I spent three days at IBC walking the audio halls, and the gap between what’s being sold and what’s being bought has never been wider. Every booth had an AI-powered something—AI mixing, AI cleanup, AI dialogue replacement. But the engineers I talked to in the coffee lines? They were mostly trying to get their existing Dante networks to stop dropping packets.
The industry has a case of solution-in-search-of-problem syndrome, and it’s not pretty.
What People Actually Asked About
I stopped by a major console manufacturer’s booth to see their new flagship. The demo was impressive—touchscreens everywhere, motorized faders that moved like they were alive, more DSP than you could shake a stick at. But the first question from the visitor next to me: “Can I still get a plain analog insert point, or is everything internal routing now?”
The answer was complicated. And that’s the problem.
Broadcast engineers are conservative for good reason. When a live event is going out to millions, you want signal flow you can trace with a flashlight and a cable tester. The trend toward software-defined everything—while powerful on paper—creates abstraction layers that make troubleshooting harder, not easier. I watched a demo of a “cloud-native” mixing system that required three different logins and a web browser to adjust a microphone preamp. The engineer next to me walked away muttering.
The Real Innovation: IP That Finally Works
For years, Dante and AES67 have been promising to make audio networking plug-and-play. At this IBC, it felt like they were finally delivering—mostly. The new switches designed specifically for AV traffic were everywhere. I talked to a guy who’d migrated an entire regional broadcaster’s radio plant to Dante over a weekend. That would’ve been unthinkable five years ago.
The difference is in the details. Earlier generations of IP audio required you to become a network engineer. Now the switches come pre-configured with QoS rules that actually work. You still need to know what VLANs are, but you don’t need to spend a week reading Cisco documentation.
But—and this is crucial—the reliability is still not quite there for mission-critical live. Everyone I talked to running IP audio at scale had a backup plan. Usually analog. Often patch bays. There’s a pragmatism that the marketing materials don’t reflect.
The AI Reality Check
Every booth had AI features. Dialogue isolation that works in real-time. Automatic loudness correction that claims to be transparent. Voice cloning for dubbing that supposedly passes the “mom test.”
I tried the dialogue isolation on a noisy clip I’d brought with me—actual location audio from a problematic shoot. The results were… fine? It removed the background noise, mostly, but left that swirly, phasey quality that tells your ears something’s been manipulated. Usable for news, where intelligibility matters more than fidelity. Not something I’d want to mix into a drama.
The voice cloning demos were more impressive, technically. Feed it thirty minutes of someone’s voice and it can generate new lines in their timbre. The use cases being pitched—localized advertising, emergency re-records, accessibility—are legitimate. But the licensing questions are a mess. Who owns the cloned voice? What happens when the talent’s contract expires? I asked three different companies and got three different answers, none of them convincing.
The Hardware Paradox
Here’s what surprised me: hardware is back. After years of everything moving to software and plugins, there’s a renewed interest in physical controls. The new control surfaces aren’t replacing DSP—they’re remote controllers for software mixers—but they feel substantial. Real faders. Real knobs. Tactile feedback.
I think we hit the limit of what touchscreens can do for audio. There’s something about mixing that requires muscle memory, the ability to adjust multiple things simultaneously without looking. You can’t do that on a glass surface. The manufacturers seem to have finally accepted this.
The sweet spot right now seems to be hybrid—DSP and routing in software, control in hardware, networking over IP but with analog backup paths for the critical stuff.
What I Didn’t See
Immersive audio for broadcast was conspicuously quiet. Lots of discussion about Dolby Atmos for streaming, but the broadcasters I talked to were mostly still figuring out stereo. The infrastructure for height channels—additional mics, more complex monitoring, bigger mix stages—isn’t there for most operations. It’s a technology looking for a business model.
Also missing: any real solution for the talent gap. The industry is losing experienced engineers faster than it’s training new ones. Everyone knows this. Nobody’s figured out what to do about it. The AI tools are being pitched partly as a solution—you need less experienced operators if the machine handles the hard parts—but that’s a dangerous bet. You still need someone who knows when the machine is wrong.
The Takeaway
IBC always has a vibe, and this year’s was cautious. Money’s tight. Buyers are skeptical. The flashy demos got polite applause, but the real interest was in reliability, backwards compatibility, and not having to retrain the entire staff.
If you’re making decisions about infrastructure right now, my advice would be: wait on the AI features, invest in good IP networking that plays nice with analog, and keep your best engineers happy because they’re getting harder to replace.
What about you? Were you at IBC this year? I’m curious whether my read matches what you saw—or if I spent three days talking to the wrong people.