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. Loudness war - Wikipedia
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, Antarestech, 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.
So, if you want a piece of software to analyze “Summertime Blues” by Blue Cheer, undo each element’s compression, distortion and make it dynamically like…wait… make it like what?
Machine learning has to learn/compare something in order to know what you want it to do.
What would you have the software use as its reference audio to apply to the Blue Cheer track?
Which of course, wouldn’t be the same instruction set to use on Bobby Vinton’s “Blue On Blue”.
Now…give it a few years and I can definitely forsee having models to analyze an existing squashed-dynamics 45…and completely regenerating an indistinguishable sound-alike with all kinds of enhancement. Karaoke 2.0. Won’t be much fun imo… but can’t stop the future.
You can do this by using a technique called sidechaining/ducking and then re-applying those eq peaks back into itself (back into the audio layer you want the dynamics restored). So for example in FL Studio there is a audio effect called the peak controller (which allows you to internally manipulate audio internally, kind-of like the idea of cv for modular synthesis) and you can use that along with fruity parametric EQ 2 to internally re-record the dynamic peaks (on a automation level) back onto itself and then re-apply those peaks dynamically. You could also use the multiband compressor to do this along with the fruity peak controller.