App/Plug-in for Restoring Lost Dynamics for Audiophiles

[15Marsak]

Audio enthusiasts have long awaited (AI-driven) software that will restore dynamics to compressed recordings, without distortion or other artifacts. That is, a plug-in or standalone app-with a small to moderate learning curve-dedicated to repairing this especially horrid and all too frequent kind of sonic damage, which producers and musicians have deliberately imposed on loads of otherwise well recorded music, at least since the early years of Motown, Brian Wilson, the British Invasion and sadly ever since. https://en.wikipedia.org/wiki/Loudness_war

As the source material and functionality of countless medium and high end home audio systems have long been computer based, such restoration software from WAVES, UAD, Todd-AO, Steinberg (Spectralayers), Sonnox, Magix (Soundforge), Audacity, Adobe Audition, Acon Digital, and other leading audio software vendors is long overdue. And with Intel poised to release their line of AI accelerating Arrow Lake processors by year's end, now would be the ideal time to begin development of this unique product for the long vibrant computer audiophile market.

Please make this very special software happen very soon.

Thank you.

[SteveG_AudioMasters_]

LMAO! Even AI relies on at least a little information before it can achieve anything...

[15Marsak]

Like what little information might it need before attempting to such a repair, or recreation? I'm no kind of experienced recording engineer, but for example, during most live performances, don't horns usually sound louder than violins, or strings in general? Likewise, even in performances without horns, aren't there usually only one or two instruments that sounds louder overall than the the rest? If so, then could not those facts be encoded into the algorithm of the AI-assisted software for it then interpolate what the relative levels of all or most instruments in a recording should be, and thereby boost their levels accordingly? Would not this be at least a logical way to begin developing such an AI-powered app or plug-in?

[SteveG_AudioMasters_]

Nope - afraid not. The reason we have compressed music is because all the dynamics - and relative differences - between instruments, and come to that, the human voice have been artificially removed. And you can't use any information you think you have about relative differences, because in reality those differences aren't anything like what you think they are. For instance, in a live situation a single triangle strike can clearly be heard above an entire orchestra; your argument doesn't take any account of the amount of actual energy in an individual signal from an instrument at all.

But your real problem is that you have no idea how much compression was being applied - nor the dynamics of it (compression has attack and release envelopes that are effectively untrackable), so reversing it isn't going to be possible. Not only that, but it's common to apply completely different compression algorithms to different parts of the same signal - you'd never treat a human voice the same way you would an instrumental backing, for instance - and yet you're suggesting that with no knowledge of any of the processing that's been applied, some sort of AI magic is going to happen?

Dream on.

[15Marsak]

I don't doubt all of that is true. But the fact is that the actuall intellectual power of AI technology is continuing to expand and at an ever increasing rate. So, in about three years, and if there was sufficient demand, enough of those variables you described could concievably be described and mapped into some number of algorithms for the software to them turn out a far more realistic creation of the recording with restored dynamics, and with inaudible artifacts.  

[SteveG_AudioMasters_]

You don't get it, do you? These recordings are created like this in the first place - deliberately. There's no wonderful recording, full of dynamics to go back to. That's how the loudness wars worked - there's nothing to undo; they were recorded like this. There are some bands who had to re-record whole albums because they realised too late what was happening - the studio takes were recorded with excessive compression at source, usually because the record labels insisted.

[15Marsak]

Of course I know that it was often done deliberately, as the Wiki report attests; Berry Gordy and his engineers like Reggie Dozier would surely admit to pouring on the compression. I just wasn't sure whether it was more often applied during actual recording sessions or later after the final mix when the stereo or mono master when into the mastering room. But if that was bad enough, Dozier was once interviewed for an early 80s issue of REC E/P, where he admitted that he even "pinned the meters" during sessions. Such perversion for the aim of turning heads programmed to think that loudness-and even clipping distortion-is music.   

[15Marsak]

I wanted to add some text to my last reply. You can't edit your last post here?

[SteveG_AudioMasters_]

You can until somebody else replies.