With the release of the eagerly anticipated L1, the flagship of the L Series, and its companion cardioid subwoofer, CS1, L-Acoustics made headlines across the sector. However, it was the reveal of the less predictable Source Intelligence – a real-time vocal enhancement technology that isolates the human voice from background noise using a unique source separation algorithm – that turned heads in the TPi office.
This isn’t L-Acoustics’ first exploration into source-based technology. That came with the release of HYRISS, and subsequently L-Acoustics DJ. Taking an input source and performing real-time stem separation, allowing a creative or artist to create an object-based mix without having to significantly change their workflow is one of the major challenges of object-based mixing. DJs, however, make up a small share of the market, whereas the voice – addressed by Source Intelligence – is present in almost every single system.
Ryan John, Director of Product Management, Software at L-Acoustics and a touring sound engineer with more than 27 years’ experience, has encountered the limitations of live sound first-hand. As a mixer, he has dealt with compressing audio, fighting feedback and contending with artists performing directly in front of the PA – a convention of modern productions.
“This is the direction that shows are going in nowadays, it’s about creative freedom and show design being bigger and better, so the toolsets that we have as mixers haven’t kept up,” John said, speaking exclusively to industry trade press ahead of the software’s commercial release. “As we started to go down this path of machine learning tools, we worked with the team to find a solution that could deliver pristine audio quality at low latency – and could do that reliably, so that you just turn it on and it works,” he added.
According to John, Source Intelligence delivers “clearer, more intelligible vocals without the need for aggressive EQ, gates, thresholds or reactive processing”. This results in increased gain before feedback and a clearer vocal mix that doesn’t lose emotional impact or the intimacy of the performance. With greater vocal isolation and audio quality, John explains, engineers gain the freedom to mix more creatively, while performers and presenters have the freedom to move anywhere on stage, even in front of the PA, without sacrificing clarity or presence, or increasing the risk of feedback.
Eager to demonstrate the creative capabilities of Source Intelligence, John was one of the first engineers to pilot the solution two years ago for Laufey’s A Night at the Symphony Tour. “Being able to use it was incredibly freeing as a mixer, you were no longer reliant on how much compression you could apply on a vocal or even the amount of effects you could apply to them,” he added. “It’s hard to go back and do shows without it!”
Running on the “trusted” L-ISA Processor II Core, the solution is designed to remove crowd noise, instrument bleed, PA spill and room reflections. Since the creation of L-ISA, the processor has also been used for other applications, including L-Acoustics DJ, Ambiance and now Source Intelligence, making it a platform rather than a single-use processor. Users can run Source Intelligence in and out of the L-ISA Processor using MADI insert, network insert over AVB Milan and/or two channels over AES as an insert.
Early in the R&D process, L-Acoustics contacted acclaimed mixing engineers including Michael Parker (Harry Styles), Joey Diehl (Benson Boone), Jamie Pollock (Nine Inch Nails), Sean ‘Sully’ Sullivan (Bruno Mars) and Johnny Keirle (Rosalía), among others, who were all on the road with challenging production designs, to deploy the solution. Many, John proudly reported, said their shows “wouldn’t be possible to get through” without Source Intelligence.
“When you go into a new realm, things like source-based processing, the only way to do this well is with a network of people and ears that you and everyone else in the sector trusts,” John reported, praising the influence of his contemporaries in shaping the development of the solution. “The feedback has been positive, meeting expectations of engineers. This has replaced five or six tools in the engineer’s toolbox. Crucially, Source Intelligence is accessible to everybody, from corporate boardroom to headline stadium tours. Since it’s running on an existing processor, any vendors that have the processor can run Source Intelligence on an L-Acoustics system, they simply purchase a licence for it.”
John was keen to explain what makes Source Intelligence tick under the hood. “We’ve built a focussed algorithm, trained using machine learning, that solves a very specific problem in a completely new way,” he explained. “Rather than reacting to signal level like a traditional threshold, it works from an awareness of what a signal actually is – opening and closing based on the voice itself.
John concluded: “Our goal is to help people put on better events – to strengthen the connection between artists on stage and audiences in the room. What Source Intelligence does is give engineers back their headroom, so they can focus on creativity rather than firefighting. As one engineer kindly put it, ‘Less fixing, more mixing.’”
Words: Jacob Waite
