LANDR might be the biggest leap forward in DIY recording technology since the development of home multitrack recorders. For decades now, artists could record, overdub, mix, and reproduce music from their basements, but mastering has remained the job of trained audio engineers working in expensive studios. This is because mastering, while simple in theory, is a vastly complicated process that requires an impeccable ear. It's what makes the difference between a good amateur recording and professional-level audio that's ready to please ears everywhere.
The founders of LANDR, who also brought us MixGenius, claim that their software automates this process with an algorithm that emulates the human ear and adapts to users' tastes and preferences. It developed out of research conducted at Queen Mary University of London that set out to demystify mastering and understanding it on a tangible, scientific level. The results are top secret, but anyone can make use of the finished product by dragging and dropping a sound file onto LANDR's website. Wait a few minutes, and out comes a mastered version of your home recording.
A lot of artists are taking notice, not least Max Henry from SUUNS, who has incorporated LANDR into his own work. "I'm always doing little tracks either for myself or contract work," he said. "So when the site went up I sent in a few to see what would come back. It's an impressive feat, kind of scary how instantaneous it all was. But it wasn't until I started to get unusually positive feedback from my clients on the LANDR mastered tracks that I was completely sold." The low budget and DIY nature of his work often means he can either mater with LANDR, or not master at all. "It's hard to say no to that."
But LANDR isn't trying to replace human mastering engineers altogether. On their official website, they claim that "the polish and balanced achieved through the subtle adjustments of a skilled mastering engineer is not something we would ever diminish," adding that LANDR files can serve as a useful reference tool for engineers. You also won't hear LANDR on the next SUUNS album. "To be honest, I'm not sure that's even the point," Max said when asked. "Ryan Morey has brought tremendous life to both our albums, and I couldn't imagine a SUUNS album any other way."
So what is this mysterious product, and what should musicians use it for? In order to get to the bottom of the matter, we spoke with Justin Evans from MixGenius, and Thomas Sontag, from MixGenius and Turbo Recordings, about how they developed LANDR and use it in the studio.
Noisey: Who developed Landr, and how?
Justin: LANDR's roots come out of a body of research done over the past 8 years by a research group of very smart graduate students at Queen Mary University of London's Centre for Digital Music (C4DM) (website: http://c4dm.eecs.qmul.ac.uk) who were investigating intelligent audio processing. Then two years ago, a new Montreal technology incubator called TandemLaunch got the exclusive rights to that research, and gave us the funding and infrastructure to co-found the company, with me and one of the students from that program. Since then, we've gained a brilliant CEO who's raised enough money, and has guided the strategy, for us to be able to transform this super bleeding-edge research into commercial reality.
What is it about mastering that's so complicated, compared to other parts of the recording process?
Justin: Mastering requires a few really difficult things - a good ear to hear and analyze things that need to be fixed, an incredible listening environment, very expensive speakers, and a acoustically balanced room, which very few people have. Why? Because your room's shape and acoustics will change what you hear, which can easily lead to really big problems in your mastering. On top of that, you need some serious sound engineering skills. The gear used in mastering is quite complicated to learn. Things like multi-band EQ's or multi-band compressors can do a lot of damage when used poorly. There's a lot of variables and serious science to contend with.
Doesn't it depend heavily on having a good ear? How can you get software to have a "good ear"?
Justin: The question is what is the "good ear" doing? A lot of what that ear is doing is identifying and solving frequency or spectral problems, or understanding what's appropriate processing for the types of song that's being worked on. Fortunately for our project, these are the kinds of problems that machine learning, psychoacoustics and digital signal processing are getting really good at. Think of Shazam. I mean, it used to take someone who worked in a record store and had encyclopedic music geekdom to tell you what that song was or what that DJ was playing. Now you just hold your phone up!
The software is "adaptive," but what does that mean? Does it adapt to a certain user's preferences? To specific projects?
Justin: Our system basically gets new data from every track that's uploaded to it, analyzes it and learns from it. To say more would be giving away our most powerful trade secrets.
Is this a solution for DIY artists who can't afford a top-notch mastering engineer? Or can it also benefit artists with basically limitless tools at their disposal?
Justin: I think it's a solution for both. We are constantly surprised by who is using the service: big record labels, A&R guys, DJ's doing their live sets, big artists who want instant previews of how their songs will sound finished, audiobook makers, video makers, podcasters, people mastering their samples for playing live. It gets used a lot, in a lot of ways we didn't anticipate while we were making it.
On the recording side of things, how have you (Thomas) incorporated LANDR into your work in the studio?
Thomas: We use it constantly. Sometimes it takes the place of pro studio mastering, but more often it replaces the 'self-mastering' process. In both cases it saves us time and money - two resources consistently in short supply at an indie record label. We use it to master live recordings of DJ sets, to boost and re-touch vinyl rips for club play, to give a quick polish to demos and new bounces we want to test out or share. Most impressively, and after careful deliberation, we're also using it on a full commercial release for the first time. We've done extensive, double-blind testing against work from some of the best mastering studios in the business and LANDR has really held its own, in some cases beating out work done by highly respected engineers. It's amazing how good it already is, and it's only going to get better.
Mastering is a notoriously difficult and vague step in the recording process - how does LANDR fix that?
Thomas: Even in the era of DIY music production, where a kid with a laptop can essentially reproduce what used to require million-dollar studios, mastering is, exceptionally, still usually outsourced. There are great tools for self-mastering but amateur's risk serious pitfalls. Without a perfect sounding room and a trained ear, it's easy for people to mess up their mixes. For unsigned musicians, booking a professional mastering studio is a big investment as well as a confusing process (not every engineer is right for every type of music). For small labels, the added costs really eat into the tiny margins. LANDR makes the process foolproof and helps level the playing field for aspiring artists and labels, making it easy and free to preview what mastering can do for their music.
What benefits does it have for you over a human mastering engineer?
Thomas: Firstly: it never gets tired. Secondly, it lowers the cost dramatically. The result is that LANDR makes it feasible to master long live recordings, DJ sets, and unfinished material that producers might want to send as demos or test in a club before committing to a final version. The idea that this once-elusive, mysterious process can now be an immediately-accessible tool in your workflow is both emotionally gratifying and hugely practical.
There is definitely the feeling, having an 'inside' view of this company, seeing the passion and intelligence of the team of developers and engineers involved, and the sophistication of the learning engine behind LANDR, that MixGenius is going to refine this service into something truly revolutionary. If we believe in the promise of AI, then it seems logical that technologies like LANDR might be the way forward not only for convenience and affordability, but eventually for quality. Magnetic coils, vacuum tubes, and tape reels do sound amazing, but are these old technologies really the way forward? Is mining the past is the way to the brightest future?
Greg Bouchard is a writer living in Toronto. He's on Twitter.