IV anywhere on the horizon?

It seems like people are missing the point if Fractal added profiling to it's feature set, it's not a "modelling vs profiling" debate on which approach yields faster results. The way I envision it is a new "rig profile" block which would basically replace an amp/cab. You would still have access to all Fractal has to offer, but this is an extra feature.

I think it's extremely short sighted to not see the potential in how awesome this would be on a Fractal unit being able to share profiled rigs/pedals/cabs etc. It puts a lot more power into the users hands to profile boutique/custom/signature rigs and share them around.
You want an extremely detailed 5153? Pull up the fractal model. You want a profiled rig of a 5153 with some boutique overdrive pedals boosting it? Pull up a user profile.
general response I guess but not really sure how my related post above, for example, misses the point or is short sighted - I've added a bit more detail there to explain further but unedited, I think it covered the bases - not everyone wishes for the same things so just because some don't prefer a given feature possibility does not make them blind to the possibilities or short sighted, and, as I eluded above, a key factor is that there's opportunity cost whatever new features are developed.
 
Last edited:
Profiling is patented and I've been threatened more than once by Kemper's attorneys.

Using neural networks is possible but training a NN takes hours and requires hardware that a consumer modeler would not have (i.e. inference accelerators).

That said, my tests have shown that white-box modeling outperforms profiling and NN in terms of accuracy and aliasing. Whether that sounds "better" is subjective.
The NN is the second best way to solve a problem. :)
 
Had a Kemper, have a QC (and of course an Axe III and FM3).
Profiling is a blast when you want instant gratification of copying 'your' amp.
But when you a/b back and forth, man that aliasing difference is real.
Play the Fractal for 10 minutes, nice and smooth on the top end.....flip to the QC and man it's got this (only word I can think of is) harshness on the top end. It doesn't bother me if I just play the QC, but if I use the Fractal first, and my ears get adjusted to it....the flip to the QC just hits me the wrong way. That does NOT happen when I go from QC to Fractal.

Would I like to have Fractal profiling? Of course! But patents are patents....and we need to embrace the killer features we already have in the Fractal.
I've said this for years with Neural DSP.

They have a super harsh 5k / 6k that's present on all of their plugins & the QC.

Any time I mix with a Neural DSP plugin, it's the first subtractive EQ move I make - destroy 6k.
 
My frame of reference is my old stage rig - a Peavey Mark IV head with a 6-10" cab for guitar. And then my organ and Leslie. It took at least 2 guys to move that stuff. These days I can use a gigbag to carry my guitar on my back, and a nice padded bag for an FM9 or FM3 over my shoulder. No comparison. My big-ass clumsy feet would kill a QC within a month of gigging.
Yeah it takes a couple "roadies" to move my organ too.
 
It doesn't bother me if I just play the QC, but if I use the Fractal first, and my ears get adjusted to it....the flip to the QC just hits me the wrong way. That does NOT happen when I go from QC to Fractal.
Of all the comparisons I've read between various other brands, and Fractal units, that comment is about the best way I've heard it explained!!
 
Profiling is patented and I've been threatened more than once by Kemper's attorneys.
What scumbags, they're just afraid :laughing:. I understand business is business but even within that spectrum it seems a bit crazy for them to do this.
I wonder how QC got around this for their profiling, maybe they're forfeiting some of their margins over to kemper. I always wondered why some 2 bit company doesn't make a profiling vst for the masses to share studio rigs, this might be part of the answer if Kemper is holding the keys so tightly.
 
Just to try to chime in profiling, what about having something like a "Comp-match" block, where an AXE captures/matches the relation between the level of the input signal and that of the output signal with the reference tone?
In combination with a Tonematch block (and a well-tweaked Amp&Cab model), I expect it would enable an AXE to recreate the character of the reference tone closer.
 
+ i dont see what they have to say as fractal don’t do profiling 🤔
At one point I remember Cliff hinting in a post that the AxeIII might someday include a form of profiling. It was only a hint so I have no idea (or details) about what he was referring to. Unfortunately I forget the name of the thread. But it was quite some time ago. Perhaps even pre-Covid, though it was after the AxeIII was released.

It sounds like QC got wind of it and had their lawyers go into attack dog mode. Odd thing is, people (or companies) seem to usually behave this way when they're insecure about something. Maybe QC realizes how inferior their product is so they won't let the market decide which product is better. And they probably also fear that Cliff is such a monster of a developer that whatever Fractal produces will kick ass on anything else on the market. But I just speculate . . .
 
I always wondered why some 2 bit company doesn't make a profiling vst for the masses to share studio rigs, this might be part of the answer if Kemper is holding the keys so tightly.
There is currently some open source machine learning profiling/capturing/whatever you want to call it, being discussed on tgp, it’s in the early stages though but sounds good. The videos give waveform comparisons, etc.

Still on the technical side not currently plug-and-play. Potentially might not be long before someone slaps it into a vst however.

https://www.thegearpage.net/board/i...odeler-v0-2-open-sourced-amp-modeler.2366776/
 
At one point I remember Cliff hinting in a post that the AxeIII might someday include a form of profiling. It was only a hint so I have no idea (or details) about what he was referring to. Unfortunately I forget the name of the thread. But it was quite some time ago. Perhaps even pre-Covid, though it was after the AxeIII was released.
I remember for sure one in the II era. He thought, starting for his own comparison algos, to make one function to "profile" an amp.
 
There is currently some open source machine learning profiling/capturing/whatever you want to call it, being discussed on tgp, it’s in the early stages though but sounds good. The videos give waveform comparisons, etc.

Still on the technical side not currently plug-and-play. Potentially might not be long before someone slaps it into a vst however.

https://www.thegearpage.net/board/i...odeler-v0-2-open-sourced-amp-modeler.2366776/
Note that some random person doing their own personal open source project is a very different thing from releasing a commercial product. Nobody is going to sue you for doing a little open source project that makes you no money but if there are patents around this stuff for guitar gear, when it's a commercial product Kemper lawyers might get involved.

Worth remembering is that patent lawyers are concerned with upholding that patent and that might involve cease and desist letters or litigation even if it does not sound fair or right. Law is not concerned with "morally right". Patent lawyers might even try these tactics for products they think might be in violation of the patent. Kemper's patent is worded kind of broad.

You can look at for example Fender vs Rickenbacker in how upholding your trademarks etc turns out when you don't do anything vs sue the shit out of anyone trying to make a Rick copy.

Profiling is patented and I've been threatened more than once by Kemper's attorneys.

Using neural networks is possible but training a NN takes hours and requires hardware that a consumer modeler would not have (i.e. inference accelerators).
How does the Quad Cortex do its captures if it doesn't have this sort of hardware? Does its capture feature just add more data to an existing model and use that to churn out the end result?

My experience was that it did a really good job replicating my real amps with those specific settings and seemed to behave nicely with things like rolling the guitar volume or adjusting the gain in the capture block. Not exact, but close enough. EQ of course did not behave like the real amp but I think it could if you could setup the capture block EQ to match your amp's tone stack behavior when making the capture.

Now with the Axe-Fx 3 I don't really care if it has profiling or not. I have proven to myself that I can dial the existing models to sound extremely close to the amps I have/had even when they are not modeled specifically in the Axe-Fx 3. Profiling at best would shortcut matching my favorite real amp settings in the Axe-Fx 3. With nearly 300 amp models of about 100 unique amps, there's not much it can't do for amp sounds.
 
How does the Quad Cortex do its captures if it doesn't have this sort of hardware? Does its capture feature just add more data to an existing model and use that to churn out the end result?
My experience was that it did a really good job replicating my real amps with those specific settings and seemed to behave nicely with things like rolling the guitar volume or adjusting the gain in the capture block. Not exact, but close enough. EQ of course did not behave like the real amp but I think it could if you could setup the capture block EQ to match your amp's tone stack behavior when making the capture.

Now with the Axe-Fx 3 I don't really care if it has profiling or not. I have proven to myself that I can dial the existing models to sound extremely close to the amps I have/had even when they are not modeled specifically in the Axe-Fx 3. Profiling at best would shortcut matching my favorite real amp settings in the Axe-Fx 3. With nearly 300 amp models of about 100 unique amps, there's not much it can't do for amp sounds.
You need time & CPU to train NN, but once trained will do the job. Or maybe there is an initial "intesive" train, and the Quad do only light adjustment.
 
And what about "analog computing" instead of digital in the next Fractal products? Voltages instead of digit & conversions... it seems like amp emulation is an ideal use of this kind of hardware...
 
Note that some random person doing their own personal open source project is a very different thing from releasing a commercial product. Nobody is going to sue you for doing a little open source project that makes you no money but if there are patents around this stuff for guitar gear, when it's a commercial product Kemper lawyers might get involved.
Computer code, which is the foundation of Kemper’s system, is normally copyrighted, not patented. And U.S. copyright law allows “Fair Use”. Fair use might offer protection through the educational aspect but I’d never go into some personal project intending to use that as a defense unless I was in school and using it as part of a class.

If they have a legitimate patent it’s trickier, but it looks like there are some exceptions to patent law that’d allow it. https://blog.jipel.law.nyu.edu/2020/04/under-the-shadow-of-a-pandemic-fair-use-in-patent-law/ is an interesting article and the papers it links to talk about it. I love looking this stuff up.

Worth remembering is that patent lawyers are concerned with upholding that patent and that might involve cease and desist letters or litigation even if it does not sound fair or right. Law is not concerned with "morally right". Patent lawyers might even try these tactics for products they think might be in violation of the patent. Kemper's patent is worded kind of broad.
Definitely.

You can look at for example Fender vs Rickenbacker in how upholding your trademarks etc turns out when you don't do anything vs sue the shit out of anyone trying to make a Rick copy.
From what I remember, trademarks are different from patents and copyrights, but the general practice of protecting intellectual property equally, and immediately, is the same. Gibson and Fender failed to do those things for years and then tried to regain control by suing a couple of big violators way too late. They were lazy and foolish.

How does the Quad Cortex do its captures if it doesn't have this sort of hardware? […] Now with the Axe-Fx 3 I don't really care if it has profiling or not.
Profiling doesn’t interest me at all. There’s almost infinite differences between points A and B on each knob on an amp. Add each knob’s effect on the overall sound and it doesn’t take a lot of thinking to figure out they’re not going to try to capture every possible sound from the amp. Reduce the profile to the “perfect” sound for one person who has different tastes and suddenly the profile is nowhere as good, or as useful.

Interpolation between one setting and another in a profiled amp is entirely different than the difference between the settings in a modeled amp. They’re trying to morph from one setting to another, whereas the model, if it’s done right, knows exactly what the in-between sound should be.

I have proven to myself that I can dial the existing models to sound extremely close to the amps I have/had even when they are not modeled specifically in the Axe-Fx 3. Profiling at best would shortcut matching my favorite real amp settings in the Axe-Fx 3. With nearly 300 amp models of about 100 unique amps, there's not much it can't do for amp sounds.
The underlined part is my thought too. It’d be the shortcut … if the profiled sound was what I wanted. Odds are really good it’s not. Tube amps are extremely dynamic, they adjust to every change we make, both to the amp’s settings, and to the guitar’s knobs, and to how we’re hitting the strings, and our guitar’s construction. A single snapshot/profile is going to capture that? Really?

It’s a much more complex situation than can be captured in a single snapshot. Math FTW!
 
Last edited:
Back
Top Bottom