I know it’s a long shot but since Line6 and Two-Notes seem to be entertaining the idea of allowing their next hardware units to load .NAM models, just throwing this out there:
Would Neural DSP be open to allow folks to incorporate .NAM profiles in their Quad Cortex units / signal chains?
I think if this gets accommodated, QC could see a surge in adoption. The Neural Capture works ok for some stuff but not so much for other use-cases; having the option to run either a Neural Capture or a NAM model would make the QC a lot more attractive.
For an ‘end user’ perspective, it would be a huge plus to to be able to run NAM profiles on the QC.
But, I can imagine that NDSP is reluctant to do this, because:
It would inevitably allow a direct comparison of NAM / QC profiling capabilities, and thus depreciate the current profiling technology of the QC. (albeit, for me, it is very different : the QC is way faster, but NAM is more accurate and based on actual IA technology)
If NDSP creates a Marketplace one day, it would perhaps cut some revenues from the marketplace, since there’s a lot a quality NAM profiles for free. (mind you, there’s also good quality QC captures for free, so I don’t see this point as important)
There might be some legal / copyright issues (even if NAM is open source, it doesn’t means that everyone can ‘steal’ the code for their own profit, but I haven’t checked the NAM licence kind…)
As a company that is (has been?) seen as ‘cutting edge’ in modeling, it would be an admission of weakness to incorporate ‘foreign’ technology in their ‘flagship’ modeler.
NAM is currently the promising technology, but we don’t know what awaits us tomorrow… If NDSP invest in this technology, they’ll have to maintain it during years, and it has a cost.
It would be much more interesting for NDSP to enhance the current capture capabilities of the QC in order to stay at the level of the competitors (ToneX, NAM) , but there is currently nothing that indicate that they will enhance anything in this area… Perhaps after the CorOs 3.x releases ? (Plugin integration is consuming a lot of resources)
Getting NAM models in (not using the tech to train models) should not be any more alien than importing an IR into the unit.
There’s value in both takes: improving Neural Capture and allowing users to import their already-trained NAM models into a QC signal chain.
There’s been a lot of talk about a Marketplace for years but nothing has come to fruition yet and even if it does, I don’t suspect it’s too much trouble to restrict that Marketplace to NDSP-proprietary Neural Capture format files.
The NAM license allows commercial products to make use of the software through the MIT license.
IRs were a “Fractal” thing right before they got wide adoption. NAM files could end up the same. If you think about it even with IRs - some are great (same goes for NAM profiles), while some have you looking elsewhere.
Had this thought when i saw two notes implemented nam models into their genome app. I think neural def should give us the option to use them. Otherwise alot of people will choose nam and whatver hardware options come out. Allowing nam models will only help keep people on the qc more.
NAM’s profiling algorithm is more accurate but also takes longer to accomplish. Able to capture fuzz is big plus. Tonex says it’s able to capture fuzz pedals too but i havent done it personally. This is the main plus i see in adding the nam algo into the QC.
This would be so amazing! Another application would be the ability to load neutron quantum speaker cabs which use NAM to model cabs in a much more realistic way than tradition IR’s