Neural Amp Modeler NAM amp capture/profile player FREE *December 14, 2023 update 0.7.7

[ladron]

I'm curious if after the capture is done, can a lower powered PC be used for playback mode?

I'm curious about this as well. The VST plugin runs at just under 2% CPU usage on my PC (Ryzen 5 2600). For comparison, Amplitube 5 uses closer to 1% CPU for a full rig simulation, so the NAM plugin is relatively heavy.

I don't know how much effort has gone into optimizing the NAM plugin code, though.

[metalifuxx]

what a great project, thanks for sharing and the efforts from all who contributed #thumbsup

quality is impressing. had a quick tryout with the very nice JCM800 v1.1 #1 at the weekend together with a York Audio MV30-Dual IR.

I have a Friedman Dirty Shirley, Suhr RL and some other nice tube stuff and I'd like to try to capture it myself if I find some time. Still need a reamp box though, having an eye at the Palmer daccapo

do the captures automatically translate to all sample rates or would it be best to play them at capture sample rate for full accuracy?

There was a discussion on this in the FB group, and it was noted that for right now, 48kKhz is the working supported rate for captures and playback so far.

[Steve Ack] “There's no resampling in the plugin at all. It's just meant to be used at plain 48kHz if you want the models to sound the way they were intended.

But, if you want to use sample rate as a creative tool to get new sounds (I'd think it'd sound noticeably different), then go for it! ”

[metalifuxx]

I have troubles getting the standalone app to receive input from my audio interface. Tried switching inputs, creating aggregate device etc., the in-built laptop mic works, but the interface inputs don't in any configuration. Audient iD4 interface, macOS Big Sur. The plugin in a daw does work as expected though. Any ideas?

Seems this a bug on Mac for anyone that has tried it. Use the plugin version for now.

[metalifuxx]

If you’re interested in the capture/shooting part of your amps/pedals hardware etc, but don’t have the technical knowledge to get the training process going, nor a faster modern GPU/CPU for quicker training, there are a few people on the NAM group, specifically Jon Arnold, who will process/train your wav files into NAM profiles for you if you send them to him.

[ladron]

From this look inside the IK Multimedia TONEX pedal, it looks like it is running on pretty modest hardware (Arm Cortex M7):

https://www.thegearpage.net/board/index ... e.2443517/

This would imply that playback of models like NAM should be able to run pretty efficiently.

[metalifuxx]

From this look inside the IK Multimedia TONEX pedal, it looks like it is running on pretty modest hardware (Arm Cortex M7):

https://www.thegearpage.net/board/index ... e.2443517/

This would imply that playback of models like NAM should be able to run pretty efficiently.

People in the group have already been discussing hardware/portable platform ideas of that sort, but it's still early for that, as the plugin is still not a 1.0 version, which I guess designates that it is still in beta, albeit mostly stable. One member said he is going to try to get it up and running on a Raspberry Pi pedal at some point. There are now some very intelligent users and some prior known devs in the group as well as a growing amount of users adding new profiles daily.

In addition there are also some content creators and people in pro studios with a vast collection of amps to model getting on board shortly, as well as youtuber/guitarist/content creator and pedal/amp/plugin demo video guy Leon Todd, who has just shared a capture of his beloved Mesa Mark IV C amp to the group. This group has grown so much in the last week or so. Thanks for the support and sharing of any info/media/videos related to NAM.

[ladron]

One member said he is going to try to get it up and running on a Raspberry Pi pedal at some point.

I've got my own DIY Raspberry Pi pedalboard, so I've been thinking along the same lines.

[Danilo Villanova]

Very cool project, hope it keeps growing. I do think that CPU optimization would be nice, here it's using 10 x the CPU that Tonex uses.

[ladron]

Very cool project, hope it keeps growing. I do think that CPU optimization would be nice, here it's using 10 x the CPU that Tonex uses.

In addition to code optimization, a lot must depend on the size of the neural network being used. Tonex may well be using a smaller network layer size.

[onerob]

Haven't downloaded yet but definitely keeping an eye on this promising project.

One thought: Does it need a name change? Neural DSP might not want the potential marketing confusion, unless NAM is effectively the name of this product?

[metalifuxx]

Haven't downloaded yet but definitely keeping an eye on this promising project.

One thought: Does it need a name change? Neural DSP might not want the potential marketing confusion, unless NAM is effectively the name of this product?

You'll try it once with the right amp profile and IR and wonder why you didn't try it sooner . It tends to be everyone's initial reaction, just look at the FB posts and on here, including mine, the first beta versions that were directly ported to VST3's first before the official now current profile loader plugin was released.

Yes, Neural Amp Modeler is the official name, but NAM is what every user has been calling it for the shortened abbreviated acronym. Neural networking is a technology process/term, and a number of plugins already use "neural" in their name, and each company has their own code and parameters for the capture/training process. And NAM is open source under the MIT license, with no money being made. Perhaps if they used the same primary initials/acronym in the name to mean something else, it would be infringing on the namesake. Steve Acktinson has said he by no means wants it to compete or step on anyone's toes. The tech being free, very high quality, and open source, may ruffle some feathers in the commercial amp sim scene, but freeware open source devs are always pushing the tech envelope to inspire other commercial devs to step their game up (or be hired and bought out). It's a win/win for everyone. The legal/moral integrity debate will come with what sources are being captured and shared by users, and his message to this community is that you please only capture the real physical hardware amps that you or someone else owns, but that can go in so many directions when you look at IR companies selling digital reverb IRs, making cabinet IRs etc, what other amp profiles from non-hardware amps that current companies allow to be captured/shared/sold etc.

[onerob]

Fair enough

[metalifuxx]

From Steve Acktinson (NAM developer):

New plugin version just got released! Version 0.7.0!

WHAT'S NEW

* There's a noise gate! More on this in a comment maybe?

* It should be more stable! If you've had issues with the output audio cutting out, or (fingers crossed!) the plugin crashing your DAW, then this should make a major difference. I caught a spot where some invalid output was leaking out of the NAM models and fixed it. Makes a lot of sense to me that it'd fix a lot of problems We also fixed the problem where .nam files were grayed out!

* There are performance improvements! 🏎 A few folks who know DSP have looked at the code for the models and told me "...that could be improved" ...and a member of the community stepped up and did just that! The power of open-source!

🛠 THIS IS A BREAKING CHANGE 🛠
Ok, time for a little bit of bad news: version numbers going up is actually bad for you as a user The fact that we went from 0.6.2 to 0.7.0 instead of 0.6.3 means that we're going to break a few eggs. Here's what to expect:

* If you've got signal chains saved with NAM in them, then they're going to load up a bit weird after you install this new version of the plugin. This will happen every time I add a new knob unless I go through a lot of trouble, and it wasn't worth it to me to keep holding onto this for that reason.

* If you've been hanging on to old "directory-style" models, they're no longer supported. Either convert them using https://github.com/sdatkinson/nam-model-utility or you'll have to stick with 0.6.2.

Enjoy!

Neural Amp Modeler (NAM) Plugin Version 0.7.0 is released!

(correct working link ( FB)
Plugin: https://github.com/sdatkinson/NeuralAmp ... tag/v0.7.0

Trainer: https://github.com/sdatkinson/neural-amp-modeler

This release sees the introduction of a new noise gate feature, as well as performance improvements on the main NAM modeling code and stability improvements to the plugin overall. Enjoy!

[metalifuxx]

Github repository of a lot of models/profiles that Chris Kenrick has been managing. Look for my post for a google drive zip file link a couple posts back if ya want most of the ones i have collected from two weeks ago. In the process of downloading all the new profiles that are constantly added every day now to update the zip file...

But please go to the NAM facebook group if you are interested and want to stay up to date on all the new amp profiles being made and shared, as well if you want to contribute to capturing and sharing, or helping with the plugin/capture tech/feature requests/bug reporting in any way. The group is now approaching 1,000 members and growing exponentially each day in the last two weeks...more members means more profiles.

https://github.com/pelennor2170/NAM_models

[metalifuxx]

Leon Todd's capture of his Mesa Mark IV, high gain channel for chug rhythms/riffs and lead (amp only, needs IR/cab sim)

https://www.dropbox.com/s/jzlwefbgx1g7c ... RKIV_1.nam