Line 6 Helix Stadium Pre-Launch Discussion

[Karl Houseknecht]

We've already been hinted that Proxy is machine-learning based - i.e. closer to NAM than what Kemper does.

The learning part doesn’t make the “model” part leaner.

 

[Karl Houseknecht]

Because a proxy or “clone” is like a “single representation” of that amp with a specific set of parameters dialed in. It’s like it’s “static”. But the model has to be able to represent a myriad of parameter changes on the fly. It’s “dynamic”.

Which one would require more processing power? Representing an amp at one particular setting? Or a model that has to compute and represent thousands of permutations of those possible settings?

Neither. They’d be the same, because there would be one or more base amplifier models that would be “dialed” in by the profiling process. It’s just that those base models wouldn’t have knobs.

 

[Lysander]

The learning part doesn’t make the “model” part leaner.

It means there's no model running underneath, which is the way Kemper operates. All in all, static profiles/captures are normally less DSP-demanding than models.

 

[paisleywookiee]

Neither. They’d be the same, because there would be one or more base amplifier models that would be “dialed” in by the profiling process. It’s just that those base models wouldn’t have knobs.

That’s not how NAM works, AFAIK. It’s not a base model, with an EQ match, like the Kemper is.

 

[Dom Mcsweeney]

You can't say one is more or less powerful than the other (white box vs black box).
It depends on how complex one or the other is.
For example, a Tonex model requires a lot less CPU power to run (approximately comparable to a Nano sized NAM model), compared to a Standard NAM model. Same goes with white box component based amp models and how complex they are.

 

[Mongillo19]

lmao. Literally sounds like a Boss TR-2.

Not sure if that's an insult or.....

 

[Sascha Franck]

I really wish people stop bringing up NAM to the Stadium discussion like, every other week

L6 has already clearly explained their stance on NAM for the platform. And 99% of users won't give a shit.

Personally, being pretty much perfectly happy with the current lineup of amps, also expecting things to only get better, I can safely say that, apart from some (obvious) curiosity, all the capturing biz doesn't even make it in my Top 10 (or even 20) of things I'd like to see. IOW, the presence of capturing tech in a unit such as the Stadium, would have exactly zero impact on my purchasing incentives.
Right now, with quite some great captures kinda sorted and ready to roll for DAW usage, whenever I'm working with plugins, I still keep resorting to HX Native (or maybe Amplitube in 5% of all cases).
I can understand that some people may want some outlier amps which likely will never show up as component modeled creatures, and I could as well imagine creating some captures of wild experiments (amps with weird gain structures for example), should that ever be possible - but for daily use, my interest dwindled towards zero quite a while ago already.

 

[Frodebro]

Personally, being pretty much perfectly happy with the current lineup of amps, also expecting things to only get better, I can safely say that, apart from some (obvious) curiosity, all the capturing biz doesn't even make it in my Top 10 (or even 20) of things I'd like to see. IOW, the presence of capturing tech in a unit such as the Stadium, would have exactly zero impact on my purchasing incentives.
Right now, with quite some great captures kinda sorted and ready to roll for DAW usage, whenever I'm working with plugins, I still keep resorting to HX Native (or maybe Amplitube in 5% of all cases).
I can understand that some people may want some outlier amps which likely will never show up as component modeled creatures, and I could as well imagine creating some captures of wild experiments (amps with weird gain structures for example), should that ever be possible - but for daily use, my interest dwindled towards zero quite a while ago already.

Couldn’t agree more. I think the hype surrounding NAM (and captures/profiling in general) is far greater than its actual usefulness outside of a very narrow set of circumstances.

 

[Chocol8]

Is this because you're thinking like a customer that likes Tone-X?

I also use NAM, and have used Kemper, so no, I am a customer who likes captures but doesn't really care about the underlying version as long as it sounds good.

If I was thinking about it selfishly, I would say Line6 should have native Tonex support so it could use all my current captures. That would not make any sense o for L6 either unless IK let them use the tech SUPER cheap.

No one is buying a Stadium because it has their favorite existing capture format native.

 

[holoholo]

I really wish people stop bringing up NAM to the Stadium discussion like, every other week

L6 has already clearly explained their stance on NAM for the platform. And 99% of users won't give a shit.

Capture players are the crossfitters of the guitar world

 

[Mooselake]

I'm curious about Proxy, neutral on the whole NAM thing. I don't know how I would use it (Proxy) unless you can do goofy things like capture Agoura Amps or even instances of whole presets and collapse them into blocks that use less DSP. Seems like there might be some fun ways to play around with it outside the box. But I don't see myself ever capturing the two amps I have.

I've been hoping for years L6 will do a Marshal 4140 model. I noticed Tonex did a capture in their Alex Lifeson product. Some day I might get desperate enough to buy that and then try to proxy it, lol.

What I thought was the topic of NAM support was still an open question for Line 6. Won't be there at launch but might be some day (or not). Has that changed?

 

[paisleywookiee]

Capture players are the crossfitters of the guitar world

Better than being the Jehovah Witnesses of the guitar world. “Excuse me, Brother. Can I talk to you about NAM?”

 

[laxu]

Better than being the Jehovah Witnesses of the guitar world. “Excuse me, Brother. Can I talk to you about NAM?”

Could be worse.

 

[Fireproof]

Neither. They’d be the same, because there would be one or more base amplifier models that would be “dialed” in by the profiling process. It’s just that those base models wouldn’t have knobs.

there’s a big difference in the profiling process and the profiling result.

 

[Dom Mcsweeney]

It would be very cool if Proxy could internally self-capture (no AD/DA), that way you could perhaps make an internal snapshot of a dual amp setup you've dialled in, then load that into one block. This may then help reduce *CPU cost compared to using 2x Agoura amp models, *if it worked out that 1x proxy model was less computationally expensive than 2x Agoura amps.

 

[Sascha Franck]

It would be very cool if Proxy could internally self-capture (no AD/DA), that way you could perhaps make an internal snapshot of a dual amp setup you've dialled in, then load that into one block. This may then help reduce *CPU cost compared to using 2x Agoura amp models, *if it worked out that 1x proxy model was less computationally expensive than 2x Agoura amps.

I'd be interested in that kinda thing - but not necessarily to save CPU cycles but rather to combine amps in wicked fashions. I'm sure I mentioned something along those lines already, but you could come up with a crossover, send the lower ended part into a cleaner amp and all things trebly into a driven one. Would give you tight low end and singing highs. Or so. Combining that into a single capture would make housekeeping easier.
I do however think that current capturing algorithms won't do that kinda stuff too well.
Hm, I could in fact check that out myself...

 

[Dom Mcsweeney]

Exactly. I'd love to have this kind of flexibility built into proxy.
I do similar things quite often with NAM because using a DAW makes this trivial, not just for NAMs but also IRs.
So much fun =)

 

[Karl Houseknecht]

It means there's no model running underneath, which is the way Kemper operates. All in all, static profiles/captures are normally less DSP-demanding than models.

So what emulates the preamp stages or the poweramp or the transformers? Is it just magic? EQ doesn’t create distortion.

 

[GTR37]

Less icky than bullet-pointing features to drum up pre-orders and sales as part of what essentially amounts to a kickstarter, and then never delivering said features.

I don’t know who would even do such a thing ?

 

[Lysander]

So what emulates the preamp stages or the poweramp or the transformers? Is it just magic? EQ doesn’t create distortion.

tl;dr Kemper uses a limited number (7, IIRC) of intenal base amps models; the profiling process involves multiple rounds of error reduction, adjusting model parameters until they best match the source signal. Quad Cortex implements a similar, more refined apprach of this same idea.

NAM, ToneX, Proxy and other use machine learning to train a neural network (normally, WaveNet). These are much more expensive to train, but can actually be made to use less resources than models at playback time.