NAM: Neural Amp Modeler

BenIfin said:
Quick *unexpected* update

In my efforts to get the VST3 working in Cakewalk Bandlab ..... I uninstalled and reinstalled 4 times ... and it finally worked for about 25 minutes before it crashed / locked up Cakewalk Bandlab ... needing a hard re-boot ... cant remember the last time I needed to do that :( ..... anyway ....... that's not my point / concern ..... reliability will doubtlessly come with time ....

I recorded some basic guitar .... set the track to Loop and messed with the Controls and found a *very unexpected* thing.

The "input [gain]" seemed to operate well.

But the Bass / Middle / Treble ... but especially the Bass and Middle responded absolutely nothing like ... Tonex ... !!

So what I hear you think ? Why do I mention this ? Who gives a f&ck ?

Well ......because one of the absolutely great things IK have got right with Tonex - and I have no idea how they have done it - is that unlike the KPA ... adjusting the Tonex G/B/M/T/P changes the sound [obviously] but it keeps the core tone and uniqueness of the Amp Capture intact.

This current 0.7.1 VST version [ I've no idea if this is a new EQ Stack being used ? ] ... is quite honestly and genuinely worse than even the KPA implementation ..... even smallish adjustments to the Bass and especially the Mids really made the Capture sound increasingly "out of whack" / very "changed" from the original Capture ... the Treble control is a better but still not "ideal".

The Tonex "eq" implementation is very musical and natural and "seems related" to the specific core Captured sound - you can adjust to pretty much any setting and the sound is still "in the Captures ball-park" ... it almost feels like you are adjusting the actual Amp controls .. even though of course you aren't because its "just" a Post or Pre-Capture EQ.

The Amps too felt surprisingly and unexpectedly "stiff" compared to the Tonex .... but that's another point entirely.

The aliasing on lower gain amps - which are bloody near impossible to find - is really not pretty .... NAM desperately needs some really accurate oversampling a.s.a.p .... but again, that's another point entirely.

I was going to post this in "the other place" and/or on the FB group where I am a "lurking member" under a different name .... but in either place, I have no doubt I would be howled down as a Tonex-shill-fan-boi-NAM-hating-heretic ... such is the "hype and heat" of some / quite a few of the current NAM "evangelists".

Anyway ..... just putting it here f.w.i.w ..... I think Steve ? or someone *really* needs to completely re-do the current NAM EQ Stack so it is guitar frequency focused .... or frankly .... just copy the way IK have done it !

If you have Tonex and NAM .... run each as VST in a track ..... use a similar'ish amp for each and the same IR for each ..... loop-toggle between them whilst messing with the Tonex EQ and then the NAM EQ .... it will be obvious very quickly.

Then again ... maybe my ears a totally f&cked ..... but I doubt it ;) :)

All the best to all,
Ben
Click to expand...
Just wanted to say that your extensive use of italic, underline, bold, and colours makes me nervous 😅
 
BenIfin said:
I recorded some basic guitar .... set the track to Loop and messed with the Controls and found a *very unexpected* thing.
Click to expand...
Surely you can post some of your playing you’ve recorded with Nam and Tonex. To better illustrate your commentary.
 
BenIfin said:
Well ......because one of the absolutely great things IK have got right with Tonex - and I have no idea how they have done it - is that unlike the KPA ... adjusting the Tonex G/B/M/T/P changes the sound [obviously] but it keeps the core tone and uniqueness of the Amp Capture intact.

This current 0.7.1 VST version [ I've no idea if this is a new EQ Stack being used ? ] ... is quite honestly and genuinely worse than even the KPA implementation ..... even smallish adjustments to the Bass and especially the Mids really made the Capture sound increasingly "out of whack" / very "changed" from the original Capture ... the Treble control is a better but still not "ideal"......
Click to expand...

I hate to be this person, but you could fire up reaper to try it out :bag. If you think something is sounding really off, it's possible that there's also something weird going on. As to the EQ - it's actually pretty much brand new as it is haha - there was no eq at all before mid january / 0.6. It behaves a little differently than what people are used to - I think the mid Q changes are narrower cut, wider boost, but I can't find the post right now. In any case, I usually just use them for minor tweaks and do post EQ with parametric personally.

The toneX EQ is actually one of the best I've ever seen put on a plugin like that - and on the hardware pedal too. Not only is it parametric and adjustable, but the ranges on that EQ aren't super big - part of the reason it doesn't seem to "alter" the tone as much. I'd love to see more flexibility like that on NAM.

One *major* suggestion - leave the output knob at 0 or less. There's some clipping that can happen if you try to increase it, on the output. Recommend boosting after the plugin or using the track fader for more volume for now.

BenIfin said:
The aliasing on lower gain amps - which are bloody near impossible to find - is really not pretty .... NAM desperately needs some really accurate oversampling a.s.a.p .... but again, that's another point entirely.

I was going to post this in "the other place" and/or on the FB group where I am a "lurking member" under a different name .... but in either place, I have no doubt I would be howled down as a Tonex-shill-fan-boi-NAM-hating-heretic ... such is the "hype and heat" of some / quite a few of the current NAM "evangelists".
Click to expand...

I know there are folks out there who will evangelize because for whatever reason, that's just the nature of where we are with all this stuff today. However, Steve is 100% open to suggestions and ideas, and we try to keep bickering to a minimum on the FB group when possible to keep the space productive. If you want to avoid all that completely, the github issues page (for the plugin repo) is actually a better place to go with suggestions anyway because it's where Steve pays the most attention. If you can demonstrate what you're referring to wrt aliasing and such, that would actually be a great place to illustrate it.

James Freeman said:
I assume most professional plugins have some kind of sampling rate conversion at the input?
Internal processing can remain at 48kHz or 96kHz if oversampled, but the DAW project sampling rate has to be independent from that.
Click to expand...

That's what this fix will be I think, with internal remaining at 48khz

Sascha Franck said:
Just to make sure: I assume there's no ways to make this run under macOS Mojave, right?
Click to expand...

Right now, there isn't - I think this is because of some dependency related issues on the C++ side and iplug. 10.15 is the min supported version for the moment
 
1680984042898.png


If you're curious about the tonestack, current implementation here:

github.com

Treble/Middle/Bass EQ knobs by sdatkinson · Pull Request #30 · sdatkinson/NeuralAmpModelerPlugin

Resolves #1
github.com github.com
 
BenIfin said:
The aliasing on lower gain amps - which are bloody near impossible to find - is really not pretty .... NAM desperately needs some really accurate oversampling a.s.a.p .... but again, that's another point entirely.
Click to expand...
If you’re hearing what you think is obvious aliasing now, when you couldn’t hear it at all earlier when running sine sweeps through plugins, then what you’re hearing might not be aliasing.

Are you sure you don’t have any gain staging problems causing clipping, sample rate conversion issues in your DAW/interface, or buffer overruns from having too small of a sample buffer in your ASIO driver settings?
 
Hey all !

Genuinely appreciate all the constructive advice and suggestions and recommendations .... I will try and get to them all in the next week or so :) Certainly a breath of fresh air from "the other place" ! Thanks.

I know this is going to sound like "b.s" or a "lie" to some people, but I have no issues at all with NAM as a potential product ..... I'm personally not interested in yet another plugin ..... but if the day comes and -someone- incorporates NAM into a pedal *anything like* the Tonex and I think it sounds and responds better, I'll drop the Tonex in a heartbeat and get that pedal - period. As I've said many times - for better or worse - I have %0 zero brand loyalty to any company - if Joyo or Nux or Donner etc.... release it in a great working format, and it sounds and feels better, I'll be the first in line to buy it and incorporate it into my rig.

An excellent useability feature in NAM I did really like - which I forgot to mention above - was the ability to turn-off the EQ. In Tonex you have to manually reset each EQ parameter to 12.00 to get it back to "original Capture" status. This NAM EQ off/on feature - to my ears at least- seems to allow you to instantly toggle between your tweaked Capture EQ -vs- the original Capture EQ. Its a small thing but extremely useful when tweaking.

Lastly, as a general comment, I have noticed there is a great deal of consternation / anger towards the idea of people making paid-for-captures because NAM is free etc..... however when many people own, tens if not hundreds of $ 000's of Amps and they take the time and care to make and release a pack properly mic'ed up in proper studio environments etc.... ... for say NAM or Tonex, I think it is absolutely right and fair and proper that they charge what they feel is fair .... no-one is forcing anyone to buy them.

Historically speak, Michael Britt almost single handedly exploded the KPA user base with his initial 5 [amazing] packs.

Having "pro" makers the likes of say a Top Jimi or TJ or Michael Britt etc.... "make stuff" for your platform can only be a huge endorsement and can only give it more credibility and help it grow further ... i.m.h.o at least.

All the best to all.
Ben
 
Last edited:
I can step in here, as a total noob on profiling/capturing whatever. I have no amps.

But consider this. All the tech talk aside.

Leo clearly proves to me, just me and my own preferences.

That 90% of what every one is talking about… simply does not matter (to me).

All 5 examples sound good and I could sit and play any of them with the sole intention of playing guitar and having a good time enjoying music, and be totally clueless about what is real or not, better than whatever or worse than whatever. Because the differences (yes I heard differences between all examples) simply does not matter to me.

I already know and decided that profiling/capturing is not for me. I’m happy enough with amp modeling as it is, cheap or expensive. So I can’t really contribute anything to discussions down to details that are discussed here other than offer you all another perspective on things. :)

Your scientific discussions. They are interesting to read, even to me. I appreciate that.

So the difference between Tonex and NAM would come down to other factors for me.

If I had to, the Tonex wins as it’s a hardware pedal that a noob like me actually can use.
 
Okay so I've been testing a few JCM800 captures from ToneHunt.
My experience is there is no gain or sound consistency whatsoever, might as well be two totally random amps, the reason is probably people are capturing at completely arbitrary re-amp levels, moreover, if there's no detailed notes I have no idea what the amp settings were, what load was used, or if it was a real amp at all.
The only captures I 100% trust and know they are accurate are my own.
I still very much prefer modeling because I have full control and know exactly what's going on.
 
the swede said:
I can step in here, as a total noob on profiling/capturing whatever. I have no amps.

But consider this. All the tech talk aside.

Leo clearly proves to me, just me and my own preferences.

That 90% of what every one is talking about… simply does not matter (to me).

All 5 examples sound good and I could sit and play any of them with the sole intention of playing guitar and having a good time enjoying music, and be totally clueless about what is real or not, better than whatever or worse than whatever. Because the differences (yes I heard differences between all examples) simply does not matter to me.

I already know and decided that profiling/capturing is not for me. I’m happy enough with amp modeling as it is, cheap or expensive. So I can’t really contribute anything to discussions down to details that are discussed here other than offer you all another perspective on things. :)

Your scientific discussions. They are interesting to read, even to me. I appreciate that.

So the difference between Tonex and NAM would come down to other factors for me.

If I had to, the Tonex wins as it’s a hardware pedal that a noob like me actually can use.
Click to expand...

Exactly the curse of A/B/C comparisons - they make us place far too much emphasis on things that actually don't really matter in anything but comparisons. Whatever fits your use case the best is the way to go. The real truth is, there is no "best" and you should play what makes you the happiest.
James Freeman said:
Okay so I've been testing a few JCM800 captures from ToneHunt.
My experience is there is no gain or sound consistency whatsoever, might as well be two totally random amps, the reason is probably people are capturing at completely arbitrary re-amp levels, moreover, if there's no detailed notes I have no idea what the amp settings were, what load was used, or if it was a real amp at all.
The only captures I 100% trust and know they are accurate are my own.
I still very much prefer modeling because I have full control and know exactly what's going on.
Click to expand...

Yep, it's gonna vary wildly based on all kinds of things, which mirrors my experience with kemper / tonex / QC. Some people (myself included) try to make collections or packs for a specific amp though, which does help narrow the focus a little, as well as documenting everything about the settings and gear. Steve's adding metadata to the nam files - and IF people fill that out, it will be helpful for informing UI instances like plugin or tonehunt. But it is what it is for now and just sort of the nature of the beast. Helps getting to know the names of creators that do good work and starting there vs diving into the abyss too, since there are a lot of people who really know what they're doing. Output level normalization is being considered too, which would balance the capture playback levels better.

Captures from tim robertson, jon arnold, helga behrens, philip priss, emile rohbe for example

A few good spots:

NAM_JonathanArnold – Google Drive

drive.google.com drive.google.com

JZ NAM Models – Google Drive

drive.google.com drive.google.com

NAM Models - PippPriss – Google Drive

drive.google.com drive.google.com

Lots should / will be on tonehunt, but I just have these handy from my bookmarks
 
Last edited:
Let's just set a silent standard for reamping, 12dBu.
That's ±0.5dB either way form Helix, Focusrite, Audient and UAD interfaces High-Z inputs, we are actually very lucky.

In practice that means if you output a -9.8 dBFS Peak sine tone form your DAW, you should adjust your ReAmp box output level to read 1vAC RMS at the 1/4" output going into your amp.
 
What kind of consistency levels are you expecting? I wouldn't expect two JCM800's to sound the same, even if the reamping levels were exactly the same. Isn't the whole point of capturing a specific amp that you're ... well... capturing THAT specific amp??
 
The goal is to preserve the gain difference between those two JCM800's.
Consistency between captures done by different people on different setups.

If I ship my JCM800 (I wish) to you and you capture it with your setup, the capture itself will be identical to my capture if we both use 12dBu as reference signal.
 
James Freeman said:
Let's just set a silent standard for reamping, 12dBu.
That's ±0.5dB either way form Helix, Focusrite, Audient and UAD interfaces High-Z inputs, we are actually very lucky.

In practice that means if you output a -9.8 dBFS Peak sine tone form your DAW, you should adjust your ReAmp box output level to read 1vAC RMS at the 1/4" output going into your amp.
Click to expand...
I think if a basic reamping guide was written (perhaps one version with an explanation as to why it’s important) and another with a simple way for most guitarists to calibrate their levels it could definitely help a load of them. I think no one has really set out a “standard” for this stuff, nor explained it properly.

Basically all we have is people on youtube giving out conflicting information.

Stage 1 should be a written .pdf guide, stage 2 should be a youtube video demonstrating how to calibrate, stage 3 should be an app/plugin that guides the user through the process.

MAYBE it’s also a good idea to have some kind of spreadsheet with commonly used interfaces and modellers. Then people can just look up whether they need to compensate for their own input chain - I’d imagine 90% or more users probably aren’t creating any models and are only playing those made by other people.

If it were me (I’m glad it’s not), I would insist that any uploaded captures to those websites have to have calibration information included. There is no reason to not have this information available, and absolutely no benefit to ignoring it.
 
Orvillain said:
What kind of consistency levels are you expecting? I wouldn't expect two JCM800's to sound the same, even if the reamping levels were exactly the same. Isn't the whole point of capturing a specific amp that you're ... well... capturing THAT specific amp??
Click to expand...
fine when it’s your own amp and you have levels calibrated for that setup - by default it’s going to be captured in the way it’s intended to be used.

if someone else was to play it with their own setup, you’d want them to have just as authentic experience. And if they capture their amp for you to use, then you don’t want to have to guess what level it wants to see.
 
You must log in or register to reply here.
Back
Top