NAM: Neural Amp Modeler

dived back into NAM today, it actually works quite nicely in Pro Tools inside the Waves wrapper thing.

made a capture of my 2203, boosted with a Mesa Grid Slammer.



I have no doubts about how accurate NAM is, and its really progressing every time I check it out. Genuinely curious how much the results matter - is NAM really miles ahead of the competition?

I think there are very valid uses for capture/profiling tech, but there are also situations where its not ideal compared to dialling things in yourself. I think its a very awesome tool to have at our disposal though and the concept is definitely some kind of a technological breakthrough.

After finding out that the Kemper is set up for reamping at 5dBu and people are making models at ANY unknown calibration, I really think the sooner people get a grip on their levels the better. There's tons of great models out there, but the process of sifting through randomly captured models AND having to adjust the input level by an indeterminable amount every time isn't the best it can be.
 
MirrorProfiles said:
dived back into NAM today, it actually works quite nicely in Pro Tools inside the Waves wrapper thing.

made a capture of my 2203, boosted with a Mesa Grid Slammer.



I have no doubts about how accurate NAM is, and its really progressing every time I check it out. Genuinely curious how much the results matter - is NAM really miles ahead of the competition?

I think there are very valid uses for capture/profiling tech, but there are also situations where its not ideal compared to dialling things in yourself. I think its a very awesome tool to have at our disposal though and the concept is definitely some kind of a technological breakthrough.

After finding out that the Kemper is set up for reamping at 5dBu and people are making models at ANY unknown calibration, I really think the sooner people get a grip on their levels the better. There's tons of great models out there, but the process of sifting through randomly captured models AND having to adjust the input level by an indeterminable amount every time isn't the best it can be.
Click to expand...


I don't think nam is "miles ahead" in terms of tone. To a certain extent I often feel that if you can only really know which is real based on isolated A/B comparison, the difference doesn't matter in a realistic sense. NAM was also never trying to be better than anything else, it really just kind of worked out that way. Steve has even stated that his implementation is pretty simple and was surprised to see it working as well as it has. Once the hype train left the station, the usual comments flowed in.

However, the fact that it's free / open source is a big deal I think. The community has been frankly amazing at building and contributing captures and compute time. Sure, plenty of issues on the table - but honestly that's fine given what it is and how it's evolved.

I have no idea what's what in your video. Waveforms might yield a clue if you know what to look for, but it doesn't really matter. All of those sound pretty good, and my ear definitely has preferences (I like A), but I can't say for sure any one of those isn't a real JCM.

Also pls reveal which is which eventually :p
 
northern_fox said:
I don't think nam is "miles ahead" in terms of tone. To a certain extent I often feel that if you can only really know which is real based on isolated A/B comparison, the difference doesn't matter in a realistic sense. NAM was also never trying to be better than anything else, it really just kind of worked out that way. Steve has even stated that his implementation is pretty simple and was surprised to see it working as well as it has. Once the hype train left the station, the usual comments flowed in.

However, the fact that it's free / open source is a big deal I think. The community has been frankly amazing at building and contributing captures and compute time. Sure, plenty of issues on the table - but honestly that's fine given what it is and how it's evolved.

I have no idea what's what in your video. Waveforms might yield a clue if you know what to look for, but it doesn't really matter. All of those sound pretty good, and my ear definitely has preferences (I like A), but I can't say for sure any one of those isn't a real JCM.

Also pls reveal which is which eventually :p
Click to expand...
I think for capturing tech its certainly in front of the competition, but maybe with slightly more demands on the user to get the best results. But yeah, to be clear - the comment is really more about the hype at the moment than anything else. Totally agree with all your points btw.

Yeah, I actually had a feeling with this test if people could see the waveforms or FFT's, it may influence their decision. Will PM you results
 
northern_fox said:
To a certain extent I often feel that if you can only really know which is real based on isolated A/B comparison, the difference doesn't matter in a realistic sense.
Click to expand...
Agree completely. We are in the happy situation where there's a lot of good gear and software on the market. It's not a clear cut "this is better" thing anymore but a "does this combination work for me" thing.

I certainly would not be able to say which is which in that video and all of them are totally fine - the differences are there but I'd be happy with any of those tones.
 
just a slight addendum to the last video - thought I'd compare NAM to ToneX and Kemper. Gotta say, its really noticeable how ToneX absolutely requires some manual adjustment of the gain after the training. It definitely opens up some margin for error that slightly reduces the benefits of profiling sounds.

I think Kemper is kind of obvious - maybe with more manual tweaking it would get closer? and I'd also imagine for many situations the differences really wouldn't matter. IMO the other 2 platforms get noticeably closer.

 
So I finally got a chance to actually use NAM last night for a very quick drumjam and I was impressed. I installed it, went to the tonehunt site, grabbed about 10 profiles based on popular added an IR and was pretty surprised to be honest. If I came across NAM before Tonex I probably wouldn't have bothered with Tonex. Now I need to figure out how to capture with it...


 
StratsoundFan said:
So how do I need to adjust the gain on my audio interface, NAM, and in Reaper. in order to get the most out of your captures? i don't really understand the "calibrated for 15 dBU" part of your post!? Thanks.
Click to expand...
Power your audio interface input with a 1volt sine wave. And set your DAW to be displaying -12.8dBFS! Calculation and methods see GOOGLE.
 
StratsoundFan said:
So how do I need to adjust the gain on my audio interface, NAM, and in Reaper. in order to get the most out of your captures? i don't really understand the "calibrated for 15 dBU" part of your post!? Thanks.
Click to expand...


If you are using your interface’s built in instrument then you don’t even need the video above - set your gain at the minimum settings and check the spec sheet for the instrument inputs maximum input level.

if the number is higher than 15, then add that difference in dB, if it’s lower then subtract. If you are using a separate DI and converters, or aren’t sure of the specs then use the video above to measure.
 
MirrorProfiles said:


If you are using your interface’s built in instrument then you don’t even need the video above - set your gain at the minimum settings and check the spec sheet for the instrument inputs maximum input level.

if the number is higher than 15, then add that difference in dB, if it’s lower then subtract. If you are using a separate DI and converters, or aren’t sure of the specs then use the video above to measure.
Click to expand...

My audio interface is Audient id4 mkii. The specs for D.I/instrument input are 12dBu=0dBFs.
 
StratsoundFan said:
My audio interface is Audient id4 mkii. The specs for D.I/instrument input are 12dBu=0dBFs.
Click to expand...
Cool, so it basically means your DI’s record 3dB hotter than @Orvillain , so you just have to lower the input gain by 3dB in the plugin and you’ll have the same gain level they were captured for. You can always adjust by ear if you like it different but that would be the closest to plugging into the amp
 
Thanks. I'll give it a try. My only gripe with NAM and Tonex is that I almost always have have to tweak input and/or output settings, to find that sweet spot.
 
StratsoundFan said:
Thanks. I'll give it a try. My only gripe with NAM and Tonex is that I almost always have have to tweak input and/or output settings, to find that sweet spot.
Click to expand...
Will always be the way unfortunately with this stuff, especially when most people aren’t aware of what their particular levels are.

FWIW I capture all mine at 12dBu - I figured it’s a common value for a lot of instrument inputs, and it’s also a fairly easy level to set the reamp box for.

But it does suck not knowing where the gain should be set, and having to guess how to get the “correct” response
 
12dBu is the sweet spot in many aspects.
Maybe not coincidentally, standard 9v stompboxs have ±4.5v peak voltage rails which is 12.3dBu, they don't go above that and even slightly lower at max.
Proper modeler design should have at least +12dBu inputs to accommodate standard 9v stompboxes and a 6db pad for up to 15v stompboxes without clipping.

Kemper's default setting of +5dBu (1.9v peak) is on the low side, not suitable for humbuckers or boosting with stompboxes.
 
James Freeman said:
He captured his amp with a microphone or applied the IR afterwards in DAW.
No, you can't remove anything form the capture.



I followed this video:

Click to expand...


DLC86 said:
I did this little step-by-step guide which sums up the procedure described on the video above:
Click to expand...

Thanks James. I got it to work but was not a heck of a lot faster than the Google link, even though I did have a RTX 2070 GPU. Maybe it was not actually using the GPU? Either way glad I have the option of being able to do it myself now.

I think I get how they inbed the IR now. By running a second CPU with an IR loader in a DAW.
 
NAM Plugin v0.7.3 was released.

NAM 073.png


Finally, fully functioning on a stereo track with correct R, L, and L+R input support.
A ton of changes.

github.com

Releases · sdatkinson/NeuralAmpModelerPlugin

Plugin for Neural Amp Modeler. Contribute to sdatkinson/NeuralAmpModelerPlugin development by creating an account on GitHub.
github.com github.com

What's Changed​

Click to expand...

I freaking love open source. :love
 
Last edited:
0.7.2 also included a big CPU reduction (about 30-40%)! The selection arrows should be on the next release too since Oli already did it. They were held on this release just in case the memory issue wasn't fixed (thought it appears to be!)
 
You must log in or register to reply here.
Back
Top