Gear is Data; Data is Gear

The more gear you have, the more data you can collect. This has always been true; when you buy a guitar or overdrive, part of the dream is not only loving that guitar or overdrive but maybe learning something even more generalizable about what you're looking for in a guitar or overdrive pedal.

But now that Neural Amp Modeler (NAM) has become widely used by guitarists, bassists, producers, and audio engineers to produce great-sounding models of real amps with deep learning, the concept of "gear as data" starts to become important in a much more literal way. Although many are quick to talk about how "modelers are killing tube amps," I think the tube amp's utility in training models on tube amp behavior will keep musicians and enthusiasts buying them for a long time into the future.

Don't let your tone be locked to one form factor

What would it mean to you if you could take your amp "seamlessly*" (there's always costs and catches but I'll talk about that more later) across different form factors? What if you could carry a small modeler/hybrid pedalboard to your gigs instead of a massive pedalboard/head/cab rig? What if you could practice with your amp's true cranked tone without waking up your kids? What if you want to use your amp's tone for rhythm in a gig but switch to a Dumble amp to play certain lead parts? All of this is made possible by NAM along with DAWs, pedals, and amps that support it. And you don't have to be a pro to get a net benefit from this.

When you work in tech in various roles over years like I have, there's a certain "runs on my machine" mentality that you become allergic to. If you are being paid to develop some sort of app or website, you better be sure it actually works on the real servers, phones, and computers that will need to use it. Neural Amp Modeler saves you from a "runs in my bedroom" or "runs on my headphone amp" situation where your tone is less portable than it could be ("portable" in both the physical sense and the sense of being transferrable from one computer/system to another).

Creating full-rig models of my combo amp and pedals has highly useful for me so far. I was partially willing to do it as a sort of "science experiment," but I've quickly reaped real world value. Last night, I wanted to play really late at night, but I was having DAW monitoring issues that I didn't immediately solve. Instead of giving up or burning an hour troubleshooting, I simply switched to my Sonicake Pocket Master and played for hours on amps that included a model of my own.

The Pocket Master also has more than 100 different effects, letting me easily remix my amp's sound with effects not on my board. I've found it particularly fun to experiment with different reverbs since I've never found the reverb on my Marshall to be particularly inspiring. This adds to the overall positive impression of this pedal that I shared in my review of it. Blackstar is also jumping into the fray with a practice amp that will natively support NAM architecture 2 when it releases, highlighting that this philosophy of bringing an open-source capture/modeling ecosystem to musicians is starting to gain real mainstream penetration in the amplification industry.

The new economics of buying pedals and amps

In the past, the value of music gear was dominated by rarity and musical utility (for a particular buyer/player). The ability to capture and model gear adds optional value to amps, pedals, and more.

To be more precise, the "musical utility" of gear is no longer monolithic. If you want an amp to move air with a real cab or a pedal to have control of knobs on your pedalboard, you do need to continually own the gear to get what you need. But if you only need particular settings of a pedal or amp as digital building blocks, then you can sell it after getting the models/captures you need. There's even the potential to retrain on your captures with future architectural improvements to NAM, as will happen on TONE3000 when the new A2 architecture is ready.

If you have a Youtube channel, you could build a rig for a video, give it away to a subscriber, and still record with a modeled version of the rig later! I think there are lots of fun ways to exploit this kind of stuff if you're positioned properly.

If NAM didn't exist, then I would see almost zero point in buying a pedal that I know can't earn a permanent spot on my board. But with NAM, owning a pedal at least temporarily can let you capture it and use it in your DAW for recording later.

I'm not sure where exactly this will drive the future of the used gear market, but one thing seems clear: liquidity should continue to trend up. There's more reason to buy amps, and there's more reason to sell them. Even if various modeling-based products and DAW-based workflows are more convenient for everyday use for most guitarists, enthusiasts will still demand to own tube amps to have data about how the real thing behaves.

One final point is that NAM currently cannot learn time-related effects (delay, reverb, etc), so not all pedals are affected equally. Is a delay pedal more valuable or less valuable now than in the past?

No free lunches

I love open-source tech and have spent a lot of my life writing about its benefits for data analysts, developers, businesses, and now musicians. It would be dishonest, however, to pretend there aren't costs. A lot of the benefits of benefits of using open-source technologies, including NAM come from increased sovereignty. This is necessarily a double-edged sword; if you gain something from being more of control of the technology you use, there is also the potential to waste time or misuse this control.

Although I am happy with the models I've been able to make of my gear so far, there was definitely a very real setup cost. It's not rocket science to train models on your gear with NAM, but it is easy to screw it up in various ways when you are new (sometime I'll write up an article cataloging some mistakes I made to try to help others new to the ecosystem). For example, I was reamping with a signal that wasn't hot enough at first, causing the models to not learn the real dynamics of my amp when played with a real guitar. These costs can be high enough that it just makes sense to buy a Quad Cortex for some people. I "wasted" a couple evenings making mistakes. But persisting through the startup costs reaps real benefits. Not only do I have models of my own gear that gives me flexibility and future time savings, but I'm now "plugged-in" to a world of others doing the same thing with their own gear with the same format.

Wrapping up

For us guitarists and bassists, gear isn't just gear. With modern software and NAM's application of machine learning, gear is the opportunity to create new gear.

Making models doesn't require you to buy more physical computing power yourself (thank God), as you can hosted compute to train on your reamped audio with this notebook or TONE3000's free service. If you already have a larger pedal collection than can fit on your pedalboard, it might make a lot of sense for you to look into modeling your pedals.

I can also see why NAM and TONE3000 could be scary from the perspective of a small (or large) pedal/amp maker, but there are also new opportunities for liquidity and distribution. I don't think this stuff is a net negative for demand for guitar amps and pedals, but it changes who will buy and why.

Impulse responses are an example of data that is derived from gear that can also act as gear. NAM is not the origination of the trend I've discussed here, but it is the expansion of it to gear where modeling non-linear behavior accurately is table stakes.

Capture your rigs and share them! Does modeling change your philosophy towards gear acquisition? I'm interested in hearing how other guitarists are navigating this part of the gear economy.