Line 6's stance on profiling

[Deadpan]

In real life each 12ax7 or 12at7 or...on and on can sound different. In some amp circuits not as much as others but in some it is night and day.

Recently I had been a bit disappointed with my Matchless HC30, was swapping preamp and power tubes and nothing was bringing the joy this amp is capable of. Swapped the rectifier tubes and bam, the glory returned.

Set in stone models can work or they don't but give me the option to make the best and you will win.

 

[pipelineaudio]

In the simplest case, for distortion in preamp tubes, I would guess that each vendor has their own mathmatical model of tube behaviour for a given brand and sample set of data they've taken for say a 12AX7. They model the effects of that tube on various imput signals to get the expected output signals, including distortion.

All of the vendor's input date will be in the same ballpark if they've measured correctly and have a good sample size of tubes. There may be variations there though. Then they will have slightly different approaches in how they model the behaviour. And that's just preamp tubes.

…

So whilst each company is trying to achieve the same thing, they all have different approaches that could manifest in a signature sound

I think you are probably onto it. In our case, where I can get our compressor to mimic my favorite Bias, Lepou and TSE amp plugins to a very high degree, we are using nearly the simplest possible transfer function for the soft limiter that I’m abusing to create the amplifier sound.

Nearly this exact same algorithm was around way back when LePou, TSE and early versions of Bias were made. Ours likely isn’t even audibly different than the old algorithms, it’s really just tweaked for stability, performance, denormal protection and aliasing.

 

[Boudoir Guitar]

This thread reminds me of folks that ask for a "model" of the feedbacker pedal, or a "model" of Strymon cloud verbs. It's basically asking "hey, can you guys copy the code of this competing product?". Put another way "can you do what Mooer apparently did to EHX?!?".

It's one thing to get a modern digital device with clean converters and loads of but depth to model the sound of old digital delays - that's still requiring creativity on the model makers part.

It's another to just...want your modern digital device to cannibalize another modern digital device.

 

[pipelineaudio]

Except that this is nothing at all what the original post was asking.

 

[Boudoir Guitar]

Except that this is nothing at all what the original post was asking.

"Why won't Line 6 put a Kemper in my Helix so I can copy Bias"

 

[James Freeman]

Interestingly, Ben Adrian actually mentioned they also wanted a quicker way to model amps. Whether that would entail using machine learning or not, he couldn’t say.
But it does imply that Line 6 also wants to be able to deliver amp and effects models faster.

Yep I've watched the interview and he did mention several way they experimented with to quicken the modeling process.
It doesn't have to be AI or Machine Learning, partial automation in any step of the process will shorten the procedure significantly.

I still think component based modeling tops any capture/profiling because the latter doesn't and can't capture the idiosyncrasies of the power supply and poweramp interaction.
Capturing excels at 'well behaved' preamp type of tones but it's not what I consider as an accurate and complete amp model, because it's not.

 

[Jarick]

Here's my take...

I'm intrigued by the ability to expand amp tones to less common models...it would be awesome to play through an obscure amp like the Demeter TGP-3 (Dean Deleo) or a Bogner modded Marshall like the Snorkler (Jerry Cantrell). But with profiling you're at the mercy of whoever captured the amp and if you prefer their settings and speaker/mic choices. And then on the flip side, you have like 10,000 different profiles of Plexis out there, and 95% of them suck. And the time and money expense of trying thousands of profiles.

I don't want a copy of Kemper's profiling tech. It sounds great when done well, but it's so hit or miss for me especially on the cab side. And you have the limitations in terms of not having the gain or EQ adjustments that match the amp. I'd be interested if it's some kind of hybrid where you can capture the vibe of an amp but still be able to change out the cab and adjust gain/EQ without everything getting sonically weird.

I still think the biggest problem to solve is to be able to quickly and easily dial in cab modeling. That seems to impact the sound and feel possibly more than anything else.

 

[Lysander]

I need something clarified, but a profile is static right? It's a shot of an amp at particular settings? Now I realise there's a tone stack in the kemper etc, but those won't behave in the way the profiled amp will?

Right. You get tone and drive controls on captures/profiles, but these don't interact with the sound like the real deal - the EQ is applied to the output, and drive is commonly boosting the input signal.

It's not like profiling doesn't have its place in the guitar world, but i genuinely don't get why people demand L6 to release a profiling product.

 

[Bruce]

For my money give me a good model over a profile any day.

 

[Digital Igloo]

This is NOT a dichotomy

I see the arguments for why they may not want to bother, or weird ones regarding not stepping on kemper's toes, but it seems like this is a road every system is going to need as at least an option going into the future.

I totally get how some customers want everything, but why is it weird for Line 6 to do our own thing instead of ripping off Kemper's IP?

For the umpteenth time, we're not against machine-learning based capture—we very well may end up going there—we're just not going to pretend we invented it. Just like with Helix's IR loading back in 2015: "Here, load a bunch of IRs. We don't deserve a cookie or anything."

 

[Fireproof]

I totally get how some customers want everything, but why is it weird for Line 6 to do out own thing instead of ripping off Kemper's IP?

For the umpteenth time, we're not against machine-learning based capture—we very well may end up going there—we're just not going to pretend we invented it. Just like with Helix's IR loading back in 2015: "Here, load a bunch of IRs. We don't deserve a cookie or anything."

That’s very fair!

 

[pipelineaudio]

*hopefully people can keep their feelings in check and concentrate on the techy stuff in here****


Thats why I said its not a dichotomy, and I don't think anyone asked Line6 to rip off Kemper's IP

I had the IUD and "threshold stepper" convolution out in a decade before Kemper was even a thing. Was Kemper ripping me off? No. I don't even know how they are doing their thing, so its not like theirs is the only way

All that aside, @Digital Igloo , I'm pretty much with a lot of others here that modeling is probably far more ideal than profiling. In the same way that an algo reverb could be so much more powerfully diverse than straight convolution reverb. So the big question for me is, why am I able to so easily approximate certain modeling tones with *some* other products and not with Helix and *some* others?

I am assuming user error, but with Helix and Overloud, there is always a certain different character to the distortion or and I hate to say it as its pretty much a nonsense term, but "hardness" in those two products which stands out regardless of the models

 

[pipelineaudio]

I have another possible wrench in the works I'm just remembering now:

I was unable with Audio Assault's amp sims to duplicate another plugin's tone and feel, but Alex at AA was able to take my sample and send me a preset (not sure if he will say its "profiling" of any sort) that was identical to my ears

SO, very big possibility of user error, but it could have been "profiling"

 

[the swede]

My view is that there is no reason for them to try to add a capturing/profiling feature into Helix as of today. If they include something like that in the next series of hardware products it would be because then it has become a thing that is expected to be included in all modelers.
And I don’t mean to disrespect the people that want this from Helix today, it’s just that there are good options out there I think.

However, as Ben said in the interview, the reasons for going into machine learning/capturing may be of greater significance for the modeling design team itself regarding the time and work they put into it. They are 3 people and occasionally 1 that helps out, so everything they can do to make it a better and faster process is good in the end. That is, we maybe will se a point where component “by hand” modeling combined with machine learning and profiling is the way they work. But i don’t know.

 

[James Freeman]

ripping off Kemper's IP? ..... —we very well may end up going there—

What context?

Captures:

 

[pipelineaudio]

you are a very very bad man! :)

 

[James Freeman]

If there is a post that deserves a clown emoji, that was it.

 

[SeeD]

When i read that a company don`t whant to step on another company`s `thing` how do L6 think about all the amps they make a 1: 1 copy of, or alot of the effects ??
Does this only aply to amps and effects ( wild west) and if someone copy a nice UI in a digital space ( a few things) it`s okey to freek out ?
Just thinking out loud

 

[Boudoir Guitar]

When i read that a company don`t whant to step on another company`s `thing` how do L6 think about all the amps they make a 1: 1 copy of, or alot of the effects ??
Does this only aply to amps and effects ( wild west) and if someone copy a nice UI in a digital space ( a few things) it`s okey to freek out ?
Just thinking out loud

1:1?!?!? I kid, I kid...

 

[sleewell]

I don't really see the need. I don't have access to enough amps to have a huge desire to profile and when I had the kemper i hated sifting though all the crap on the exchange to find a few good ones. I think line 6 does a great job of giving us the amps that are needed in a super polished form.