SoOoooN! [NDSP Morgan Amps Suite]

[MirrorProfiles]

I think they started releasing plugins around 2018? Looks like their first mention of machine learning was with Nolly's tape saturation, 2019. Then around 2020 with the QC, they started mentioning machine learning a lot and how it was being incorporated into their plugins. Machine learning stuff wasn't in their first products and wasn't what drew people into their products.

Clue is in their name. I’m pretty sure you can find the academic papers from the founders on how they do it, although I’m sure it’s evolved a lot since (they’ve posted patents for robots to adjust knob positions during capturing). I think some of the pedals and FX may be algo based but generally their bread and butter is in ML capturing of amps and I don’t think it’s been any different. With hindsight, the oversampling and aliasing points towards that.

Steve from NAM said he landed on something quite similar with his approach just by chance. Mercuriall do some machine learning based stuff for their poweramps but as far as I’m aware Neural have paved the way for these amp models that have full controls. The dev behind Bogren digital has something of his own that’s similar, but obviously a good few years later than NDSP.

 

[mikah912]

I very, very much agree with this.

To be fair, it’s not like I’m trying to create my own, new tone when I write/record, but as I’ve been playing through that Shiva the last couple days and hearing tones I’ve never gone to but thoroughly enjoy, I can definitely see how someone’s own tone can take shape in that way, just discovering.

It’s definitely going to have an effect on how I dial in models. Previous to getting this amp I’d dial in a model based on how I thought it was supposed to sound, rather than exploring what sounds it were capable of. There’s no forcing a tone to happen when you’ve got a physical amp and one cab, all you can do is see what it offers and I do believe I need to make my preset dialing-in more in line with that.

I think a lot of it is the hyper-subgenre categorizing of guitar music these days. Each of these micro-genres kinda has a prototypical guitar tone associated with it. Straying from that path will get your contemporaries to say that your tone doesn't "right".

So if you play P&W, the "correct" tone is some sort of EL84/Vox-ish amp with a good amount of reverb and delay, plus a clean boost overdrive to kick on for solos.

You play rock/hard rock, the "correct" tone for, like, forever should be a cranked Marshall or variant, probably modded for extra gain into Greenbacks or V30s.

You play modern metal/djent/prog, you "need" to have something based on a 5150 or Recto with a drive in front going into V30s.

There were just a lot less "instructions" and waaaaay less resources when the signature tones of the past happened. I still hear singular tone guitar players these days (e.g. Julian Lage, Josh Homme), but they're outliers that just prove the rule.

 

[MirrorProfiles]

Studio Devil, S-Gear, etc. had the format already... NDSP put a pretty modern skin on it.

I mean, compare these:

to these (which ALMOST look like they could from the same company):

 

[mikah912]

Honestly, I think the most direct predecessor to the Neural look and feel was Overloud TH-U, but they just lost prominence amid slowed-down development and increased competition.

 

[MirrorProfiles]

Neural absolutely nailed that "first impression" aspect of their plugins. You load it up for the first time - it looks the part, you have to do zero work to hear it as intended, its simple+intuitive. Its all basic stuff we take for granted but they clearly put a lot of thought and focus on getting those things right. They could probably save a TON of money by doing unlicensed sims but the way it looks and behaves has a tremendous impact on how we use it. The same is true all over the plugin world, particularly for analog modelled stuff. I dont think Neural were the first to do it like this, but they got the formula right for what they offer (its a sum of all the parts).

 

[Whizzinby]

When you consider how difficult it was for many to record at all until relatively recently, I wouldn't undersell how cool it is to have high quality easily recordable tones right at your fingertips. For non-pros simplicity and ease of use is a lot of more important than trying mine for truly distinctive tones.

I think Neural, as a plug-in company, killed it. They have serious issues with market saturation now, given how many offerings they have available, but I think they’ve done a really good job of creating instantly accessible tones in an easy to use format. Can’t knock that. IMO

 

[GTR37]

Honestly, I think the most direct predecessor to the Neural look and feel was Overloud TH-U, but they just lost prominence amid slowed-down development and increased competition.

That was a weird one DV Mark Multiamp was the hardware version
And apparently Overloud was partnered and bundled in there
Then Overloud went plug in only and DV never released the new multiamp and abandoned it

 

[James Freeman]

 

[yeky83]

Clue is in their name. I’m pretty sure you can find the academic papers from the founders on how they do it, although I’m sure it’s evolved a lot since (they’ve posted patents for robots to adjust knob positions during capturing). I think some of the pedals and FX may be algo based but generally their bread and butter is in ML capturing of amps and I don’t think it’s been any different.

No, they've said in interviews that their first plugins were without ML. And their marketing reflected that. So again... machine learning stuff wasn't in their first products and wasn't what drew people into their products.

I mean, compare these:

View attachment 16018
View attachment 16019

to these (which ALMOST look like they could from the same company):
View attachment 16020


View attachment 16022View attachment 16024View attachment 16023View attachment 16021

Yes, they made it really pretty and others followed. But the "'pre-routed standalone app' format" existed beforehand.

 

[Whizzinby]

No, they've said in interviews that their first plugins were without ML. And their marketing reflected that. So again... machine learning stuff wasn't in their first products and wasn't what drew people into their products.

Yes, they made it really pretty and others followed. But the "'pre-routed standalone app' format" existed beforehand.

Yeah we are probably thinking of the same interview where Doug seemed to indicate they realized that the only way they’d be able to model fast enough to realize the QC was to turn to the machine learning approach.

They mentioned at TOP that the Gojira plug was the first to be modeled by the same processes as those on the QC, but I don’t know if that meant it was the first plug to use ML or not. (Or if they used different variations of ML before that on plugs)

 

[MirrorProfiles]

Yeah we are probably thinking of the same interview where Doug seemed to indicate they realized that the only way they’d be able to model fast enough to realize the QC was to turn to the machine learning approach.

They mentioned at TOP that the Gojira plug was the first to be modeled by the same processes as those on the QC, but I don’t know if that meant it was the first plug to use ML or not. (Or if they used different variaruins of ML before that on plugs)

I could be wrong but I seem to remember them mentioning machine learning and neural networks well before this, and the AATI update with gojira was more to do with the capturing process being automated and more efficient when dealing with sample rates.

 

[Mongillo19]

Remember when they said amp models can take as short as 8 hours (or so)

 

[Whizzinby]

I could be wrong but I seem to remember them mentioning machine learning and neural networks well before this, and the AATI update with gojira was more to do with the capturing process being automated and more efficient when dealing with sample rates.

Totally possible. They mentioned they unified their approach on Gojira, not necessarily that it was the first plug to use ML.

But Doug did indicate they had to pivot early to ML when they realized they didn’t have the ability to model fast enough with traditional methods.

 

[Whizzinby]

Remember when they said amp models can take as short as 8 hours (or so)

Ahh, the good ole days.

 

[mikah912]

Remember when they said amp models can take as short as 8 hours (or so)

That was accurate, actually.

They just forgot to mention that "performance issues" can delay them by, uh, going on 3 years now.....

 

[MirrorProfiles]

quite an interesting interview with one of Neural’s machine learning guys

 

[TubeStack]

quite an interesting interview with one of Neural’s machine learning guys

That Explorer is .

 

[Boudoir Guitar]

Yes, they made it really pretty and others followed. But the "'pre-routed standalone app' format" existed beforehand.

I think there's still a bit of a gap between something like S-Gear ("we don't need to offer a million amps; most guitar players, can cover most sounds they need with a plugin that has a tweed, a full Mark, a marshall, a soldano, and a Dumble; add in some classic basic effects and call it good") and the Archetype stuff ("a plugin that is designed to give one set of sounds to one specific group of guitar players; nothing more and nothing less.")

 

[laxu]

I think we've all said it before. Generally speaking, NDSP (plugins and hardware) do reach professional quality in terms of tone.

I just don't think they could have done any worse in the PR/support department. If they just wouldn't even had a social media presence I think the company wouldnt be as despised by some. In the end they did it to themselves

I'd say their aggressive marketing is a huge factor in their success. We, as guitarists who are already keeping track of what's going on in the digital modeling world, of course really hate it because it's like a stubborn salesman knocking on your door constantly.

But if you look at e.g Reddit's guitar subreddits where the demographic skews younger with a lot of newer players, they are the target audience. They are the ones buying their first Focusrite and trying some NeuralDSP plugins. They are the ones who see the QC ads on Instagram, Facebook and probably Tiktok and then start asking "should I buy a QC?"

I have seen plenty of random NDSP ads, but never an ad on social media for Line6, Kemper or Fractal. They of course have a lot of word-of-mouth promotion from users, but that's not going to end up in the ears of someone new to the whole thing.

Finnish companies are often absolutely terrible at promoting their products (how many Finnish companies can you name besides Nokia?), but NeuralDSP has clearly done a lot of things right in that department. Even if we hate it.

 

[DrewJD82]

I'd say their aggressive marketing is a huge factor in their success. We, as guitarists who are already keeping track of what's going on in the digital modeling world, of course really hate it because it's like a stubborn salesman knocking on your door constantly.

But if you look at e.g Reddit's guitar subreddits where the demographic skews younger with a lot of newer players, they are the target audience. They are the ones buying their first Focusrite and trying some NeuralDSP plugins. They are the ones who see the QC ads on Instagram, Facebook and probably Tiktok and then start asking "should I buy a QC?"

I have seen plenty of random NDSP ads, but never an ad on social media for Line6, Kemper or Fractal. They of course have a lot of word-of-mouth promotion from users, but that's not going to end up in the ears of someone new to the whole thing.

Finnish companies are often absolutely terrible at promoting their products (how many Finnish companies can you name besides Nokia?), but NeuralDSP has clearly done a lot of things right in that department. Even if we hate it.

Hahahah it’s not just us that hate it-

Soon (NDSP Teaser)

New plugin? Plugins on QC???

unity.neuraldsp.com