NAM: Neural Amp Modeler

James Freeman said:
Oh you mean when capturing a real amp, yeah I understand analog-to-digital gain staging well.
I was capturing Helix Native for testing which is completely in the digital domain in 32bit floating, quantization noise is irrelevant there.
Click to expand...
No, I mean even when capturing an amp sim. Yes helix native runs in 32 or 64 bit floating point, but I assume when you bounce and/or export the output track it is 24 bit fixed point, and that's where it's relevant.

EDIT: if previously you meant you normalized the volume before bouncing or exporting the track then that's ok
 
BenIfin said:
Very true ..... I also think that the IK tech's who built Tonex are also going to continually refine and improve it to.

And if I could attempt some hopefully humorous sarcasm for a moment ..... this "open source" argument is also the exact same argument Linux users were saying and doing 15 years ago when they were all saying Linux would destroy Windows in a matter of no time due to its massive open-source user developer base .... I'd imagine the NAM open-source-user-base is a bit smaller than Linux's ;)

All the best,
Ben
Click to expand...
Exactly. There’s obviously comparisons to be made, but ToneX is a product that will receive updates and support - I’m sure it’ll improve (as far as accuracy and training times etc) but likely at a slower speed because of the need for it to be reliable as a paid product.

NAM doesn’t have the same constraints, but as a result may be more volatile or vary in quality. That said, anything that pushes things forward and keeps IK (and other companies) on their heels is only a good thing for users.

As much as I like ToneX, I’d really like to see other companies try their hand at implementing this stuff, which I think is inevitable. IK don’t have a good track record for user interfaces, I don’t really like how they charge their customers upgrades or how they handle certain aspects of their business. That’s taking nothing away from the quality of their technology - I own almost all the software that they make and use it regularly. I just feel like they somehow manage to shoot themselves in the foot rather than reaching their full potential.
 
DLC86 said:
Just to be clear, saying these things is not the equivalent to bashing tonex and IK multimedia, I'm still pretty impressed by what they're able to achieve, especially compared to older profilers/capture devices, it's just that I'm even more impressed by nam and tonex definitely has room for improvements, considering they probably run the same base code too.

PS: and just for the record, I looked at RTA graphs for the first time now so I don't think my brain is tricking me into "making me hear what I see", previously I spotted those differences just by ear and actually I expected to find bigger differences honestly.
Click to expand...

Its all good :)

I can only put it like this ...... take, say, 10 x well done Tonex Captures -and- take 10 x well done NAM Capture - each [obviously] of the same Amps, Settings etc.... etc...

My genuine question remains:-

=> without resorting to spectral analysis or any other kind of non-human-ear-analysis ...... do you believe that in an A/B playback or play-through test, you would, with any degree of certainty or accuracy, be able to tell which is which from each other (?) ..... or indeed from the "real" Amps, Settings etc.... (?)

My answer is I never would. And I also don't think anyone anywhere on the planet would be able to do it either.

All the best,
Ben
 
BenIfin said:
Its all good :)

I can only put it like this ...... take, say, 10 x well done Tonex Captures -and- take 10 x well done NAM Capture - each [obviously] of the same Amps, Settings etc.... etc...

My genuine question remains:-

=> without resorting to spectral analysis or any other kind of non-human-ear-analysis ...... do you believe that in an A/B playback or play-through test, you would, with any degree of certainty or accuracy, be able to tell which is which from each other (?) ..... or indeed from the "real" Amps, Settings etc.... (?)

My answer is I never would. And I also don't think anyone anywhere on the planet would be able to do it either.

All the best,
Ben
Click to expand...
I think I've already answered here:

DLC86 said:
PS: and just for the record, I looked at RTA graphs for the first time now so I don't think my brain is tricking me into "making me hear what I see", previously I spotted those differences just by ear and actually I expected to find bigger differences honestly.
Click to expand...
So yes, when doing an A/B while playing easily, by just listening to the recording maybe less easily but still..
And I also think a lot of folks would be able to tell the difference especially while playing, I have no special ear.

If you want try them both and judge for yourself, you can find both my VH4 NAM and Tonex captures here (pick the V2 files):

NAM Captures - DLC86 – Google Drive

drive.google.com drive.google.com

TONEX Captures - DLC86 – Google Drive

drive.google.com drive.google.com
 
DLC86 said:
I think I've already answered here:


So yes, when doing an A/B while playing easily, by just listening to the recording maybe less easily but still..
And I also think a lot of folks would be able to tell the difference especially while playing, I have no special ear.

If you want try them both and judge for yourself, you can find both my VH4 NAM and Tonex captures here (pick the V2 files):

NAM Captures - DLC86 – Google Drive

drive.google.com drive.google.com

TONEX Captures - DLC86 – Google Drive

drive.google.com drive.google.com
Click to expand...

Real JCM 800 Amp vs Tonex ..... you've probably already listened to this video so its probably pointless as it wouldn't be a blind unknown A/B test ..... but you think "a lot of folks" would be able to "hear" the difference in the mix or isolated track part of this video [note the reveal is in the isolated track part] ? .... and more-so ... if it were done with 10 x different Real Amps and 10 x Tonex's .... in an unknown blind A/B ..... you / "a lot of folks" could "hear-tell" them apart with consistent accuracy too ?

Am only clarifying as I *think* that's what you're saying, but I don't want to put words or assertions "in your post" as they say.

Ben
 
Last edited:
While it's nice if one tool is more accurate than another, there's a ton of people happily using less accurate tools. We are pretty much splitting hairs here and that's academically interesting but not relevant to playing or recording guitar.

I just listened to @DLC86's Bassguy Bright capture comparison and I can hear slight differences there - also finding the Tonex in that case to produce a better result, but overall if I had any of those tones I would not be unhappy. It's a case of "different, not worse" really.

Maybe IK should consider hiring the NAM guy to tweak their ML setup to perfection?
 
BenIfin said:
Real JCM 800 Amp vs Tonex ..... you've probably already listened to this video so its probably pointless as it wouldn't be a blind unknown A/B test ..... but you think "a lot of folks" would be able to "hear" the difference in the mix or isolated track part of this video [note the reveal is in the isolated track part] ? .... and more-so ... if it were done with 10 x different Real Amps and 10 x Tonex's .... in an unknown blind A/B ..... you / "a lot of folks" could "hear-tell" them apart with consistent accuracy too ?

Am only clarifying as I *think* that's what you're saying, but I don't want to put words or assertions "in your post" as they say.

Ben
Click to expand...
Yes, I've already seen that video, and in the isolated comparison I can definitely hear the differences on the high and low ends, in the mix nope.
I don't know if I'm able to always tell them apart, as I said I'm not claiming they're worlds apart or that I have golden ears, all I am saying is that in all my captures and in a lot of video I've seen I always hear that slight high frequencies roll-off and "rounder" bass, and the RTA shows that differences, no more no less.
Does it mean that tonex is useless for me? No, in a mix I probably couldn't be able to tell the difference. But since NAM produces more accurate results with less training time and no effort to match the gain, why shouldn't I prefer to use or suggest others using that? It's also free!
 
DLC86 said:
Yes, I've already seen that video, and in the isolated comparison I can definitely hear the differences on the high and low ends, in the mix nope.
I don't know if I'm able to always tell them apart, as I said I'm not claiming they're worlds apart or that I have golden ears, all I am saying is that in all my captures and in a lot of video I've seen I always hear that slight high frequencies roll-off and "rounder" bass, and the RTA shows that differences, no more no less.
Click to expand...
DLC86 said:
Does it mean that tonex is useless for me? No, in a mix I probably couldn't be able to tell the difference. But since NAM produces [extremely marginally] more accurate results with [yes] much less training time and [no effort (?) (!)] to match the gain - [totally disagree and can find no objective tested and demonstrated evidence anywhere to support this opinion] why shouldn't I prefer to use or suggest others using that? [of course you should - no issues at all there] It's also free! [yep, excellent]
Click to expand...

^^ [My] comments in [red] above.

Its clear to me that we will just have to agree to disagree, so I'll try and stay out of this thread from now on.

Ben
 
BenIfin said:
[totally disagree and can find no objective tested and demonstrated evidence anywhere to support this opinion]
Click to expand...
That's not an opinion, it's a fact, and I can also prove it if all video reviews showing that you need to adjust it after the capturing process aren't enough.
Tonex is totally unreliable in matching the gain, if it wasn't why would they put a gain adjustment after the training at all (as do QC and kemper too, btw)?
And do you always set that knob the same way and always obtain exactly the same amount of gain as the target amp?

PS: maybe they made it so the gain matching is accurate only thru their Tonex Capture box (even if that would be quite silly)? Do you have one of those?

PPS: nope, I don't think that's the case, just look at this video from min 11:35. Both the gain difference and the high freq roll-off are pretty evident to me and he's using the tonex capture box.

 
Last edited:
This is a neat concept, but IMO it's going to be more important to have a widely adopted platform for creating/sharing captures, and to have hardware integration. That's where ToneX seems to have a huge advantage (based on price being substantially lower than Kemper).
 
Found the problem.

NAM Plugin doesn't like when the track is set to Stereo, so set it to Mono before you load the Plugin and the bug will be gone.

mono.png
 
James Freeman said:
NAM Plugin v0.7.0 has been released.

Can anyone try to reproduce this bug I reported?

github.com

[BUG] Cakewalk (DAW) - Plugin goes silent on stereo tracks · Issue #102 · sdatkinson/NeuralAmpModelerPlugin

Describe the bug The VST (NAM Loader) stops processing sound and behaves like it is bypassed. All the UI controls stop responding EXCEPT input knob and input meter. To Reproduce Add NAM Loader to a...
github.com github.com
Click to expand...
I have this issue depending on what plugins are loaded prior to nam being inserted.
 
Deadpan said:
I have this issue depending on what plugins are loaded prior to nam being inserted.
Click to expand...
Yes, if other plugins in the fx rack are forced internally to Stereo, even if the track is set to Mono, NAM Loader will stop processing sound.
For now, set the track to Mono and don't put anything in the fx rack with NAM, use a Bus for everything else like IR loaders.

It is very buggy, I wouldn't use NAM for actual projects.
 
Last edited:
As if there’s no way a “real” amp
Could be recorded and end up with less low end than a capture player plays.
 
zib said:
A NAM amp is around 500ko. Just saying (courtesy of James Freeman).
Click to expand...

I think there is an option to lighten the capture to run on a DSP.

@DLC86
Did you try the 'Light' and 'Feather' options?
For some reason I'm having trouble with Google Colab, I can't upload files.
 
  • Like
Reactions: zib
James Freeman said:
I think there is an option to lighten the capture to run on a DSP.

@DLC86
Did you try the 'Light' and 'Feather' options?
For some reason I'm having trouble with Google Colab, I can't upload files.
Click to expand...
Nope, I run the training on my laptop with an RTX3070, never used coogle colab
 
I'm also on RTX3070 but I don't want to install 5GB of software I will never use just to train a handful of models.
 
James Freeman said:
I'm also on RTX3070 but I don't want to install 5GB of software I will never use just to train a handful of models.
Click to expand...
Well, actually it's more than 5GB, the Anaconda folder alone takes 17 GB on my hard disk 😅
But I think it's worth it since the training is much faster than google colab with a 3070
 
You must log in or register to reply here.
Back
Top