You are here

Off The Record

Music & Recording Industry News
By Dan Daley

First mastering went online, now it's becoming automated. Should we be afraid?

Mastering was once the province of the golden-eared pro-audio elite. Since it moved online, however, it's come to look more and more like Craigslist without the sex. There are hundreds of sites offering some version of 'mastering' services, often for as little as $20 a track. What some would consider the most critical stage of a recording can now be had for less than the cost of a decent lunch. And it would likely be far less satisfying, too.

Back in 2006, when SOS Features Editor Sam Inglis looked at the burgeoning landscape of online mastering services in these pages, it was far sparser and populated mainly by well-established mastering facilities, like Metropolis and The Sound Masters in the UK, venturing into Internet commerce. It was around the same time that US facilities like Masterdisk were tweaking the 'unattended' mastering session concept, where clients would send their WAV or DDP files electronically for mastering sessions that would be handled by the staff in between the booked sessions of more affluent (ie. major-label) customers. It was a cost-effective strategy in reaction to a traditional record industry that was running out of money to pay card rates. The unattended session would cost the client less, and while they may not have been guaranteed the 'star' of the facility, they were at least assured that whoever did their mastering was a professional with a long track record and lengthy credit lists. More recently, a quick Google session reveals that it's increasingly the mastering engineer's side of the equation that's unattended, with MP3's likely being run through a Waves L2 or L3 while the 'mastering engineer' is out back having a smoke.

Big Data Masters

But is that a completely fair assessment, in the age of Big Data? At a time when Shazam and other algorithm-based software is being used to not only find hits but to predict and to manufacture them, too, is a robotic approach to finishing off a record perhaps the best way to help avoid finishing off the record business?

Justin Evans seems to think so. The Montrealer's MixGenius site is turning its machine-learning and artificial-intelligence (AI) analytics on the mastering sector, which, along with algorithmically based automated music mixes, bundled into a larger, amorphous (and somewhat sinister-sounding) category he calls 'audio refinement,' he says can come to represent a $1.2 billion market worldwide.

"Our approach to music intelligence is based on a big-data process similar to Shazam, just flipped — it's mastering based on feature analysis,” he says, talking as comfortably in the patois of the Internet startup as in the lingo of pro audio. Evans declines to be more specific about exactly how the algorithmic mastering process works, other than to say it was developed by a research group of faculty and graduate students at Queen Mary University in London, UK, and that it analyses key frequency and dynamic features of a track and then chooses what it considers the appropriate parameters and processors for mastering it. Beyond simple processing, however, it also involves what Evans calls "supervised” machine learning, so that as the processor racks up mastering sessions, it learns from them, with minimal human guidance.

Evans says that this systems approach to audio mastering is presented with utter respect for what mastering engineers do with their ears. He sees it as complementary and in some ways inevitable: as the overwhelming wave of recorded music continues to build, the thresholds of conventional mastering simply cannot accommodate the massive amount of material that's pouring out of home studios, laptops and even mobile phones. What Evans is essentially saying is that all this music can either go out into the world without a finishing polish to it, and by implication continuine to reinforce the lo-fi culture of the digital masses, or it can get that last sheen applied robotically and inexpensively, thereby raising the bar for sonic quality.

The True Cost

It is cheap: the MixGenius site will master an MP3 for free; users can get up to four uncompressed tracks finished for C$9 per month, or unlimited tracks for C$19. The next logical question — is it any good? — is harder to address. With Spotify's inventory at over 20 million songs (many of them indie), and YouTube's seemingly infinite number of tracks (many of which are downright dodgy), almost any sonic improvement would be welcome. Apple are already using automated processing, including levelling, as part of their iTunes Radio service. In large part it's there as a counter to the loudness war, which continues to be waged via artists' insatiable demands for ever-louder records, as well as the proliferation of semi-pro 'mastering' services, who are all too happy to help them achieve just that, because if they don't their competitors will. Perhaps an automated approach, at least regarding functions such as level normalisation, might be a good idea. But how far to take it past that? EQ? Compression? At what point does the omission of human ears cross a line? To Evans, even the question itself is so 20th century. Instead, he compares it to what Instagram can do with digital photography.

He may be right. Artists at all levels have for years been putting stems of their songs online for consumers to remix. Mash-ups have been around even longer than that. Putting music into the hands of others (even machines), whether for remixing or mastering, could be viewed as simply part of a culture that no longer wishes to own anything but still have access to everything. Why should music be any different? For music that's professionally produced and intended to compete at the upper echelons of the marketplace, there will be the facilities and the talent needed to make it as good as it can be. For the rest, which might otherwise come into the world a bit ragged, an automated final polish can help make it as good as it needs to be.   

Published September 2014