Human versus AI mastering
Artificial Intelligence (AI) is a function of most of the online digital technology we interact with. This includes music production. There’s AI for fixing glitches and noise in audio recordings, AI for out of tune instruments and vocals, AI will even write you a song and you can choose a virtual vocalist.
AI has also landed in the world of music mastering. There are online services that will master your track(s) with little or no interaction from the user. Just upload your music to a website and, hey presto, ready to release masters will be provided, and sometimes for FREE!
There are also AI tools that are being used by musicians and mixing engineers who want to take ownership of mastering. Izotope Ozone (and there are many other examples of software in this realm) has a ‘mastering assistant’ which will listen to a portion of your song and provide suggestions for EQ, compression, stereo imaging, and final limiting, as well as many other creative tweaks. AI in this regard is comparing your track sonically to thousands of other tracks in the same genre.
AI will try to match the tonal balance and feel of your track to those that have gone before. This type of software will often advise to use the mastering assistant as a starting place, rather than being the “one-shot” solution to provide a ready-cooked master. The rest is up to you, but you could simply export the suggested master and the job would be done.
So, does this all spell the end of the road for mastering engineers of the human variety? For some people it may mean just that. After all, it’s likely to be very tempting to bypass the need for an extra stage (and expense) in the music production journey. Software which provides a mastering solution, like Ozone, will of course involve an initial investment, if that’s your favoured route, but it’s then quick and easy for almost anyone to produce a software-led master.
There’s also AI embedded in many individual plug-ins, especially those for compression and limiting, that could be used in the mastering chain. In this regard, AI is a useful tool in a human-led process.
Will AI-led mastering give you results that are good enough? Well, it all depends on what good enough means. You might be happy with mastering via a fully- or partly-automated service. But, and this is the big but, how will you know if your music has reached its sonic potential? If you are a seasoned mastering engineer you’ll probably know straight away, but then you wouldn’t be using this route in the first place.
Most people will never know. Yes, the music might sound good, but there’s every chance that it could sound better. This is essentially why it’s not the end of the road for mastering engineers. Indeed, AI mastering has been around for over a decade (e.g., LANDR) and the business of human mastering seems to be going from strength to strength.
An experienced mastering engineer will get you a little bit more, and possibly a lot more.
It’s also important to mention the role of musicians in directing the process and recognising the value in human-led mastering. Musicians may have very specific instructions for their mastering engineer. There may be reference tracks by the client or other artists to compare to in terms of the target tone and feel of the master.
Yes, AI can do some of this for you, but to get it right will usually involve an experienced mastering engineer to work out the, often specific, order of audio processing to achieve the desired goals. I’ve noticed that some AI masters may achieve a good result for the desired tonal balance, but the music might, for example, still lack some clarity or there might be harshness at certain audio frequencies.
Is AI mastering likely to work better on individual tracks than on albums? Not necessarily, but there’s probably more chance for sub-optimal results across an album. AI mastering may result in less sonic cohesion between the tracks than via human-led mastering.
In summary, you might get lucky and end up with genuinely good results via AI, but 99% of the time a mastering engineer will likely give you improvements over AI.
Perhaps you’ve tried AI mastering online or used an automated process via software. If you’d like to compare the results with those from a dedicated mastering engineer, using premium analogue equipment, then complete the form via the link below to get a free test master.
