AI Mastering: Why It Isn’t Worth It…

It seems like AI Mastering options are everywhere these days. Today we take a look at why it doesn’t replace a human.

AI Mastering: Plugins Galore…

---

Audio mastering is the most important step in recording and producing music. It often gets looked over, and that is because mastering is hard and takes time to learn.

Just a few short years ago, there was only two ways to get your songs mastered-you paid for it (probably an enormous fee) or you did it yourself. But there are tons of plugins these days that make mastering a one-step process.

But are these AI mastering plugins worth it? They certainly seem inexpensive at first glance, and it takes away hours of work or truckloads of money. What are the downsides?

These AI mastering plugins make some bold claims, and I happened to have the opportunity to try one, 8 years ago in early-access. It did not go very well, but we will get to that.

Most home recording artists probably have a suite of plugins that they use regularly, plugins that might automate a lot of the recording process. Don’t get me wrong, many of these plugins can be useful!

But we are not talking about something like amp sims, or compressors that circumvent audio hardware equipment that can be expensive. We are talking about an entire process! It would be like letting AI mix your song!

Like those silly commercials that graced every DVD about pirating copies. You wouldn’t let AI record a song for you, would you? You wouldn’t let AI rob a bank!

But in all seriousness, AI mastering plugins are getting more sophisticated. But does it actually work to replace a real-life human? Today we are going to look at early versions, as well as newer variants, compared to a mastering engineer.

Should we heed the warnings about the dangers of AI? Just how far can deep learning go when it comes to being creative?

I speak in jest, but fully AI music is everywhere. The algorithm for deep learning has become so accurate, it can write entire songs based on your description and lyrics.

Tell the AI model to write a song that sounds like Deftones, and it can probably get it right with a little tweaking on your end. The AI tools can create almost any genre you ask it to make.

Music has always been a way for humans to express emotion, whether it be sublime glee, the deepest pain, or extreme anger. Certainly, AI cannot express itself.

So why would you let it perform the most important step in production? Mastering is the final step in production, meant to enhance your music for listeners.

But let’s dive deep, and take a look at where this began. It has been a long journey, but these tools cannot replace a human…yet.

---

LANDR: One of The Firsts

LANDR was first offered to me when I uploaded a song randomly to Soundcloud. It was mentioned that the program was in early access, and that I could try it for free.

I signed up, since I did have some “actual” songs that were going to be produced professionally in the studio and they were already mixed down. So I decided to try it out, and see how it worked.

Back then, AI mastering was very archaic. This was in early 2016 or so, a time when AI was still mostly regarded as science fiction. So I decided to master a track the hard (but human) way, and then master that same track in LANDR.

The results were mixed, to be frank. The version that I had mastered myself took about 3 days, and it was the absolute best I could do at the time. I was pleased with my mix, as well as the final master I created.

The LANDR version was similar to mine in volume, and I imagine LANDR started as a sophisticated limiter since it was basically brick wall compressed with slight EQ tweaks.

LANDR had missed some of the smaller adjustments I made with my master. I did a massive amount of work on the EQ, as well as some added tape saturation on the drums. LANDR missed these things completely, because it was not in the program’s toolbox.

But that was years ago, and LANDR has become much more complex. Now the plugin has a built-in EQ, with a recommended version of your mix. I imagine the plugin has a lot going on in the background to make your song sound professional.

But LANDR was only the first, and now even major DAW companies support a “mastering assistant. So if there are so many plugins that master a track for you, what are you missing by using them?

---

Mastering Is Personal

Working with mastering engineers over the years, I found that most act like an extra “band member” or producer. A good mastering engineer will know what style you are going for with your music. They may even be a fan of your music.

On the other hand, mastering engineers are certainly expensive. I personally charge $30 an hour and that can rack up a bill rather quickly. So I see the appeal of AI mastering for bedroom producers on a shoestring budget.

I suppose it depends on how seriously you take your music, and just how far you are willing to go for your art. Your favorite bands use mastering engineers, and once that band finds one that works for them? They hold onto that engineer, and use them on every album.

That personal touch does not have to be from a mastering engineer fi you record at home. Learning how to master your own songs can be a daunting task, but it will pay for itself over the years, with dividends.

“Nothing worth anything is ever free or easy”. At the sake of getting philosophical here, becoming a great musician was not easy either, was it? Learning how to record at home is difficult, as well as mixing.

Mastering is the final step before your audience hears your music. Do you really want to put that trust into an AI learning tool that has simply “studied” mastering?

---

AI Mastering is Quick & Dirty… And I Tested It.

Since it had been a while since I tested AI mastering, I decided to do that today with a track that I have mixed down. I have already mastered it the traditional way, for our podcast.

There are a few programs that allow you to “test” your mix for free, and I compared it to my master that I “did by hand”. There were a few that were better than others, but like AI music it was lacking nuance and complexity.

So I tried to figure out exactly how it was working. What elements did the ai mastering plugin use to achieve the results? Well, it was a little hard to tell what was going on behind the scenes.

Programs like Emastered and Cloudbounce produced similar results to LANDR. They produced a good volume for streaming, as well as some added EQ to the track.

But it missed a few things that my personal master featured. I added saturation to the top end that enhanced the vocals. I also put a slight compression on the track that added some low end that was missing from the AI mastering.

So maybe that personal touch is the main thing that is missing. I am intimately familiar with the track because I made it. If it were a client, I would have a notebook full of wishes from the artist for the final master.

I have several notebooks from old clients, and they have page-long details about each song. They chronicle how each song should feel and sound, something you cannot convey to AI.

---

How AI Mastering Works

If you are reading and still think that AI mastering is a good idea versus learning it yourself or taking it to an engineer, let’s explore how it works. Imagine throwing all music into a blender, every genre and style.

Then you dump the contents of the blender into a sorting rack, that puts all of the pieces of music back together the way the machine believes it should be organized. When your music hits the blender, it gets filed under similar songs by the sorting process.

Like most AI tools, it seems very impressive at first glance. But if you have ever tried to write with AI, you know it has some serious flaws. Mastering is a human technique as much as writing, music, or other art.

AI will analyze your track, and take very streamlined notes on the EQ, limiting, and stereo field. This is a cool trick, and it is backed by “feeding” the AI tons of music and having it focus on the mastering qualities.

Once it has done that, it knows the areas where your track would “traditionally” need enhancing or adjustment. You can then dial in the amount of these areas that need help in your master.

This creates a “one size fits all” type of mastering process however, based on your music’s style, genre, and overall sound. If your track is raucous, aggressive punk rock then the AI will compare your track to similar artists.

So does it work? Can you use the quick and dirty mastering plugin to publish your song?

Sure, you can. But how much of “you” is still in the song after you have ran it through an AI filter? Again, I do not want to get “deep” with you on this, but machine learning and emotional creativity just shouldn’t coexist.

It may seem like the plugin has done the work for you, and it has! But the result is something generic and thoughtless. The plugin just does what it is told, to every song.

---

AI And Art: Final Thoughts…

I don’t think that the jobs of mastering engineers are in peril. AI mastering is a quick way for amateurs to get their songs on streaming services. Professionals are not using AI mastering for the final product.

It is impossible to say if AI mastering will ever be as nuanced and expressive as a human. But I seriously doubt it will replace a human, and there will always be a market for engineers.

AI mastering works, without a doubt. But it is akin to eating a microwave meal because you are in a hurry instead of cooking a meal from scratch.

Not all of us recording at home have the time or ability to master our tracks. You might just be starting out, and mastering is something you plan to learn later down the road.

I think that is who AI mastering is made for, artists that do not have the time, resources, or knowledge to master their track themselves.

In my analogy above, you do not have a personal chef at home, so you make a microwave meal. There is nothing wrong with that, and sometimes the quick option is the only option..

But as an artist myself, I don’t think AI has a place in the arts in most cases. If it is being used as a tool to simply ignite creativity, then that is great. But AI cannot replace human creativity.

Mastering is a creative process, not a mechanical one. I have an AI plugin telling me right now that this sentence can be better, but I am dismissing it. In fact, I dismiss it frequently because it would ruin creative writing.

Creativity involves making mistakes, and breaking the rules. Trailblazers in music rarely stick to a safe path, honing their eccentricities however possible.

So while I currently cannot say that AI has no place in art, it certainly should not be the sole source. If AI mastering helps you to understand how to learn “actual” mastering, then use it!

But AI mastering is not a replacement for an actual engineer, a producer, or yourself. Your tracks should drip with your personal touch, even at the end of the creative process, when mixing and mastering.