Curious: modeling & profiling. What’s the difference?

DallasGlenn2005

Experienced
Not looking to create debate which is better/worse or argue for/against the merits of one or the other. Only looking for someone to explain what profiling is vs modeling. Until a recent post, I never realized that they are two different things.
 
Profiling is like taking a snapshot (or audio photograph) of your amp with the amp controls set to whatever they were at the time of taking the snapshot. Modelling however is reproducing the sound of of the amp based on the components that make up that amp. The advantage of modelling in the Fractal sense is that you can adjust the amp controls and it responds like the real amp does, whereas on a profiled amp the controls don't get profiled.
 
black box <> white box.

Black box (profiling) is like a single "audio photo" of an amp. It sounds fine but it's not longer authentic if you start tweaking. You need multiple profiles to get accurate "audio photos" of different settings on the amp.

White box (modeling) is building a digital version of an amp, including every component (every transistor, resistor etc.). Everything acts like the original amp.
 
How I think about it is if you already have your sound, you have a few amps and tones that you like and just want to make it more portable/ more consistent for gigs or recording, whatever, then profiling works for you. If you don't quite have your tone or you want to keep playing around, I see modelling as the better platform as that.

Both are great, it just depends on who you are and what you want. I like being able to fiddle around with everything and I had no idea what I wanted to sound like, so I went Fractal with modelling. However there are plenty of people who have been touring forever that know their sound, and just want that in a little box. All the power to them.
 
"Profiling" is how kemper brands their approach. Ck has refered to profiling as "our modelling". I find that people, at times, tend to believe the 2 approaches are totally distinct. I don't quite think so.

Modelling typically involves creating digital representations of amp structures and components, the aim being to emulate the amp feel and sound. Fractal does this for us. Profiling instead targets a given tone "on the spot" and -- I believe -- aims to get there by automating selection of parameters that are also built in. Additional steps -- again, from what I believe; call it a semi educated guess, not being CK -- include eq matching and/or cab capture similar to IRs.

I see advantages to both depending on use case. But you're not asking about which is better, so that's irrelevant.
 
Last edited:
black box <> white box.

Black box (profiling) is like a single "audio photo" of an amp. It sounds fine but it's not longer authentic if you start tweaking. You need multiple profiles to get accurate "audio photos" of different settings on the amp.
White box (modeling) is building a digital version of an amp, including every component (every transistor, resistor etc.). Everything acts like the original amp.

Does FAS use AI to improve component algorithms? I'm sure writing code to create algorithms to replicate the nonlinear response of electrical components is a tall task. And each FW update contains improvements on those algorithms to move away from an "approximation" and converge on an exact digital recreation. Is AI capable of doing this, or would the act of writing the AI be more difficult than just DIY the algorithms?
 
Last edited:
I always describe the difference as:

  • Profiling: go to a restaurant, order from the menu, take what you get. Maybe great, maybe not. Essentially, "sold as is" based on what is available as a completed pre-prepared meal.
  • Modeling: go to a restaurant, walk into the kitchen, and make your own meal with the tools and ingredients available. Essentially, "make it your way"
 
Does FAS use AI to improve component algorithms? I'm sure writing code to create algorithms to replicate the nonlinear response of electrical components is a tall task. And each FW update contains improvements on those algorithms to move away from an "approximation" and converge on an exact digital recreation. Is AI capable of doing this, or would the act of writing the AI be more difficult than just DIY the algorithms?
I don’t believe he uses AI. The stuff he is doing involves a grasp of the physics that I don’t think AI is capable of doing.


See this recent one as an example:
https://forum.fractalaudio.com/threads/you-guys-are-in-for-another-treat.171270/post-2063967
 
I always describe the difference as:

  • Profiling: go to a restaurant, order from the menu, take what you get. Maybe great, maybe not. Essentially, "sold as is" based on what is available as a completed pre-prepared meal.
  • Modeling: go to a restaurant, walk into the kitchen, and make your own meal with the tools and ingredients available. Essentially, "make it your way"
You can tweak a profile hm (just saying)
 
Does FAS use AI to improve component algorithms? I'm sure writing code to create algorithms to replicate the nonlinear response of electrical components is a tall task. And each FW update contains improvements on those algorithms to move away from an "approximation" and converge on an exact digital recreation. Is AI capable of doing this, or would the act of writing the AI be more difficult than just DIY the algorithms?
I don’t believe Cliff uses AI. From what I’ve read Cliff is an engineering and math wizard! My guess is it would take longer to create the AI to make the models, as the AI would need to be programmed to do the process, which would need to be checked, whereas a human doing it can make all the measurements and make the model without needed to create AI. And I’m sure it‘s wicked fun to open up these amps and see What makes them tick!
 
Back
Top Bottom