I’ve been thinking a lot about NAM A2 lately, and something finally clicked.

Steve Atkinson created Neural Amp Modeler. Tone 3000 is the company behind the massive public library of NAM captures — and the team that sat down with Steve and built NAM A2. The version number is not the story. What they did with it is. They took a competitive advantage — something that was hard to develop and hard to do well — and turned it into a feature. A feature any company making a $150–$300 unit can put in the box. Want Tone 3000 access on top of that? Add Wi‑Fi, a little more DSP, point people through your software, and you’ve got cloud too.

That’s the quiet revolution. Some of the smartest people in the room took capture accuracy — the thing the big platforms used to own — and put it on the shelf as something the affordable units can ship.

So where does capture technology go from here?

Why you wouldn’t start from zero

I’ve been talking this through with people in the space: I don’t think we see another brand-new capture engine walk into the market. If you’re entering now, why sink a quarter million or half a million into building from scratch when open-source NAM and NAM A2 are sitting there? Fork it. Keep developing. A lot of people already have.

ToneX, Quad Cortex, Kemper, Helix Proxy — they already have their methods. They’re going to keep iterating on those. NAM A2 is not the end of the road. Guitar tech isn’t “done.” There’s still a lot you can do.

Here’s what I think we see next from those camps.

Modeled tone stacks on top of captures

Kemper already does something unique: modeled tone stacks on top of a profile. Tell the unit your Deluxe Reverb was at 5‑5‑5 and it knows what rolling the treble actually does, because bass and treble on those circuits are mostly high- and low-pass filters. Know the slopes and the corner frequencies and those knobs stop being a black box.

Guitar players treat EQ as endlessly interactive. A lot of it isn’t. Marshall and Fender stacks especially — you can know what those knobs do.

It gets messier with a bright cap on the gain control. The unit needs enough breakup in the capture to hear what that circuit is doing. Feed it something 100% clean and the modeled stack won’t track right. Still: other platforms could steal this idea tomorrow.

Parametric captures and stitching

We’ve already seen pieces of this — parametric captures, multiple NAM captures stitched into something closer to a working model.

It reminds me of the Helix and Axe-Fx cabinet engines. You think you’re moving a mic around a speaker. A lot of that is IRs layered and blended, sometimes with EQ on top. Impulse responses all the way down.

That same stacking idea gets more interesting if you stop treating the speaker as a linear IR. Deduce a speaker from a direct capture plus a cab’d capture and you can keep breakup and non-linear response — then blend characters that don’t exist as one physical speaker. A Greenback that goes more ceramic in the lows and more Alnico on top. Parallel snap above ~1.2 kHz with Greenback presence underneath. IR blends are common. Non-linear speaker models you can ride in and out would be a step past that.

HeadRush’s Super Clones pointed the same direction for amps: several captures of one amp, blended between points so you get more of a model than a single frozen snapshot. Accuracy and workflow issues held that back — including having to leave the unit for another piece of software — but the idea was right.

ToneX could go there without reinventing physics. Capture the amp, then calibrate the tone stack live: bass at five, then three, then zero, then seven and a half, then full. Same for treble. Listen to what the knob actually does. On a Marshall or a Fender, that’s enough to nail the feel of the stack. A highly interactive Vox AC30 EQ is a harder ask. Stitching captures Super Clone–style on ToneX is a wild prediction, but it’s a clean next step.

Modeling layered on captures

You can also put modeling on top of a capture. Tag the capture as 6V6s, then model what happens to breakup, EQ, sag, and compression when you raise or lower plate voltage across 6L6, 6V6, EL34, EL84. Same family of idea as Kemper’s tone stacks: engineered behavior riding a real-world capture.

Kemper is still ahead in some of that deep engineering thirteen years later. A lot of good ideas are just sitting in that unit. That’s another piece on its own.

Stop only emulating analog

Capture and modeling both get stuck in “make it exactly like the amp.” Fine. Necessary. Not the ceiling.

Computers don’t emulate abacuses anymore. They do things abacuses can’t. Same question for digital guitar: why not push into sounds and controls that would be awkward or impossible in the analog world but useful for players?

Example: a smart “brightness” control that isn’t a dumb tilt. First lift is mostly 2k and up. Push further and it starts clearing low mids for a more pristine feel — not a telephone high-pass. Shelving bass might pull deep lows first, then add low mids back so it feels like a smaller amp instead of a thin DI. Adding mids might duck a little 300 Hz and add presence so the honk still has pick attack.

That’s the “magic mic EQ” trick: non-static, frequency-aware moves that players already do by ear. One knob. Programmable now.

Line 6’s Hype knob is an early version of “only we have this.” More of that, please.

The white space after NAM A2

NAM A2 took the big platforms’ hardest competitive advantage — accurate capture — and made it a feature on the affordable units. So what’s left on the whiteboard for QC, ToneX, Proxy, Kemper, Line 6?

Creative controls. Deeper stacks. Stitched captures. Digital-first EQ that actually helps. Stuff that makes each box unique for players.

Great time to be a guitar player. What we’ll have in a decade is going to be ridiculous.

-HW out