On Benson Boone's tour this year, front-of-house engineer Joey Diehl started getting a vocal sound he'd never mixed before. But the change came from software, not from a new microphone or a different console.
"[It] gives me a vocal sound like no other," Diehl said. "It's basically studio quality vocals now."
Diehl is mixing front of house for Boone's tour, and the tool he's crediting is Source Intelligence, a machine-learning-based vocal isolation system L-Acoustics announced on 21 May 2026. It runs on the company's L-ISA Processor II and strips out up to 40 dB of everything that isn't a voice, including PA bleed, stage noise, and crowd noise, at less than 8.5 milliseconds of latency over MADI. His tour is one of at least six named productions already running it in front of paying audiences.
AVIXA has already covered beamforming through Audio-Technica's ATND1061, a 9-inch by 9-inch ceiling array microphone that focuses on whoever's talking in a quiet conference room and filters the rest, replacing the oversized ceiling tiles that kind of coverage used to require. Source Intelligence takes on a related but different problem. It listens continuously inside an acoustically hostile live venue and separates voice from noise in real time, closing in on the trade-off touring engineers have fought for decades–gain before feedback. But this happens from the software side rather than the pattern-based one.
What The Software Actually Does
Ryan John, L-Acoustics' Director of Product Management for Software, described the shift in the company's own announcement.
"Source Intelligence removes that trade-off entirely," he said. "The machine is continuously listening and working."
The system processes up to four vocal channels at once and works across handheld, headset, lapel, and distant-pickup microphones, according to L-Acoustics' press materials and TPi Magazine's coverage of the launch. It ships as a licensed software subscription through L-Acoustics' All Access program rather than as new hardware, which is itself part of the pitch. A venue or touring rig that already runs an L-ISA Processor II doesn't need a truck full of new equipment to get it, just a software key.
TPi described its capability as being built to "continuously identify the voice in a microphone signal and remove everything else." That continuous, always-on quality is the actual technical departure from beamforming, which listens in a fixed direction rather than picking a signal apart from the noise around it moment to moment.
The system also needs at least three L-Acoustics Amplified Controllers on the network to validate and run, and it ships with power supply, network audio, and processor redundancy built in as standard. Audio routes in over MADI, Milan-AVB, or AES, and engineers get separate controls for refining the high- and low-frequency edges of the isolation rather than a single on-off toggle. That's the kind of failover and routing flexibility touring audio equipment has needed for years, built into a processing layer that used to just be an EQ curve.
Not Just For Concert Stages
L-Acoustics doesn't pitch Source Intelligence as a touring product alone. Its own product listing names four intended sectors: touring, house of worship, theatrical, and corporate events, and describes the house of worship use case in almost the same language as the touring one: gain-before-feedback control without compromising on stage noise.
The corporate events case is the more interesting one for AVIXA's own readership. L-Acoustics frames it around multi-presenter scenarios, handhelds, lavaliers, and headsets mixed in the same room, which sits close to the exact conference room problem that beamforming looks to solve. If this approach scales down from a Harry Styles arena show to a boardroom, the two technologies end up competing for the same room before long: a spatial microphone picking a direction against a continuously listening model separating signal from noise, on hardware that’s a lot smaller and cheaper than an L-ISA Processor II.
A Season Of Named Deployments
Source Intelligence isn't a demo unit doing the rounds of trade shows. It's already mixed into a season's worth of major tours, each running through a different production company.
L-Acoustics named two beta partners at launch: Solotech, running it on Harry Styles' tour, and Clair Global, running it on both Benson Boone's and Rosalía's tours. FOH Online reported two more named engineers since launch, on productions where the production company wasn't specified. Jamie Pollock deployed it on Nine Inch Nails' Peel It Back tour, and Sean "Sully" Sullivan, who mixes Bruno Mars, was among the first engineers to run it on a major touring production.
That's five separate artists, two confirmed production companies, and three named front-of-house engineers, all inside the technology's first several months in the field, before it's even had a full year to spread by word of mouth.
A Different Stage Entirely
The deployment that breaks the pattern isn't a tour at all. It's Broadway.
PRG ran Source Intelligence on Dog Day Afternoon, and Tony Award-winning sound designer Cody Spencer used it to solve a version of the same problem theatrical sound has always had: keeping an actor's live vocal intelligible over a loud stage without the actor sounding artificially boosted.
"I'm very impressed with how much headroom it gives me," Spencer said. "I was able to get crisp, clean vocals in any scenario, including very loud stage environments."
Adapting a touring product to a fixed eight-shows-a-week theatrical run is a genuinely different test than a stadium tour, where the whole rig gets rebuilt from scratch at every stop with a full crew reload. A Broadway house keeps the same room, the same seats, and the same acoustic quirks for the length of the run. Spencer's account is the only public record so far of how the system performs under that kind of sustained, unchanging use.
The Trade-Off That Used To Be Physics
Gain before feedback isn't a software problem by nature. It's a hard acoustic limit, the reason engineers pull vocal mics back defensively, ride faders all night, and treat "loud enough to fill the room without howling" as a permanent trade-off against clarity.
An always-on machine learning layer that claims to separate voice from everything else in under 8.5 milliseconds is a real claim against a limit that used to just be physics, not an incremental EQ trick. The honest question a trade audience should ask is whether that holds up identically across a stadium tour, an arena show, and a fixed Broadway house, or whether the touring-versus-theatrical split already suggests the system performs differently depending on the room. Every account on the record so far, though, comes from L-Acoustics' own named beta partners, not an independent or skeptical source, worth keeping in mind before treating five satisfied tours as proof the trade-off is solved for good.
There's a real constraint worth stating plainly too. Source Intelligence can't run at the same time as L-ISA or Ambiance on the same processor, according to L-Acoustics. It is an exclusive operation mode rather than one more layer stacked on top of an existing mix. That's less a flaw than a reminder that continuously listening and deciding what counts as voice is expensive enough, computationally, that the processor doing it can't be doing something else at the same time.
What Comes Next
Since its May 2026 launch, Source Intelligence has moved from a product announcement into a working part of the touring and theatrical audio stack, running on tours for five different artists across at least two named production companies, and on a Broadway run that has nothing to do with a stadium at all. If the trend holds, the interesting story a year from now won't be which tour adopted it first. It'll be whether "the machine listens and decides what's voice" becomes the default assumption for how a live vocal mix works at all, the same way beamforming quietly became the default assumption for how a conference room microphone works.