You are thinking of TINA. Since NDSP models are proprietary, we don’t know how much of their sound is based on data collected by TINA vs how much is “white-box” component modelling. But insofar as their models are using data collected by TINA, yes, that is operating on the same principle as a parametric NAM model: capture the gear at many different settings and let machine learning interpolate the “gaps” between each of the captured settings so that the user doesn’t have to actually create a capture of every single micro-setting on every potentiometer in order to have fairly accurate tones when the pots are set in between the captured settings.
NAM creator Steven Atkinson wrote in that blog post I linked above about how with modern capture methods the parametric capturing process doesn’t need to be roboticized because it can be done with much fewer captures than a naive approach would otherwise suggest:
A related misconception is that it is practically impossible to collect enough data to make a model like this. For a single-knob model (e.g. of the “drive” knob), one could imagine sweeping the knob from 0 to 10 in increments of 1, requiring a total of 11 reamps. With the standard reamping file I’ve provided for NAM, this could be done in under an hour. However, to do this for 2 knobs, one might imagine that they would have to do all combinations of the knobs, making for 11x11=121 reamps. For this model, which has two knobs and two switches, this logic would suggest that I ran almost 500 reamps, recording over 24 hours of audio.
One way around this is to reduce the number of points–instead of increments of 1, I could do increments of 2 (0,2,4,6,8,10) and reduce the number of points by a factor of about 4 overall. But this is a losing game, since adding one more knob multiplies the work by a factor (of 11, or 6, in this example.) With only 2 values per knob (min, max), the 7-knob model above would have still taken over 100 reamps (and might have pretty dubious accuracy interpolating between those extremes!) This challenge has a name: the curse of dimensionality.
Since that’s a really big problem, there’s been a lot of work to fix it, falling largely under the scientific field of optimal experimental design. It’s beyond the scope of this blog post to get into the details, but the punchline is that using some advanced methods from this field allowed me to trim the time I spent (including the time spent moving the knobs between reamps) to just over an hour. Work smarter, not harder!
Of course, an important difference between QC models and NAM is that unlike QC models, NAM is open source, meaning that anyone can build on NAM and NAM models are freely interoperable between different NAM players. So while QC models might be functionally similar to NAM models in some ways in how they work in isolation, their differences quickly become apparent once the user has any desire to use their QC models outside of the QC ecosystem: of course, they can’t, they are stuck using them on the QC hardware with models and effects created by NDSP for that hardware. Whereas NAM is usable in a large variety of different software apps and hardware devices made by dozens of different developers and manufacturers.
I think it is plausible that they could go in the NAM direction, especially if enough users request NAM support. There have already been a few companies that have their own proprietary modeling tech but have also either already added NAM support or said they will in the future e.g. Fractal Audio, Blackstar, Hotone, Tonocracy, Two Notes… so it wouldn’t exactly be unprecedented if NDSP decided to as well.
NAM is already empirically superior to NDSP’s capture technology according to objective metrics such as null tests and more subjective measures such as listening tests. Sure, NDSP could continue waging an arms race against NAM and try to get ahead. Or, they could embrace NAM as an open standard and refocus the bulk* of resources they would have spent fighting the capturing arms race into other lines of work, such as continuing to improve their time-based effects (which NAM cannot model), quality of life improvements across the board, compute optimization, more PCOM support, artist collaborations, etc etc… things that open-source fundamentally cannot improve upon because they are truly unique to the NDSP ecosystem.
* If I were in their shoes I would also make sure to set some slice of those resources aside to invest into continuing to improve the NAM open-source algorithms to ensure that NAM remains the best capturing tech available. Since there is a large and growing ecosystem of companies investing into NAM development already, no single company needs to shoulder that entire burden like they do with proprietary capturing tech, resulting in significantly reduced contributions required from each individual company to achieve comparable (in actuality, even better) results.
Seriously tho the benefits of being interoperable with this large and growing ecosystem of open source models and players cannot be overstated. Even if it doesn’t happen in the current generation of devices, perhaps the support this request gets can inspire an upgrade in the next generation QC.