INTERACT FORUM

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1]   Go Down

Author Topic: A wild idea to combat dynamic range compression  (Read 3024 times)

InflatableMouse

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3978
A wild idea to combat dynamic range compression
« on: September 30, 2014, 10:56:18 am »

I've had this idea but I don't know if its possible ... I figured I'd throw it out there and see what others think of this.

I suppose many of us know about the loudness war ... music is being dynamically compressed to sound louder or fuller. Especially remasters of older music suffer from this. Many people don't like this and Online streaming services often serve dynamically compressed versions of songs that have much better dynamic range. If you want to know more about it, search for "loudnes war" in google.

Oke, what I was thinking is would it be possible to create a fingerprint of the dynamics of a song during audio analysis (optional choice, tickbox), upload this fingerprint to an online database and later apply it to a version of that song with less dynamic range to restore it?

Obvously there are challenges and I don't have all the answers, I just wondered about this idea and I thought it would be neat if its possible!



Logged

Matt

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 42053
  • Shoes gone again!
Re: A wild idea to combat dynamic range compression
« Reply #1 on: September 30, 2014, 11:06:57 am »

Oke, what I was thinking is would it be possible to create a fingerprint of the dynamics of a song during audio analysis (optional choice, tickbox), upload this fingerprint to an online database and later apply it to a version of that song with less dynamic range to restore it?

So what's happening exactly?  Is it reading the dynamic range from one file and applying it to another file? 

I think the nature of the sound is already lost if the range is overcompressed.  I think turning it down wouldn't be the same as restoring it.
Logged
Matt Ashland, JRiver Media Center

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5181
  • "Linux Merit Badge" Recipient
Re: A wild idea to combat dynamic range compression
« Reply #2 on: September 30, 2014, 11:14:09 am »

So what's happening exactly?  Is it reading the dynamic range from one file and applying it to another file?  

I think the nature of the sound is already lost if the range is overcompressed.  I think turning it down wouldn't be the same as restoring it.

I think he's imagining you could create a "finger print" from an old recording with a DR of 13 that somehow captures necessary and relevant information about it's dynamic range.  You'd then somehow use that fingerprint with a newer DR 9 version of the same song to recreate the original dynamic range.  In it's grossest form it would be kind of like making a diff of the two files.  

It couldn't be the exact same as a copy of the original unless the fingerprint was really a diff of the two files (which would not be drastically smaller than an entire file, especially if there are other factors contributing to the difference between them (different masters, etc)).  

My guess is that it would be space-intensive and would require access to the original files, at which point, why not listen to the originals?  Maybe to "rehabilitate" streaming audio on the fly?
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: A wild idea to combat dynamic range compression
« Reply #3 on: September 30, 2014, 11:20:48 am »

I don't think there is much to be done for dynamic range compression really.
If it is compressed to the point of clipping, there are tools which can declip and make the files a lot more listenable though.
 
Perfect Declipper and iZotope RX are two tools which can do a good job.
 
http://www.youtube.com/watch?v=m7s3rLdWrQ4
 
I'm not sure about Perfect Declipper, but iZotope RX works well until you enable Volume Leveling in Media Center, since it works on a fixed level (e.g. the top 3dB) rather than dynamic, and Media Center does not let you apply declipping before leveling is applied.
 
Leveling often reduces the volume of compressed tracks quite a lot, which moves the peaks in the track below the range that the declipper is active.
Of course you can do offline declipping by creating processed files, but then you have duplicates to deal with.
 
 
At one point Matt seemed interested in this, but I don't think I provided very compelling samples, so I don't think anything ever came of it.
Logged

InflatableMouse

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3978
Re: A wild idea to combat dynamic range compression
« Reply #4 on: September 30, 2014, 11:53:51 am »

I'm not talking about declipping or a diff.

What I'm imagining is to "sample" the gain or strength of a signal throughout the song and apply it to another.

Songs for which this matter almost always have "identifiable chunks". What I mean is, when you look at that waveform zoomed out, you can see the quiet parts and loud parts. In a dynamically compressed version, these "chunks" are similar in loudness. Now, in my imagination when I would analyze that song with good DR, the algorithm would identify these chunks, calculate their loudness and store it with a time index and length, etc. If I then wanted to apply that to a low DR version, it would search the song, identify these chunks and apply the previously recorded gain.

I hope I explained it well, sorry if I'm using incorrect terminology.
Logged

InflatableMouse

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3978
Re: A wild idea to combat dynamic range compression
« Reply #5 on: September 30, 2014, 11:59:31 am »

To elaborate, I understand what is lost due to compression is not coming back. If they made it so loud its clipping, I understand.

But as long as that's not the case, I would think at least the dynamics could be restored if you knew what it was (which I tried to explain in my previous post).

I'm thinking that it might even have another beneficial effect; there are remasters out there that are actually better sounding, more detail, better tonal balance, better soundstage but are dynamically inferior to an older version. If my idea would work, you could have best of both worlds, increased dynamics from the old song with the better tonal balance, soundstage etc from the new remaster  ;D
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: A wild idea to combat dynamic range compression
« Reply #6 on: September 30, 2014, 12:33:48 pm »

I feel like it would require so much data to do this well, that you may as well be replacing the track.
 
It seems that Perfect Declipper now also performs dynamics restoration, not just declipping, as I downloaded the latest version to give it another try. I have a handful of tracks that I'd like to test it with to see how well it works now.
Logged

InflatableMouse

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3978
Re: A wild idea to combat dynamic range compression
« Reply #7 on: September 30, 2014, 12:34:51 pm »

And how would it know what to restore it to if it doesn't know what it was in the first place?

Edit:

A few seconds of what I mean illustrated in the attached image. As you can see within the yellow lines would be the chunks, the peaks in the lower song match the peaks in the upper song. If the upper song's chunks would be restored to the level of that of the song below it, wouldn't that effectively restore its dynamic range?
Logged

-JK-

  • Junior Woodchuck
  • **
  • Posts: 67
Re: A wild idea to combat dynamic range compression
« Reply #8 on: September 30, 2014, 01:45:19 pm »

What you're thinking of already exists, it is called decompression and (mostly) comes in a plug-in form for DAWs. It could be used for consumer purposes, but a fingerprint wouldn't do, you would need to know (or extract from the song information about) the way it has been compressed/limited, so for example the ratio used, attack, threshold etc., and this would of course have to be done at the mastering stage. While I think that it could be achieved as a plug-in feature or something like that, it would have to be standardized and used by a company like Apple or the like to get the attention it would need to justify such kind of invention, and I don't think that this will happen. :-) You can of course play around with decompression plug-ins and an audio player that allows insertion of VST plug-ins in its audio-path, see if you like what you hear. :-)
Logged

kstuart

  • Citizen of the Universe
  • *****
  • Posts: 1955
  • Upgraded to MC22 Master using preorder discount
Re: A wild idea to combat dynamic range compression
« Reply #9 on: September 30, 2014, 02:17:11 pm »

People already have tried doing this, it does not work very well.

You state: " Especially remasters of older music suffer from this."

Much easier is just to get the release prior to the remaster.

Steve Hoffman Forum has threads on almost every older music album, with discussion about which release is best.

theoctavist

  • World Citizen
  • ***
  • Posts: 228
  • a bad liver and a broken heart.
Re: A wild idea to combat dynamic range compression
« Reply #10 on: September 30, 2014, 10:06:09 pm »

Only if it doesn't have clipping or frequency shaping in the compressed version. The right way to do this is in fact to include the compression information in an UNcompressed file. But nobody does that, to speak of, although it's been patented for so long it's almost gone.Dolby AC2 and AC3 have level control side information. This is one such technique.  It's only used in theatres, though.
Logged

Walt Whitman

  • Recent member
  • *
  • Posts: 20
Re: A wild idea to combat dynamic range compression
« Reply #11 on: September 30, 2014, 10:46:26 pm »

I've worked occasionally with Pro Tools in the past. I used an Izotope add-on called Ozone that had something they called "Harmonic Exciter", where you could divide the audio band into segments of your choosing by frequency, and apply different parameters to add some "excitement" and, for lack of a better term, "ass", to the music. What was nice about it was that it just wasn't a "better" equalizer; it used the musical peaks and frequencies that were occurring in real time and applied the exciter just to what was happening in the music to great effect when used judiciously, I thought.

Using something like this could bring some life to ridiculously overcompressed music, without all the overhead of having a musical "fingerprint" for every piece of music. While it would not "uncompress" the music, it would add dynamics, power, warmth, etc. It's truly something you have to hear to realize its effect, comparing music before and after the Harmonic Exciter is applied.

It's funny that the OP brings up "adding" more dynamic range to heavily compressed music at this time, as I was recently thinking the same thing and how that would be a great addition to JRiver Media Center!

From the .pdf on the Izotope website, their product literature on "Harmonic Exciter" as it appears in Ozone 5:

10: HARMONIC EXCITER

An exciter can add “energy” to an audio signal and, in the context of the mastering domain, to an entire mix. This doesn’t necessarily mean increasing the volume, or boosting the signal, however. A sense of “energy” can be created by adding additional combinations of odd or even harmonics to the frequency content of a mix, and that’s what a harmonic exciter is designed to do.

What’s the Goal of Harmonic Exciting when Mastering?

If one of your mastering goals is to add power, punch, and other subjective terms such as “warmth” and “brightness,” the Harmonic Exciter might be a good tool to use.

There are many design strategies behind exciters, from dynamic equalization, waveshaping and distortion to short multi-band delays. When people speak fondly of recording music to tape machines, or reminisce about the sonic characteristics of tube equipment, part of that enthusiasm is a result of the harmonic distortions that these and other types of analog equipment impart on audio, as further described below.

How could distortions (a sometimes-dirty word in audio) be so appealing? Well, in truth, distortion in small doses isn’t necessarily a bad thing. If designed correctly and applied with restraint, distortion can introduce harmonics that add excitement or sparkle to a mix.

As an example: Ozone provides a selection of exciters modeled on tubes, triodes and tape saturation. When tubes saturate, they exhibit a type of harmonic distortion that is generally described as “warm” or “musical.” If used with care, this distortion creates additional harmonics that add presence or sparkle to the mix while still preserving a natural characteristic. You can perhaps see why boosting high frequencies with an EQ is not going to achieve the same effect. Boosting an EQ simply turns up the existing harmonics, whereas a Harmonic Exciter can actually synthesize additional harmonics. Tape saturation provides a similar effect, although the harmonics that are created are more “odd” than “even.” That is, tube saturation typically generates even harmonics that are an octave apart (again, “musical”), while tape saturation is a slightly more aggressive excitation, generating odd harmonics that are a fifth apart.

Principles of Harmonic Excitation

It’s very easy to overdo an exciter. What may sound good at 3.0…might sound even a little better at 4.0…and once you get used to that, you find yourself pushing it up to 5.0 to keep the “excitement.” Before you get caught up in the excitement (pun intended, we guess) and send it off to the duplicator or to SoundCloud, do a little reality check:

1. Compare it to some commercial mixes. Okay, in some cases these are overdone as well, but it depends on the genre and sound you’re shooting for. What works for a dance mix probably isn’t going to sound as appropriate on an acoustic jazz number.

2. Live with the “excited” mix for a while. At first listen an exciter is, well, exciting, but over time it can really sound fatiguing or even harsh and annoying.

Using the Ozone Harmonic Exciter

This is a very easy effect to use. (That could also be why it’s often overused.)

As a multi-band effect, the Exciter has up to 4 bands to work with, and each has a pair of controls. In most cases, you’re going to use the Amount control. In addition, you’ll probably apply excitation to the upper one or two bands, although there are some cases where tube saturation in small amounts across the entire spectrum can be musically pleasing.

With your mix playing back (of course), adjust Band 3’s Amount slider upward. As you move the slider up, you’ll hear what starts as sparkle and excitement, but it can quickly turn against you as you go up too far. Take note of the point where it starts sounding “annoying” and then turn it back down to 0.0.

Note that higher bands can usually bear higher amounts of excitation.

Clicking on the ‘Oversampling’ button engages more processing power. This will greatly increase the accuracy and sound quality of the analog modeling in the harmonic exciter. If your computer is powerful enough, you may run this in real time. Otherwise, it’s good to engage this tool before rendering/bouncing.

As you work with these effects, you can use the checkbox labeled “B” to bypass any multi-band processing applied to that band by Ozone. As a counterpart to the Solo checkbox, the Bypass checkbox is a useful tool for hearing what sections of your mix are being processed through each band. Note that this Bypass applies to ALL multi-band processing, including multi-band harmonic excitation, stereo imaging, and dynamics.

You can right-click on the multi-band spectrum display to remove or insert bands. For additional guidance, there is also a right-click and “Learn” feature that will automatically analyze your audio signal and determine the most appropriate, transparent places to place your crossover points.

General Harmonic Exciter Tips

• In most cases, excitation to the upper bands will give the desired effect. However, since Ozone offers an analog tube saturation model for the harmonic excitation, you can achieve a type of “tube emulation” on the low-mid bands as well. For this scenario, you’ll want to try a very small amount of equal excitation across all bands. (In other words, keep the Amount control low and work in single band mode.)

• You can get a “dirty” bass effect by applying some excitation to the low band. If you’re simply going for more bass level, use an EQ or maybe the Dynamics module, but if you’re looking to add grunge on those bass harmonics, sometimes you can turn to the Harmonic Exciter. You can also check out iZotope’s Trash for extreme multi-band distortion fun (www.izotope.com/trash).

• The Tape excitation mode provides a bright, saturated tone; the Tube model provides a thicker midrange rich tone; and the Retro model provides a heavy saturation character. The Warm exciter mode is similar to Tube, but is unique because it generates only even harmonics that decay quickly, and you can use Warm to drive tube style saturation quite hard, without creating “mud” in the low mid-range.

• Electronica mixes, or mixes that are comprised of more soft synths and samples than acoustic instruments, can sometimes sound a little lifeless. Use the Harmonic Exciter as described above to change this!
Logged
Pages: [1]   Go Up