What is Neural Amp Modeler, and do you actually need it?

Short version: NAM is how you get a copy of one specific amplifier, measured rather than approximated. It is free, it is open source, and there are thousands of captures online made by people who own amps you do not.

The longer answer is about what that costs you, and when you would rather have something else.

How does a capture actually work?

You play a known test signal through a real amp and record what comes out. NAM compares the two and trains a small neural network until its output matches the recording. That network is the capture.

It is not a model of how an amp works. Nobody told it about valves. It is a measurement of what one amp did, on one day, with the knobs where they were — which is the source of both its strength and its limits.

What is a capture good at?

Being that specific amp. If someone captured an amp you have heard and want, a good capture of it is uncannily close. Closer than a description of that amp could be, because it is not a description.

Existing at all. Plenty of amps are rare, expensive, or in a studio you will never book. A capture is a file.

Being free. The format is open and the libraries are free. This is genuinely unusual in guitar gear.

What is a capture bad at?

Having knobs. This is the big one, and it surprises people.

A capture is a snapshot of an amp at one setting. Turn the gain up on the real amp and you get a different circuit behaviour; turn it up on a capture and there is nothing to turn, because the knob was never captured — only its position on the day. That is why a loaded capture in RigShed shows just Level and Mix: those are the only two honest controls, and showing a fake gain knob would be a lie about what the file contains.

Most capture libraries work around this by shipping many captures of the same amp at different settings. It works, and it is a lot of files.

Being cheap to run. A neural network is doing arithmetic on every sample. RigShed’s own Clean amp is an analytic circuit model instead, and it costs 0.10% of one core — roughly 30× less than running a capture. On a phone, that difference is battery.

Telling you what it is. A capture is a file with a name someone chose. There is no standard for what the name means.

Captures or a modelled amp — which should you use?

Not a real competition. They answer different questions.

	A capture	A modelled amp
What it is	A measurement of one amp at one setting	A circuit simulated from its behaviour
Knobs	Level and Mix	All of them, and they do what they say
Changing the sound	Load a different capture	Turn a knob
CPU cost	High — a network per sample	Low. RigShed’s clean amp: 0.10% of one core
Where it comes from	Someone with that amp and a day spare	Fitted against a measured reference

Use a capture when you want a specific thing — a named amp a track was made on, or a sound somebody shared.

Use a modelled amp when you want to work. Turning a knob and hearing the circuit respond is how you find a tone, and a capture cannot do it.

Do I need NAM to use a guitar amp app?

No — and this is worth saying plainly, because the NAM-native apps tend to imply otherwise.

Several iOS apps are NAM loaders: they are excellent at playing captures and they require you to bring your own. If you have a library, that is exactly what you want. If you do not, you have bought a very good player with nothing to play.

RigShed works the other way round. Eight amps, each fitted against a measured reference, are there when you open it — and captures load on top if you want them. Both are free, and there is no cap on how many captures or impulse responses you keep, on any tier.

Will my captures load?

Yes, whatever they were trained with.

NAM has several model architectures, and they have changed over the years — which is what people are asking about when they ask about “A2”. RigShed links all five, including the one every modern capture uses.

This is not a version badge on a page. It is the difference between a file someone shared this morning opening and erroring, and it was found the hard way: an early version of that list left one architecture out, and “every modern capture in a nine-pack test library” failed to load. Now they do not.

Drop .nam files into the RigShed folder in Files or the Finder and they appear in the app. No import dialog, nothing to convert.

So what should I actually do?

If you have never used a capture: start with a modelled amp. Learn what the knobs do. You will understand what a capture is for much faster from there, and you will not have spent anything.

If you have a library already: bring it. Keeping captures and IRs uncapped on the free tier is a deliberate decision, not a trial limit.

If you are choosing an app on NAM support alone: that is table stakes now, and most of the cluster has it. The thing worth comparing is what the app gives you when the captures are switched off.

---

RigShed is a guitar rig for iPhone, iPad and Mac. Eight amps fitted against measured references, 41 pedals, and a vocal channel — see the gear, or the pricing, which starts at free and stays there.