Helix Needs a Capture Block: Hear Me Out.

[Orvillain]

I use ToneX with my custom captures when I'm feeling lazy, or I'm travelling and just need something quick and dirty to get myself up and running. The rest of the time, I'm more interested in modelling than profiling.

 

[la szum]

Pretty sure, Eric, mentioned L6 would not be doing what anyone else
has done in that recent interview with them 2 P & W guys whose names
I can't recall, or probably never learned in the first place. I took it to mean
profiles/captures.

People can change their minds, but he seemed pretty staunch on that point.

 

[dronerstone]

Pretty sure, Eric, mentioned L6 would not be doing what anyone else
has done in that recent interview with them 2 P & W guys whose names
I can't recall, or probably never learned in the first place. I took it to mean
profiles/captures.

People can change their minds, but he seemed pretty staunch on that point.

They've been saying that for a while. Legit points, ymmv tho.

 

[CakeEater]

I wouldn't mind a capture block, it wouldn't have to be NAM. I'd just like to see the likes of Worship Tutorials and Tone Junkie roll out captures of their esoteric collections like they do for Kemper.

 

[dronerstone]

I'd certainly appreciate the possibility, but it's definitely not a 100% necessity.

Usually when I need to "profile" anything, I either do via A/B-ing the preamp section using fx loops, or I go by ear until it's close enough for my tastes.

I'm a simple man.

 

[EOengineer]

Pretty sure, Eric, mentioned L6 would not be doing what anyone else
has done in that recent interview with them 2 P & W guys whose names
I can't recall, or probably never learned in the first place. I took it to mean
profiles/captures.

People can change their minds, but he seemed pretty staunch on that point.

If we’re going to take those comments literally I’ll wait here while L6 removes the modulation, distortion, cab, reverb, amp, and delay from Helix because most processors have those too.

L6 doesn’t necessarily need to enter the profile capturing space as there may be actual hardware limitations in Helix that make that a non starter, but if something approaching a “standardized” universal profile format materializes via NAM or someone else, then they should at least consider adding a profile player block in addition to their very capable circuit modeling, just as they’ve accommodated impulse responses.

 

[2dor]

If we’re going to take those comments literally I’ll wait here while L6 removes the modulation, distortion, cab, reverb, amp, and delay from Helix because most processors have those too.

L6 doesn’t necessarily need to enter the profile capturing space as there may be actual hardware limitations in Helix that make that a non starter, but if something approaching a “standardized” universal profile format materializes via NAM or someone else, then they should at least consider adding a profile player block in addition to their very capable circuit modeling, just as they’ve accommodated impulse responses.

Two-Notes may pull one off if Genome is in any way indicative of their future intentions. Line6 may not want to go down this path but folks like Two-Notes and Atomic (looking past their track-record since COVID hit) may pull some nice surprises in this regard.

 

[EOengineer]

Two-Notes may pull one off if Genome is in any way indicative of their future intentions. Line6 may not want to go down this path but folks like Two-Notes and Atomic (looking past their track-record since COVID hit) may pull some nice surprises in this regard.

It will be interesting to see how this shakes out long term. I wouldn’t expect L6 to approach supporting playing profiles until the smoke clears and we’ve settled on something approaching a universal standard - and I don’t blame them.

Given the rapid iteration that we’re seeing with NAM development in the open source space, and signs of adoption from competitors, I suspect they’re eventually going to be forced to at least add support for playback.

 

[mbenigni]

QC/Kemper dont use the same technology, they are using distortion/amp models running through a bunch of secret sauce filters, its not an actual neural model, so the end result is probably less intense to run than a standard NAM model.

Interesting - thanks for the clarification!

So my takeaway would be, Line 6 (or anyone else) could opt to add a NAM loader block, but it would be at the end user's discretion to use content (profiles) that meets their needs, without making other objectives impossible. Which may be a usability/ support liability concern Line 6 would rather not take on. Then again, it's not really all that different than e.g. Poly Capo on HX Stomp, which is known to use enough CPU as to preclude presets with certain other blocks. (Though I am completely neglecting the potential for increased latency here.)

 

[2dor]

It will be interesting to see how this shakes out long term. I wouldn’t expect L6 to approach supporting playing profiles until the smoke clears and we’ve settled on something approaching a universal standard - and I don’t blame them.

Given the rapid iteration that we’re seeing with NAM development in the open source space, and signs of adoption from competitors, I suspect they’re eventually going to be forced to at least add support for playback.

Of course. That said, NAM profiling has been pretty much constant (the profile "architecture" types the trainer spits out by default without tweaking under the hood) from close to when it started getting traction (last year in Jan / Feb).

In a sense, you could say the dust has settled at least as far as the quality of the profiles and their performance hit "out of the box" which is probably why we're starting to see hints of what may be coming next.

Those Dimehead Electronics folks are about to release their NAM Player pedal soon which supposedly can play full-size standard NAM profiles, IR, convolution reverb with ~0.5ms latency; price though is about 500 EUR + VAT which is a bit pricey but understandable for a boutique builder.

 

[Lysander]

I’m not sure this is true. Can anyone here say with authority that a NAM profile is more CPU intensive than a component-based model?

NAM is based on WaveNet neural networks, and these are quite resource intensive, particularly for DSP hardware. This is also the technology used by ToneX, as far as i know.

Note that NAM now supports the so-called "nano" (i.e. lightweight) models, which require less hardware at the cost of slightly less fidelity.

Normal capturing technology is based on error delta reduction and EQ matching, which results in low DSP requirements. This is the approach used by Kemper, Headrush... and Neural DSP, despite all their "biomimetic AI technology" mumbo-jumbo.

tl;dr: NAM is (quite) more DSP/CPU intensive than a component model. Other capture solutions are the opposite.

 

[mbenigni]

NAM is based on WaveNet neural networks, and these are quite resource intensive, particularly for DSP hardware. This is also the technology used by ToneX, as far as i know.

Note that NAM now supports the so-called "nano" (i.e. lightweight) models, which require less hardware at the cost of slightly less fidelity.

Normal capturing technology is based on error delta reduction and EQ matching, which results in low DSP requirements. This is the approach used by Kemper, Headrush... and Neural DSP, despite all their "biomimetic AI technology" mumbo-jumbo.

tl;dr: NAM is (quite) more DSP/CPU intensive than a component model. Other capture solutions are the opposite.

Good to know, thanks.

 

[2dor]

NAM is based on WaveNet neural networks, and these are quite resource intensive, particularly for DSP hardware. This is also the technology used by ToneX, as far as i know.

Note that NAM now supports the so-called "nano" (i.e. lightweight) models, which require less hardware at the cost of slightly less fidelity.

Normal capturing technology is based on error delta reduction and EQ matching, which results in low DSP requirements. This is the approach used by Kemper, Headrush... and Neural DSP, despite all their "biomimetic AI technology" mumbo-jumbo.

tl;dr: NAM is (quite) more DSP/CPU intensive than a component model. Other capture solutions are the opposite.

Nano models do require less CPU but fidelity goes out the window on these profiles. They are good for boost pedals though but wouldn't use them for amps

 

[bikescene]

I would be all for NAM playback on Line 6 hardware. I’ve only got an HX Stomp now, but I’d definitely have to consider the next gen flagship.

I’ve captured a fair amount of my equipment in NAM, and it’s be nice to have that and the next gen Helix Native all under one hood.

 

[EOengineer]

Nano models do require less CPU but fidelity goes out the window on these profiles. They are good for boost pedals though but wouldn't use them for amps

So is Tonex using nano (my brain immediately jumps to the command line editor here lol) for its lower quality captures? I’ve noticed little to nil difference between the two quality levels across my Mark III and 2204 capture attempts. I swear my ears work, and this is on a pair of ADAM S2A. Maybe I should do another comparison to double check my memory.

 

[EOengineer]

In a sense, you could say the dust has settled at least as far as the quality of the profiles and their performance hit "out of the box" which is probably why we're starting to see hints of what may be coming next.

Unacceptable. Everything near my fat white body needs to get constantly more betterer or I’m going to be left here to work on my personality.

 

[Digital Igloo]

Pretty sure, Eric, mentioned L6 would not be doing what anyone else
has done in that recent interview with them 2 P & W guys whose names
I can't recall, or probably never learned in the first place. I took it to mean
profiles/captures.

More like we won't copy someone else and then falsely claim we invented it. Fractal did IR loading before we did and at no point have we tried to convince people we did it first.

 

[2dor]

More like we won't copy someone else and then falsely claim we invented it. Fractal did IR loading before we did and at no point have we tried to convince people we did it first.

This post has literally put a big smile on my face. Thank you for being awesome.

 

[EOengineer]

 

[texhex]

More like we won't copy someone else and then falsely claim we invented it. Fractal did IR loading before we did and at no point have we tried to convince people we did it first.

If Line 6 creates a capture method, just know that it's going to be put to this level of scrutiny if not higher. It would make sense that the Line 6 folks do something like this level of comparison in house before shipping it.