Before vocal cloning, text-prompted song generation, and musical style copy-paste, there was AI mastering. Back in the halcyon days of 2014, mastering became the canary in the coal mine of the music industry when LANDR launched an automated mastering service.
The streets are in panic, industries are collapsing, outboard motor racks are being set on fire, and a band of vicious mastering engineers roam the streets seeking revenge against the new AI overlords. just kidding. In fact, the initial launch was more wailing than shocking. There were a few complaints and some wringing hands, but everyone quickly got back to business as usual. But AI fears are back, bigger than ever, and perhaps this time they’re justified.
Mastering has long been described as the “dark art” of music production, but in theory it all seems pretty simple. After a song is recorded and mixed, mastering his engineers perform a series of relatively minor sonic adjustments to finalize the track for release. A little he EQ, stereo imaging, just the right amount of compression, etc. You’ll have songs ready to stream, and ideally they’ll sound smooth on anything from cheap earbuds to high-end studio monitors.
Of course, achieving professional-quality results is much harder than it sounds, and some mastering engineers have attained near-legendary status for their ability to perfect their tracks. For this very reason, mastering has always had an exclusive feel, or at least it has.
For those of you who don’t know, the power of AI technology has grown exponentially, and automated mastering services have made it possible to achieve results that were impossible just a few years ago. . And the field is even more crowded, with Soundcloud, Cloudbounce, Roex and others joining his LANDR to offer affordable mastering with increased quality levels.
One of the most interesting startups on the block is Masterchannel, a Norway-based company founded in 2022. The company uses an approach known as “reinforcement learning” to achieve results.
“Many of our competitors base their technology on huge datasets containing millions of songs,” says Christian Lyngstad-Schultz, co-founder of the company. “When you upload a track, it finds the most relevant songs from your dataset, applies those presets, and then tweaks them.”
In contrast, Masterchannel says it worked with thousands of human engineers to achieve “benchmark results,” i.e. objective standards against which AI output can be measured. Once the track is uploaded, Schultz says, the algorithm works like a mastering engineer. Experiment with different approaches and receive positive or negative feedback on decisions made. “With this approach, instead of working with one mastering engineer, thousands of engineers can work on his one automated system.”
One of the main advantages is the genre-agnostic system. “There are new genres popping up every week,” Schultz points out. “Usually you have to pick a whole new dataset and tune it to its sound. You just have to tweak the parts and everything becomes much more flexible.”
Schultz said reinforcement learning is central to how MasterChannel works, but the company also uses a general dataset of songs for deep learning, and the two approaches complement each other. “You can use reinforcement learning and add a deep learning model on top of it,” says Schultz. “This could reduce costs, further accelerate the process, and have many positive effects.
While AI companies are experimenting with different technological approaches, customers are also busy leveraging these services for an ever-wider range of uses. Daniel Rowland, LANDR’s head of strategy and himself a seasoned audio engineer, said, “I’m not sure how people will use what I build, as opposed to how I thought it would be used. It’s really interesting to see how they’re doing,” he says. “For example, many people use his LANDR for mix references, individual musical stems like drums and synths, movie sound effects, and more.”
The primary user base for AI mastering is also shifting from bedroom producers and early career artists to people working in all areas of the music industry. “Some famous composers use LANDR simply because they need to cycle through cues and revisions very quickly,” Rowland says. Similarly, Masterchannel was originally launched as a direct-to-artist service, but is now used by many established artists, producers and even mastering his engineers, says Schultz. “It’s every area,” he says. It’s been used for prototyping as a kind of his AI co-pilot in production, but also for the final release. It’s also been of interest to engineers looking to “replicate” their own production style in a similar way to what we’ve seen with Drake. ”
Perhaps one of the biggest changes with the longest lasting impact is the use of AI mastering services by the industry’s largest customers: film and television studios, game developers, and record labels. “It’s already happened,” says Schultz. “We are currently testing with some of the biggest companies here in Norway. What matters to them is the amount of material and how quickly it can be finished, which can reduce costs. .”
Now back to the rampaging gang of newly surplus mastering engineers. We don’t expect an extinction-level event to happen overnight, but changes are clearly happening. As the company that started the whole conversation, LANDR has actually made significant moves to push its commitment to human engineering, rather than siphoning business. “There are dozens of mastering engineers on the LANDR website that you can hire,” Rowland emphasizes. “We’re very open that we don’t want to replace mastering engineers. We just want people to find what works best for their project within their budget, AI or not.”
Mr. Schultz’s take on the issue is a little more blunt. “I think there will always be a need for mastering engineers for people who can afford it. could result in fewer jobs.”
While this may be a bitter pill that is difficult for some to swallow, Schultz’s logic makes sense. Subpar skills are sure to struggle to compete with consistent (and affordable) mastering, but studio academics will likely still maintain their niche in the upper echelons of the industry.
There is one problem with this scenario. A highly skilled mastering engineer does not arrive at the birthing suite fully prepared. It doesn’t work for everyone at first. Everyone needs practice and experience to achieve the skills of Bob Ludwig or Bernie Grundman. How will the next generation of engineers develop their skills in an environment where very few are paid? That is an open question.
However, all this assumes that future generations will also have mastering engineers. “I think a lot of what we call mastering can diminish over time,” Schultz speculates. If you look at the new generation coming up, they don’t necessarily know or care about the term ‘mastering’. It’s just about making a professional sounding track. So many of these production steps I think he merges into one. ”
The possibility of simply doing away with mastering as a separate step in the production process isn’t all that far-fetched. Schultz says the ultimate goal of MasterChannel is to make the process more seamless and simpler. “Long-term, we are building technology that can be applied to all domains where audio is used. Playing videos on YouTube, narrating games in Unity, releasing music on Spotify. , I need a button that automatically detects and optimizes the audio type.”
This is an amazing vision of the future with undeniable positives. Roland points out: “Labels are mastering and releasing content that would never have seen the light of day before because of the exorbitant cost of mastering.” Providing affordable mastering options for other artists is equally beneficial. And more audio is being produced for a wider range of media than ever before. Simply put, we need an industrial-scale solution to this ever-growing flood of content.
While considering these advantages, it’s worth taking a step back and considering what the purpose of music mastering really is.
Generally, a well-mastered track aims to meet the “industry standard” in terms of loudness, EQ, stereo imaging, etc. However, industry standards have changed a lot in his decade. A song mastered in the 90s or early 2000s will definitely sound different than a song mastered in 2020. Why? Because someone broke the rules. An engineer did something different from the standard, others heard it, liked it, copied it, and a new industry standard was born. Alongside genres, styles and instruments, mastering is central to how the sound of popular music evolves.
The question is, can we expect “one-click” AI mastering to creatively break the rules? Will any songs be mastered in 2030? sound Is it noticeably different from the one mastered in 2060? I hope you can code a little creative chaos.
