NAM: Neural Amp Modeler

I have the "Jeff Beck" one & I absolutely love it but man... with the Treble cranked, Gain at 2 and Output higher than 2 o'clock it can clip a 19 dBu input. That pedal's insane - so is the Fortin Grind.
Ha yeah; there’s some pedals that definitely benefit from a pad and a lot of headroom to capture! Or you could measure how much level they add at max, and then factor in the difference with your return metadata. Any of the TC Integrated Preamp circuit based pedals are another level of insane, must be more level than you can get out of a single 12ax7
 
it can clip a 19 dBu input. That pedal's insane

That's not insane in an analog tube world where an amp input can handle a huge range of levels that no A to D conversion can. That's the problem with trying to port gain and boost pedals into a digital world.
 
That's not insane in an analog tube world where an amp input can handle a huge range of levels that no A to D conversion can. That's the problem with trying to port gain and boost pedals into a digital world.
Handle is a kind of loose term, it’s more that we like the result of how it craps itself when tons of level is driven through it. If we were to make them equal, you’d only count the range of the amp where it stays linear.

A/D can handle it but you just have to pad/attenuate accordingly before.
 
Handle is a kind of loose term, it’s more that we like the result of how it craps itself when tons of level is driven through it. If we were to make them equal, you’d only count the range of the amp where it stays linear.

A/D can handle it but you just have to pad/attenuate accordingly before.

Yes, obviously a tube input stage can "handle" a tremendous amount of voltage without the magic smoke being released and this is only relevant because many of us don't mind or even like the sound of it clipping. We mostly don't like the sound of a digital converter clipping.

Padding and attention is problematic if you are trying to model the behavior of a system with and without the boost dynamically. Sure you could set up a rig where kicking on a boost also engages a pad, then adjusts the digital level of the amp so it sees the input get relatively stronger despite the pad, and also adds some boost at the output to correct for the change to the amp block level...etc.

We need more bits!
 
Sure you could set up a rig where kicking on a boost also engages a pad, then adjusts the digital level of the amp so it sees the input get relatively stronger despite the pad, and also adds some boost at the output to correct for the change to the amp block level...etc.
Or you could just use 2 input channels….

IMO things like the TC integrated preamp are kind of an edge case - it’s an actual preamp+eq which was never really intended to go before an A/D. Situations where you’d have to record the output of it, while being able to switch to a DI on a single channel are even more fringe. Easy enough to solve without being plug in and go situations. I’m not even sure that circuit is really adding much grit itself - it’s just a powerful eq and a lot of level from what I remember.

I suppose it could be fun to do some models using mic preamps with calibration and we could have 80dB of Neve gain in a plugin. TC IP has made a bit of a reputation for something you’d run into a guitar amp but I’d imagine most cases of using preamp, the preamp itself would be where the sound is made (rather than a shaping tool). Like Beatles Revolution etc
 
I use this Pedal Capture from @2dor all the time! I really like it at as a flexible TS style pedal in front of an overdriven amp capture. Or just as an OD flavor on an EOB amp capture. Flexible because there are so many settings combinations included in the capture, I can easily find the right sound.

But I agree. There's not much focus on Pedal Captures, even though they can work really well. Which is a shame since there are even more OD pedals out there than there are amps - at least I think so, don't actually know... :LOL:
 
Acustica incorporating NAM into their own modelling, interesting. Timestamped the vid here where he talks about it.


 
Last edited:
Acustica incorporating NAM into their own modelling, interesting. Timestamped the vid here where he talks about it.



I suspect there's many plugins out there using wavenets that people may not be aware of. Steven recently shared that he was involved in developing a plug-in for a company that uses a spin off the A1 models. Basically any company doing distortion related emulations has probably tested it out by now and tried different takes on it. That's just normal given the open source nature. Concept wise these are simple networks and experimenting with architecture variations would be the first obvious choice for many developers
 
Haven’t checked it out but I saw his demo of it. What an absolute legend. I thought someone would have done this years ago, but glad it’s done now.

I've been gushing about the last couple Cattle Decapitation albums in the "What are you listening to now?" thread since this forum opened and when I saw his name pop up in the New Member approval I did a double take. I didn't even know he was doing stuff like this, legend status agreed.

....now we gotta get some caps from those albums! :rofl
 
Hey dudes! Awesome you guys already found out about NAM-BOT!

I've been lurking here on and off, mostly digging into nerdy stuff about NAM with custom training architectures and what not. I was still experimenting with training on my own GPU for super heavy models and I had been really getting into vibe coding lately, so I figured I'd give this a shot. Turned out pretty sick, I think!

BTW, Just updated today with macOS support in beta with version 0.4.0.

Please check it out and let me know if you run into any issues!
🤖 https://github.com/daveotero/NAM-BOT
 
Hey dudes! Awesome you guys already found out about NAM-BOT!

I've been lurking here on and off, mostly digging into nerdy stuff about NAM with custom training architectures and what not. I was still experimenting with training on my own GPU for super heavy models and I had been really getting into vibe coding lately, so I figured I'd give this a shot. Turned out pretty sick, I think!

BTW, Just updated today with macOS support in beta with version 0.4.0.

Please check it out and let me know if you run into any issues!
🤖 https://github.com/daveotero/NAM-BOT
Stoked to try the mac version. Great to see you post here Dave, myself and a few others are also old Sneap forum guys who i’m sure all remember you.
 
Hey dudes! Awesome you guys already found out about NAM-BOT!

I've been lurking here on and off, mostly digging into nerdy stuff about NAM with custom training architectures and what not. I was still experimenting with training on my own GPU for super heavy models and I had been really getting into vibe coding lately, so I figured I'd give this a shot. Turned out pretty sick, I think!

BTW, Just updated today with macOS support in beta with version 0.4.0.

Please check it out and let me know if you run into any issues!
🤖 https://github.com/daveotero/NAM-BOT
Welcome to the forum Dave! I'm pretty sure I remember you from the Sneap forum days too. I would've been Amberience over there.

NAM-BOT looks great!
 
I am about to go to Floriduh for a week to see the sun for a change, but I will try this out (mac and pc) when I get back.
 
Back
Top