Not so long ago, it seemed as though studios would become entirely virtual. Instead, hardware has had a resurgence. DAWless synth setups are a thing. Producers are integrating outboard into computer‑based studios. Sales of reel‑to‑reel tape are up year on year. Even in live sound, there’s a growing trend to incorporate analogue rackmount kit into front‑of‑house rigs.
AI seems likely to accelerate these trends. The smouldering perception that using computers to make music is somehow inauthentic is being fanned to a flame by the rise of generative music creation services and virtual session musicians. But AI has also opened up a gap in capabilities between computers and dedicated hardware.
Even the most DAWless of setups often incorporate multiple pieces of digital kit. Some are based around DSP chips or FPGAs. Many are essentially embedded computers. Yet few are remotely as powerful as the average Mac or PC. A machine running code that is tightly optimised for a single task simply doesn’t need the same raw processing capability as a general‑purpose computer.
A consequence of this is that it’s difficult to retrospectively add AI capabilities to existing music hardware. There is no spare computing capacity available to execute machine‑learning models, even if it’s practical to code them for the platform. This doesn’t just mean that we can’t generate AI backing tracks or synthesize voices. It also limits the use of AI for more mundane but arguably more useful jobs, such as navigating huge preset libraries, or real‑time interactive help. And whereas we’re now used to accessing cloud‑based services on our computers, that seems like a less attractive option on a workstation keyboard.
The JBL BandBox is the first standalone device I know of in our sector to be designed around AI.
Which is why I think this month’s cover product is so significant. The JBL BandBox is the first standalone device I know of in our sector to be designed around AI. Stem separation is no longer groundbreaking in itself, but its implementation in a Bluetooth amplifier very much is. This is a grounded, practical application for AI that makes obvious sense, solves a real‑world problem and helps to differentiate the BandBox from its competition. It’s also one that plays to the traditional strengths of hardware, being portable, simple and direct.
As I understand it, the BandBox has been developed from the ground up as an entirely original design. In general, I suspect this approach will prove easier and more productive than to try to bolt machine learning onto existing platforms. Is it the first in a wave of disruptive products? Will we see new manufacturers entering this space? How will the existing giants adapt? And will the perception that hardware is more authentic than software survive such a paradigm shift?
Sam Inglis Editor In Chief