How to capture your own amp
How do I make my own NAM capture?
You play a known test signal through your amp, record what comes back, and train a model on the difference using the Neural Amp Modeler trainer. It needs an audio interface, a computer, and roughly an hour. The result is a file any NAM player can load.
What is actually happening
A capture is not a recording of your amp. It is a model trained on how your amp responded. You send a known signal through it, record what comes out the other side, and hand both files to the trainer. The trainer adjusts a small neural network until feeding it the input produces the output, and the result of that is the capture file.
This is why captures reproduce the way an amp reacts as you dig in, rather than just its tone at one volume: the test signal covers a range of levels, and the model learns the whole behaviour.
What do I need?
An audio interface with an output that can drive your amp's input, a way of getting the amp's output back into the interface — a load box, or a microphone — and a computer to run the trainer on. The bar is much lower than people expect: no treated room is required, because you are capturing the amp rather than the room.
What matters far more than the gear is level discipline. Capture at a level that clips somewhere in the chain and you have trained a model of a clipped chain, and it will sound wrong in a way that is very hard to diagnose later.
Load box or microphone?
Through a load box you capture the amplifier alone, and pair it with an impulse response of a cabinet afterwards. That keeps the two separable, so you can change cabinets later without recapturing anything.
Through a microphone in front of a real cabinet you capture the amp, the cabinet, the microphone, the position and the room, all fused into one file. It can sound wonderful and it is completely inflexible. Most people who share captures do both.
How long does it take?
Setting up is the slow part. The actual test signal runs for a couple of minutes, and training runs unattended after that — minutes to an hour or so depending on your machine and how hard you push the settings. The first capture takes an evening because you are learning the process; the tenth takes a few minutes.
Why can I not do this on the phone?
Playing a capture and training one are very different jobs. Playing means running a small network a few thousand times a second, which a phone manages easily. Training means running a much larger process over and over while it gradually converges, which is a laptop's job.
So the division is a practical one: capture on a computer, then play the result anywhere, including a phone.