.
Do you know what a square wave looks like when it's sent through a high pass filter with an extremely low cut-off frequency?
You know what? I´m sorry, you´re right: I confused high-pass filtering with low-pass filtering - again. Considering that, all of your following sentences make sense. I´m really sorry.

Feel free to search my posts on this topic, once concerning the vinyl version of Metallica's Death Magnetic and once concerning Pearl Jam's Vs.  Both of these discussions took place with Axon, with whom I trust implicitly with regards to this sort of thing, and you will see that he agrees with me.  Would you like me to show you some plots?  They will look just like what I provided in the Death Magnetic discussion: clearly a high-pass response.
No, I´m perfectly fine now 

I did a quick test by running an IR through Relife:

(Red: right / Left: blue)

Algothm was set @ 3 & no preclean applied.
Also I noticed some phase skew as well:
http://j.imagehost.org/0026/relifez.png
Fascinating. Can one correct this? And please explain - if you like - to a noob like me why phase skew is bad for audio.

EDIT: I just did a quick test with iZotope Ozone - I used the EQ (with the option "Analog") and applied a high-pass filtering at various frequencies (20, 25, 30, 35): Voilà, ReLife! Then also the phase skew makes sense (though I still don´t understand it fully) because iZotope points out that the analog EQ will introduce phase distortions.
marlene-d.blogspot.com

VST which promises to fight the loudness war

Reply #20
Now, who wants to be the party pooper and  the  &  at kvraudio?

VST which promises to fight the loudness war

Reply #21
Now, who wants to be the party pooper and  the  &  at kvraudio?
I´d love to do it since I fell for it completely - but I lack some basic technical knowledge to back it up properly. So I´m not the right person.

Guys, did I mention that you are great?
marlene-d.blogspot.com

VST which promises to fight the loudness war

Reply #22
It's funny that it's as simple as that - I never cared to verify.
Theoretically, pretty much any kind of processing (phase shift, EQ, dynamic processing) applied to a brickwall-limited signal will change its peak levels and make it look better. But such a filtering would not restore any audible dynamics, unless it's a true expander.


VST which promises to fight the loudness war

Reply #24
(bought popcorn, I have Jaegermeister already, waiting for humorous quotes here...) 
People, this thread is the reason why I joined HA in the first place... trying to figure out how audio stuff works is something I really like.





Bringing audiophools to their knees is just sort of a bonus.
Error 404; signature server not available.

 
SimplePortal 1.0.0 RC1 © 2008-2021