Neural Amp Modeler: Plugin, Profiles, Hardware – Our Complete Guide
How to create your own profiles, what NAM pedal is the best?
3,7 / 5,0 | 
Neural Amp Modeler: Plugin, Profiles, Hardware – Our Complete Guide · Source: RunawayThumbtack / YouTube
There has been quite the buzz in the amp modeling world in recent years: Neural Amp Modeler. The open-source technology is giving Kemper, Neural DSP, and IK Multimedia a run for their money. Not only is the plugin and modeling platform completely free of charge, but it also surpasses the expensive plugins from the competition in terms of quality more often than not. And the number of pedals with NAM support is growing, too. This guide will get you up to speed on what NAM is all about, which software and pedals support it, and how you can create your own NAM profiles.
Everything about NAM:
What is Neural Amp Modeler?
NAM stands for Neural Amp Modeler. It’s just a relatively new format for digitally emulating amps, to put it simply. On the one hand, that allows you to play amp sounds from legendary Marshall or Fender amplifiers, for example, that you could never afford. On the other hand, using it to model your own amps for the road helps touring musicians. They don’t have to lug their heavy amps around with them when they use NAM tools.
Contrary to its name, at least the early versions of Neural Amp Modeler seemed more about amp profiling than amp modeling or amp capture. Here, an acoustic impression of a single amp or pedal setting is taken, rather than virtually replicating the individual components of an amplifier. But that’s what makes NAM so good: even trained ears often prefer NAM over ToneX, Kemper, or Quad Core in direct comparisons.
In addition, both the NAM format and the plugin are open source, meaning they are free of charge, and anyone with the relevant knowledge and motivation can get involved. In addition to the official NAM plugin as a VST solution for DAWs, there is a growing number of hardware devices with Neural Amp Modeler support.
How NAM came about
For a full rundown on how things came to be, you can read the complete, detailed history of NAM on the official website. In short, Steven Atkinson first developed the format back in 2019 and uploaded it to the programming platform Github. As a passionate musician on the one hand and an AI researcher on the other, Steven combined his focus on machine learning with his hobby.
What started as an occasional evening pastime quickly became more. Because the results (loudly) spoke for themselves. Simulating guitar amplifiers using neural networks sounded so realistic that Atkinson was more than impressed. But then the project lay dormant for three years.
However, Steven couldn’t let go of the idea and, above all, the sound results. And so, in early 2022, he developed the first plugin version of NAM. This made the technology accessible to a much wider audience. He then began conducting comparative tests and publishing them as videos. In direct comparison with Neural Amp Modeler, Kemper, Quad Core, and ToneX almost never sounded quite as realistic. What a revolution!
2023 – Social Media Amp Modeling
Atkinson continued development and integrated a feature into NAM called Parametric Models, which made it much more versatile. It allowed amps to be acoustically photographed not only with a single setting but across the full range of their controls and settings.
Further improvements to the technology followed in 2023, but most importantly, Steven redesigned the NAM plugin, including the EQ and IR loader. And he and his collaborators improved the workflow for creating your own amp profiles. Meanwhile, the popularity of Neural Amp Modeler grew dramatically. There were initially a few hundred members in the official Facebook group, but by the end of 2023, that number grew to over 15,000! Today (as of March 09, 2026), we are at almost 25,000.
The number of guitar YouTubers who talked about NAM also grew, with some even inviting Steven for an interview. Neural Amp Modeling went viral. The first hardware solutions with NAM support also started to appear. And the number of profiles created by the community (see below) grew rapidly.
Software and amp profiles
If you’re looking to get started in the world of Neural Amp Modeler, you can do so completely free of charge on the software side. The plugin (AU and VST3) is available for download on the Neural Amp Modeling website. Install it and load it as an audio effect in your DAW.
Then you’re off to Tone3000 (formerly ToneHunt), the main hub for NAM profiles. There are over 6500 NAM profiles to choose from, from Fender Reverb to Marshall JCM800 to Vox AC30 or Roland JC 120B. Individual distortion pedals and preamps have also been modeled using Neural Amp Modeler, such as the ProCo Rat or the Boss Blues Driver.
Hundreds Impulse responses for cabs are also available. So, to get started, you select an amp and a cab, download them, and then load them into the plugin. Plug your guitar into the interface (Hi-Z input) and off we go!
Creating your own NAM models
Whether you want to participate in the community or simply don’t want to lug your expensive gear home from the rehearsal room, creating your own Neural Amp Modeler profiles is relatively easy. The workflow might not be quite as accessible as with ToneX, Quad Cortex, or Kemper, but with a little patience, even the more technologically challenged of us can do it. Basically, there are two (and a half) ways to create your own profile.
The simplest way is to do it directly via Tone3000. Go to the website, click “Create Tone” in the top-right corner, then “Capture.” Next, you download a WAV file of a sine sweep. Afterthat, you need to mic your amp, connect everything (i.e., the output from the interface to the input of the amp, then the microphone in front of the amp, and finally the microphone cable into the input of the interface), and finally set up your DAW this way:
- Load the sine sweep onto an audio track.
- Set the DAW’s routing so that the audio signal from the sine sweep is sent directly to the amp.
- Now create a second audio track to which you route the microphone’s signal.
- ATTENTION: Be very mindful of the volume of your monitor speakers in the process! Better yet, mute the audio output on your DAW’s stereo master for the modeling process; otherwise, ear-damaging feedback might occur!
- Start the recording in the DAW.
- Export the audio from the second track of the microphone recording (preferably mono, 24-bit, 48 kHz). Make sure that the new audio file has exactly the same length as the sine sweep.
Now upload this audio file to Tone3000 in the Capture area. As an alternative to the sine sweep, you can also upload a dry/wet pair. In other words, you split the signal from your guitar via a DI box and record it simultaneously before and after the amp.
Neural Amp Modeler without Tone3000
Option two comes from the official Neural Amp Modeler website. I’ve included the video with detailed instructions here; the process is relatively straightforward. Similar to Tone3000, you need a dry and a wet signal, i.e., WAV files of guitar recordings before and after the amp, of the same length. And here, too, you need a DI box and an interface with two inputs.
The actual NAM profiling process runs via a dedicated Google Colab notebook. What sounds cryptic is actually just a solution for running complex neural networks and code on Google’s powerful servers with just a few clicks. As far as I understand it, this is not a process in which your data, i.e., recordings of your amps, is used by Google for training. However, I know from my own experiments that a free Google account is required to use Google Colab.
Last but not least, there is also the option of performing the Neural Amp Modeler process completely offline using the open-source programming environment Anaconda. Check out this video:
Hardware for NAM
Neural Amp Modeler is not only popular for its outstanding modeling quality but also for being free software. However, the number of hardware on which NAM profiles can be loaded is also growing. First and foremost among these is the NAM Player from German company Dimehead, one of the first NAM-loading pedals on the market.
In addition, Darkglass recently introduced the Anagram, a bass multi-effect that can load NAM profiles. Budget multi-effects Valeton GP-5, GP-50, and GP-150 can also load NAM profiles. However, Neural Amp Modeler profiles must first be converted by Valeton’s app to a lower-resolution format so the pedals’ CPUs can handle them. There was also a recent firmware update for the GP-200 that made it NAM-compatible.
As for Hotone multi-effects, the Ampero II, Ampero II Stage, and Ampero Stomp II models are now NAM-compatible thanks to a firmware update as well. Just like with Valeton pedals, however, NAM profiles are downscaled during the import process so that the multi-effects’ CPUs can process them. This may result in a slight loss of quality.
In addition, there is the recently announced Sonulab Stompstation Pro. Its first batch sold out immediately, which, given its feature set and price, isn’t a surprise. It offers a NAM Loader that can load two NAM profiles simultaneously, a stereo IR loader, full support for NAM A1 and A2 (see below), a color display, MIDI sync, 192 kHz-32-bit converters, and a complete stereo signal flow. Under 600 euros. The just-announced Blackstar Beam Mini supports NAM as well.
The Future of Neural Amp Modeling
Just think about it. Neural Amp Modeling only really saw the light of day in 2022. That’s four years ago. And now we already have a growing number of pedals, a huge passionate community, and more and more converts from Kemper, QuadCore, and ToneX.
And it is highly likely that the capture process will continue to be simplified in the coming years. I would also venture to predict that the Neural Amp Modeler format will eventually put an end to the current amp profiling/capture/modeling situation, in which every manufacturer does its own thing.
I predict that, sooner or later, neither ToneX, Quad Core, Kemper, nor Helix will be able to avoid integrating NAM, at least as an option. Tone3000 has also made its API public for some time now, which allows NAM profiles to be loaded directly from there into guitar software/apps and then onto pedals.
A2: The next NAM generation
Just this year, in January, Neural Amp Modeling A2 (Architecture 2) was announced, which promises improvements in three different areas:
- Less power-hungry: Profiles created with A2 can be loaded on less powerful CPUs without conversion. This opens the door to full NAM compatibility for many more multi-effect and amp modeling pedals.
- Better sound: A2 will dramatically improve the realism of the virtual amps, including dynamics, gain structure, frequency response, and transients.
- Slimmable NAM: Gone are the days when manufacturers had to offer their own converters to downscale NAM profiles to CPU-friendly formats. With Slimmable NAM, the NAM loader automatically recognizes the CPU’s performance and adapts the profile’s resolution accordingly.
When A2 was announced, the developers aimed for a March 2026 release, so it should be right around the corner. It might take a moment or two for hardware makers to implement it into pedals, but we should be seeing plenty of firmware updates in the coming months.
Conclusion
Now it’s your turn, dear readers!! What are your experiences with the Neural Amp Modeler? Does NAM sound better or worse to your ears than other modelers? Let us know in the comments!
Info about the Neural Amp Modeler
*Disclaimer: This post about the Neural Amp Modeler contains affiliate links and/or widgets. When you buy a product via our affiliate partner, we receive a small commission that helps support what we do. Don’t worry, you pay the same price. Thanks for your support!
This might also interest you
