MASTERING WITH A BRAIN!?
We recently received an initially skeptical (but ultimately favorable) review of Elevate from a popular Youtube account, and I love that. There's a lot of marketing nonsense in plug-in-land and it's important to be skeptical about companies claims. And to be fair, I do make some bold claims about our technology. So what do I mean when I say it's a Mastering Limiter with a Brain!?
Obviously it doesn't come with a fleshy lump of pink or grey matter. 🧠
Also, it's not what I would call AI (and I haven't). AI (or artificial intelligence) is not a well defined term, but in the popular imagination it means replacing a person with a computer, or at the least, having a computer do a human's job. Elevate doesn't try to replace a mastering engineer's brain, instead it uses brain-like intelligence to do a job none of us would ever consider doing manually - solving a particular underdetermined problem that comes up when trying to create a loud master that sounds good.
WHY IS A BRAIN BETTER THAN AN ALGORITHM?
(SOLVING UNDERDETERMINED PROBLEMS)
If I told you that x + y = 7, what are the values of x and y?
I know, you didn't think there'd be a quiz. 📚
If you asked your high school algebra teacher this question she'd tell you that it can't be solved. There are an infinite number of values that x and y can take and still add up to 7. You can't solve this with or without an algorithm, this is called an underdetermined problem - heavy sounding stuff.
However, I'd bet that a lot of you said, "x and y are 3 and 4", or maybe, "7 and 0". I bet no-one said, "x = 2,705,783 and y = -2,705,776", and I think we'd all agree that "3 and 4", or "7 and 0" are much better answers than "2,705,783 and -2,705,776". It turns out that brains are great at finding good answers to underdetermined problems. In fact, that's what they do every time we ask them to get us to the grocery store, or which checkout line to choose once we're there. Solving underdetermined problems is just using mathematical language to describe common sense.
HOW DOES THIS APPLY TO MASTERING, THOUGH?
Elevate breaks the input signal into the 26 critical bands of human hearing (which we talked about here), then for each sample we need to find the best gain value for each band so that when we add them all back together the output isn't greater than 1. This is just like our x + y = 7 problem above, in fact this one is:
a + b + c + d + ... + x + y + z < 1
And this is where the common sense from a "brain" helps us make good choices for those gains.
To get mathematical for a second, finding the "best" solution to an underdetermined problem is called solving an optimization problem, and to solve it we apply a "model" to the problem that gives the computer an idea of what a good solution looks like. Elevate has a model which tries to define "best" as best-sounding, then it sets up an optimization problem and solves it every sample to give you the loudest, best sounding master.
And that's what I mean when I say MASTERING WITH A BRAIN, but what's most important is that it's a difference you can hear.
Thanks for coming to my TED Talk,