How to Create Your Own NAM Profile

Level: advanced. You'll need a bit of gear and about an hour of free time — but you'll end up with a digital copy of your own amp's sound that you can use anywhere.

What you'll need

A real amp (or pedal) you want to capture
An audio interface with an instrument input (Hi-Z) — e.g. Universal Audio, Focusrite, RME
A DI box/reamp box to send the signal back into the amp's input
A computer with the NAM Trainer installed (part of the Neural Amp Modeler project)
A quiet room and about 30–60 uninterrupted minutes (the recording itself takes a couple of minutes, but dialing in levels takes some care)

Capture the cab or not? If you want a ready-to-go "straight to the mix" tone, record the signal from a mic in front of the cabinet. If you want flexibility (e.g. swapping cab IRs later), capture the signal right after the amp (via a speaker-out through a load box/attenuator) — no cab.

Step-by-step process

Set up the amp

Set every knob on the amp to the position whose sound you want to capture, and don't touch them again until the recording is done.

Wire up the reamping chain

Interface output → DI box/reamp box → amp input. Amp output (speaker-out through an attenuator, or a mic in front of the cab) → back into the interface's input.

Dial in your levels (gain staging)

This is the most important step. The signal shouldn't clip going into the amp or when recording the response. Too quiet and the model will be noisy; too loud and you'll bake in distortion that wasn't in the original.

Record the test signal (reamping)

NAM Trainer generates a special test WAV file with a complex frequency spectrum. Play it through the amp and record the response at the same time — that's your training data.

Run the training

Load the original test file and the recorded response into NAM Trainer and start training. The process is fully automatic and takes anywhere from a few minutes to a couple of hours depending on your hardware (a GPU speeds things up considerably).

Check and export the .nam file

Compare the model's sound against the original on new material (not from the test file). Once you're happy with it, export the finished .nam file and use it in the plugin.

Common capture mistakes

Clipping the signal: Too hot a level going into the interface or amp — the model will "remember" extra distortion that wasn't actually there. Leave yourself some headroom.
Noisy signal: Too quiet a recording level, or noise picked up from bad cables — the model will learn the noise too, and it'll be audible even between notes.
Changing amp settings mid-recording: If the amp's knobs move between recording the test signal and checking the result, the model will sound "off," even though training technically succeeded.
Wrong test signal level: The test file needs to be played at the same volume you actually play at — otherwise the model won't "see" the dynamic range and nonlinearities you actually use.

Once you have a finished model, consider sharing it with the community — upload it to ToneHunt or Tone3000 and note the gear you used to capture it. It helps other guitarists and grows the open NAM library.