Before we discuss the current state of AI, please read the following instructions:
- Insert “At least as of now…” before every sentence.
- Insert “…although this could change at any time” after every sentence.
Thank you for your attention to this matter.
AI is now pervasive, but maybe not for the reasons I first proposed when I wrote about it ten years ago for Pro Sound News. (I called it “self-driving audio” because the term AI was not yet common, and I was trying to be clever with a pun on “self-driving auto,” which, as it happens, is also now advancing by the minute.) No, the general public isn’t ditching humanmade music for AI, although AI-generated music has certainly made inroads in areas like jingles and sonic logos.
Sadly, part of the pervasiveness isn’t due to the public discovering the joys of music, but from miscreants vomiting AI-generated slop all over streaming services. Now the concern is less that AI-generated music will replace human creators, but that it will drown out human creators.
Consumers have already made it clear they don’t like being conned, and in blind tests many listeners prefer human-generated music over AI-generated music—even when they think the AI stuff “sounds” better. That just confirms what I’ve always said about consumers caring more about music than sound. Or as B-movie/cult-classic director Ed Wood Jr. once said, “Filmmaking is not about the tiny details. It’s about the big picture.” He’s remembered to this day not because his films were good, but because they prioritized passion over polish, with a heavy dose of DIY vibe and unfailing enthusiasm. In other words…human qualities. Algorithms haven’t figured that one out. (For the last sentence, don’t forget to follow the instructions at the beginning!)
AI music gets the tiny details right. But if consumer acceptance is any indication, it fails at the big picture. So, let’s put that discussion on hold, and go into the studio, where AI has made considerable headway.
• Discover more great stories—get a free Mix SmartBrief subscription!
The industry, and the law, is starting to recognize the line between AI-generated and AI-adjacent. Consider IK Multimedia’s ReSing, which is an AI-adjacent signal processor, not a signal generator. You don’t type prompts. It doesn’t generate vocals. AI transforms the vocal you sing into a different-sounding voice. There are zero alterations to your expressiveness; if you slide a note, it slides. If you add vibrato, it adds vibrato, and if your note is off-pitch, it remains off-pitch. ReSing doesn’t do anything that isn’t in your performance.
Another AI-adjacent area is assisted mixing and mastering. I’ll quibble, though, and say this is about machine learning and pattern matching, not intelligence. AI can’t make a judgment, like, “Do you really want that self-indulgent guitar solo to last 24 bars?” The biggest danger is that people who aren’t familiar with the mixing or mastering process will think AI can do as good a job. Well, it may do a job that’s “good enough,” but that’s not the same as the undeniable artistry of a world-class engineer.
Which brings us to the next dividing line: good enough vs. striving for excellence. Yes, AI can sometimes deliver mixing and mastering choices that are “good enough.” But it’s incapable of self-reflection. By definition, AI thinks that what it offers is the right answer, because after analyzing a zillion files, that specific answer was chosen more often than not. But it wasn’t always chosen.
That’s what separates the artists who think, reflect, revise, tear down and build back up again while being guided by their vision. AI doesn’t have a vision. You can tell it to have one, but it won’t second-guess itself the way you will if you find yourself straying outside your intentions and need a course correction.
• Jerry Garcia Venue/Museum Highlighting Live Sound Planned
However, AI has real potential in the studio for accelerating workflow. When AI helps me deliver higher-quality work faster, that’s great. Tools like stem extraction, tedious noise cleanup and diagnostic tools are welcome.
Sure, if AI could replace my judgment, I’d see it as a threat. But AI can’t replace mic placement, taste, arranging, performance coaching, mix decisions, or production dreams. It can easily do “how to,” but it struggles with “what if.” Of course, AI could suggest mic placement, based on information (accurate or not) scraped from the web. But can it take into account the room? The artist’s state of mind that day? What happens when mics sum to mono?
Still, despite AI’s potential in the studio, generating music with AI strikes me as the audio equivalent of elective plastic surgery (i.e., surgery done for non-medical reasons). Plastic surgery can appear mostly real and may even look attractive. But sometimes, the quest for “perfection” takes the humanity out of someone’s looks. Similarly, we all know the dangers of polishing a production “to perfection” until it loses its soul. But if a production starts without a soul…is there any real value to “perfecting” it?
• • •
P.S. — Lest anyone think I am overly anti-AI, here’s a true story: ReSing has one voice option whose sonic character is very similar to my late wife’s voice. I miss her vocals so much, but when I needed a vocal part she would have done, I sang a part the way she would have—and for a few precious moments, I could almost hear her by my side again. AI made that possible. Does this mean a singer lost a gig? No. Only AI could have done what I wanted.
