The Soul of Sound: What Remains When Music Becomes Data

How music becomes whole through what lingers, accumulates, and is never fully heard on its own

It all started with a simple question about audio bit depth—16-bit, 24-bit, or 32-bit—but, like most things in music, the answer wasn’t isolated. It lingered. One thought bled into the next, creating a kind of residual effect. Over the course of an afternoon, that question spiraled into a deeper reflection on psychoacoustics, vintage monitors, LANDR’s Pro mastering, the ethics of artificial vocals, and my quiet philosophy of finishing songs in a 150-square-foot room with an angled ceiling.

I sit three feet from a pair of Event 20/20 monitors I bought in 1996. They still work perfectly—powered, honest, a little bright in the highs, but completely trustworthy. My studio is small and efficient, with professional sound treatment and roadband absorbers wrapped in cloth, enough to swallow the worst bass nodes without turning the space into a coffin. I mix on open-back headphones too, and sometimes I leave the monitors running to feel the low-end air move around my ears. When a mix feels ready, I upload it to SoundCloud, hop in my car, and let road noise, engine rumble, and tire hiss decide its fate. The worst place to play a song is in a noisy vehicle—but if it cuts through that, it cuts through anything.

I record at 96 kHz, sometimes 32-bit float, because I want to give the converters every possible bit of resolution. I know full well that by the time a track hits Spotify, the system downsamples it to 44.1 or 48 kHz, applies lossy compression, and plays it through phone speakers or earbuds half-dangling from someone’s ear on a subway. Still, I want my master file fat, pristine, future-proof. I let the platforms handle the codecs. I refuse to make those decisions.

Lately, I’ve been relying on LANDR’s Pro tier for mastering and distribution. The real game-changer for me is STEM mastery. I export up to eight separate stems—lead vocal, backing vocals, drums and percussion together, bass, grouped pianos, grouped guitars—and upload them straight to the cloud. The AI processes each stem individually with its own EQ, compression, saturation, and transient shaping, then glues everything back together with a final, shared space that keeps the whole track living in the same room. I send mixed stems, reverbs, and all, because I want my choices to stay intact; LANDR adds the final polish and loudness. The in-DAW plugin helps me preview ideas quickly inside Reaper, but the real magic happens when I feed it those clean, grouped WAVs.

I stay pragmatic about it all. If a mix already sounds good, I let it ride. I doubt I’ll chase endless revisions when five new songs are already demanding my attention. Perfection remains a luxury; finishing stays a discipline.

My mind drifts back in time. I remember trading a beautiful ES-335 for a Kurzweil K1000 module in the late ’80s because that rackmount box housed the best grand piano sound money could buy at the time. I regret the trade now—the guitar would fetch a fortune, the module gathers dust as a relic—but those sampled pianos shaped my ear. I started sequencing in 1992 with an Alesis drum machine, programming rolls and fills I could play live, never notating anything I couldn’t perform. I still don’t program anything I couldn’t play, though sometimes I roll my eyes and let the piano double a guitar lead just to make it punch harder.

I always write songs on the piano first. I sit down, build chords that move like living things, and let each note travel its own small journey through time, interacting with its neighbors. Counter-melodies emerge naturally in the inner voices—alto lines descend while soprano holds, tenor leaps just ahead of the beat. By the time I lock the chord progression, the vocal melodies and harmonies already hide inside it. The vocals wait for me to sing them out loud. I add guitars last; their strums and arpeggios decorate what the piano has already declared complete.

I love psychoacoustics—the “residual effect” Apple describes in Logic Pro’s Vintage B3 manual, where the brain reconstructs a missing fundamental from its harmonics. Stack the right overtones, and the ear invents a bass note that isn’t physically there, freeing up low-end space in the mix. That same idea—the mind filling in what’s missing to create something whole—inspired my book on leadership. Just as the ear supplies the absent fundamental to make music feel complete and powerful, great leaders often create unity and direction by helping people connect the dots between what’s said and what’s truly needed, even when the “fundamental” isn’t spelled out. Producers have used this trick since tiny transistor radios somehow conveyed the thump of a kick drum. Today I use it deliberately on small speakers and earbuds, though I haven’t applied it systematically yet. The seed sits planted.

The discussion always turns to AI, and I remain unfazed by the panic. We’ve used artificial intelligence since the ’80s—MIDI is AI, sampling a piano and playing it from a keyboard is AI, drum machines, synth presets, and ACE Studio’s uncanny vocal renders all continue the same line. What has changed is how much better AI has made sampling itself. Neural synthesis now generates infinite expressive variations, seamless articulations, and realistic physics on the fly. At the same time, stem-separation tools extract clean instruments from old mixes, and generative models create brand-new textures from text prompts.

Sampling a voice is far harder than sampling a piano. The human voice carries infinite subtle variations—breath, vibrato, timbre shifts, emotional inflection, the unique shape of a throat and mouth—that change with every phrase and feeling. A piano note, once struck, follows a predictable decay; capture it once and trigger it ideally every time. Yet a sampled piano is every bit as much a copy as an AI-generated vocal—an analog instrument frozen in digital form, divorced from wood, hammers, and felt. I often point out that even Rick Beato layers four pitched versions of a low male voice to create an impossibly deep timbre. What’s the difference between that and an AI doing it in one pass? There isn’t one. The soul lives in the intention behind the note, not in the hammer striking the string.

I don’t lean heavily on AI vocals—I want the final product to feel like me—but I refuse to fear the tools. Drum loops save hours. A sampled ’80s grand does the job. If ACE Studio ever captures my exact phrasing and breath, I’ll use it without apology. My goal is communication, not purity.

By the end of any long session, the sun shifts across my small, angled room. The Event 20/20s wait quietly for the next mix check. Five new songs grow in progress, each one born on piano, each one carrying invisible vocal lines in its inner voices, polished by LANDR Pro’s stem magic, listened to in cars and kitchens and subways by people who will never know how many hours I spent making them merely good enough.

And that, I realize, is the entire point. What we hear is never just what is present in isolation, but what remains, overlaps, and resolves in the listener’s mind—the residual effect at work. Music isn’t about perfection frozen in amber. It’s about conversation—between notes, between harmonics, between maker and listener. All the bits, all the sample rates, all the AI black boxes and vintage monitors help me keep that conversation going a little longer, a little clearer, across whatever noisy room the world happens to be in today.