Will AI mastering replace real mastering engineers?
Making music is the most human and human occupation in the world, but in recent years, artificial intelligence has also begun to play on this field. Algorithms are involved in almost every stage of musical production – from creating beats on drum machines to writing the melodies themselves, which can hardly be distinguished from people’s creations. Now AI is starting to work in mastering, and this raises the question of the need to use qualified specialists in one of the most difficult areas of music production.
Mastering is the final step in post-processing audio, and its task is to make the final track sound balanced no matter what you listen to: even in Spotify, even in iTunes, even on disk. The purpose of mastering is to create a solid, coherent musical experience, leading the listener from song to song. The process combines both accurate calculation and the personal approach of a mastering engineer. With a good mix, the specialist will easily understand what kind of sound the performer is aiming for and will help to achieve it. Without mastering, the tracks would be more faded and quiet. Engineer Ian Cooper notes that “mastering is a bit like a photo: you can make the sky bluer, and the grass is greener.” And mastering can also be expensive. Depending on the experience of an engineer, the cost of his services can range from hundreds of dollars to tens of thousands per track – it all depends on the listening and talent of a specialist. For indie artists and musicians, the amount can be overwhelming.
However, in recent years, automatic algorithms have evolved quite seriously and now promise artists access to professional mastering for much smaller amounts. Some use deep data analysis networks that process incoming data for a long time, others are based on a carefully designed human signal sequence translated into a program format. But despite the different principles of work, they have one task: to issue a ready-made mastered track after a couple of mouse clicks.
Landr became the most popular among such services. Fill the song, give the Landr algorithm to analyze it and choose from three options, how bright you want the effects in it. After that Landr gives the result. The approach is not the most flexible – if you do not like the result of Landr, it will not be possible to ask you to tweak something, unlike working with a living person. Portal ArsTechnica in 2016 published a review in which he called Landr’s auto-testing “auto boxing”, but others are not so critical and say that the service copes with its tasks). And in theory, with each processed song, the Landr algorithm becomes smarter. “In 2017, we conducted a series of blind testing with large labels and professional mastering engineers, and some preferred Landr to larger and well-known mastering studios,” said service director Pascal Pilon (Pascal Pilon) portal The Verge.
Some people worry that AI mastering will take jobs from live engineers, but London-based Streaky engineer compares it with buying a suit from a shop window, and not to order (see video below). Those who are important in fitting the suit for themselves or the quality of the fabric, in any case, will go and order the suit, but for most, the more affordable option is quite acceptable. The company already offers a set of plug-ins under the name Ozone, and in 2017 introduced the Master Assistant feature. The assistant does not do all the work for you – he gives starting points that can already be adjusted as you see fit.
Due to this, producers can make decisions based on the choices that AI made. “There is no competition with people,” a spokesman for iZotope said, “for professionals, this technology simplifies the work of cleaning material that takes a lot of time, and they can concentrate on the creative part.”
The founder of MajorDecibel Adam Love (Adam Love) agrees with this statement: “This is not a substitute for mastering from a mastering engineer. Engineers can give performers feedback, work in a particular style, adjust or highlight some specific points. Man is slow and methodical, but unlimited. Automation works faster, but in a much narrower framework. ”
So this is just an affordable alternative that will help the music sound better. “We do not select jobs and change the industry, but rather create a new niche, allowing those who cannot pay for high-quality mastering to get it,” said eMastered Representative Colin McLaughlin. “But for the best mastering, of course, you need to contact a real mastering engineer.”
It is difficult to say whether the AI will ever learn to listen with the same diligence as a person, but he probably does not need it. Modern AI mastering is already quite complicated and even now is an excellent option for many musicians.