Gee whiz, modeling is dead!

Captures/profiles will NEVER be as versatile or fun as a full component models.

I'd sort of disagree with the "fun" part.
One of the reasons I'm interested in capturing is that I could capture some wild, nested sound creations that would take up plenty of blocks and likely look all confusing in a component modeling environment. That's also why I find all the neural learning stuff interesting as the KPA wouldn't capture these things properly.
 
this is a false dichotomy, theres no reason all these switches cant be taken into account on a profiling system
Except that they aren’t yet. I can’t adjust a 5-band EQ in a capture, or pull bright knob, etc. it’s got to be yet another capture to sift through. Maybe someday though, sure.
 
I can’t adjust a 5-band EQ in a capture, or pull bright knob, etc.

Well, you could.
It's all a matter of the amp topology the capture would be embedded in.
Tonex is a semi-decent example to start with already, as you can at least chose where to place the EQ. If this concept was vastly expanded (fwiw, something I also wish was available for component modeled amps, but from all I know, only FAS does it), captures would be a vastly more flexible thing to deal with.
You could be given a choice of which "amp environment" the capture would be embedded into. There could be a few pre-made ones and you could have an option to roll your own, too.

As some blunt examples:
Amp environment (1) would be just the typical pre-gain, post BMT and level, done.
Amp environment (2) would be pre-gain and pre-mids, B and T would still be post.
Amp environment (3) would be a double tone stack with pre- and post-BMT stacks.
Amp environment (4) could add an embedded compressor pre or post.
Amp environment (5) could add a boost of some sorts (think of the way FAS does it, with a number of drives being tied to the amp).
Etc.

Would that give you the same tone manipulating experience the original amp would give you? In some cases it might, in many cases it won't, for the very same reasons we all know about. But it'd still make up for an incredible flexible environment. And quite sometimes it'd go vastly beyond what the original amp would allow you to do.
In all these cases, the captures would sort of serve as a basic character kinda thing.

Now, all that is possible with the current capture tech already, but you'd have to deal a lot with additional EQs, compressors and what not to get there. If you could just select from a library of captures and amp "containers", it'd be much easier to get much more flexible results. I'd certainly enjoy something like that a lot.
 
I’ve wondered about this - rather than machine learning to be used to brute force copy the sound of an amp, using machine learning to code algorithms that bridge any differences between what the schematic SHOULD do and what the real amps actually do.

I think a mixture of white box and black box modelling has been how UAD and other plugin emulation companies have done things for some time.

Relab have posted a bit about how they’ve used their own machine learning to reverse engineer complex reverb algorithms to the point where they are 1:1 to the originals
 

"it takes the neural network a couple of weeks to generate the final algorithm for each amplifier"

Sounds familiar, NDSP are doing something very similar.

@FractalAudio explained it very well here:

That's an enormous amount of data.
NDSP have a patent pending for an automatic system that twists the knobs automatically and takes many samples.
No wonder it takes them months to model a single amp.




ndsp machine.png
 
Which capture platform currently allows me to capture a Mark V, and adjust the 5-band after the fact?

None. Hence "could".
Still, you could just add one on your own.
And fwiw, the Boogie 5-band EQ is something pretty trivial as it's sitting behind all gain stages (just in front of the power amp, which, in typical settings, shouldn't have that much of a big effect).
If I was to capture any Boogie with a 5-band EQ, I'd probably just leave it off.
 
None. Hence "could".
Still, you could just add one on your own.
And fwiw, the Boogie 5-band EQ is something pretty trivial as it's sitting behind all gain stages (just in front of the power amp, which, in typical settings, shouldn't have that much of a big effect).
If I was to capture any Boogie with a 5-band EQ, I'd probably just leave it off.
And Pull Bright? Mid boost? Point is, unless it’s a simple Marshall/Fender/Vox circuit, it’ll be a problem until one of these companies figures it out.
 
And Pull Bright? Mid boost? Point is, unless it’s a simple Marshall/Fender/Vox circuit, it’ll be a problem until one of these companies figures it out.

See:

Would that give you the same tone manipulating experience the original amp would give you? In some cases it might, in many cases it won't, for the very same reasons we all know about. But it'd still make up for an incredible flexible environment. And quite sometimes it'd go vastly beyond what the original amp would allow you to do.
 
Except that they aren’t yet. I can’t adjust a 5-band EQ in a capture, or pull bright knob, etc. it’s got to be yet another capture to sift through. Maybe someday though, sure.
Another capture is fine. Maybe several of them. This wouldn't be super different than interpolated samples of instruments or 3d digitization of existing sculptures for mass production
 
Yeah dude. I read your post. 😂

So there.
Really, I am quite aware of the shortcomings of captures, so you're more or less preaching to the choire. And no, I don't think that technology will be improved so much as to really rival component modeling in those very areas any day soon. Yet, capturing is a pretty valid approach for many things - and IMO the way the captures are treated (at least given the currently available units) could be vastly improved. Sometimes even so much you possibly wouldn't need a component modeled version anymore.
That's all I was saying.
 
Back
Top