There exist a fair amount of programs or web services to automatically master music. Is mastering with some automated mastering programs/services as good as mastering by professional humans nowadays? If that depends on the music genre, my main interest is electronic music.
asked Feb 8 at 3:44
-
"Electronic Music" is a rather broad category, about as broad as "guitar music". Also, the market matters: Is the recording headed for a Spotify playlist, a major motion picture, or the dance floor?– TheodoreFeb 14 at 17:13
-
@Theodore Mostly electro house and trance, for clubs. But I'm interested in radio edits (web distribution eg Spotify) and other EDM genres as well.– Franck DernoncourtFeb 14 at 19:26
Add a comment
|
2 Answers
No. There is actually no true AI in the world at this time. Everything that is called “AI” is actually just machine learning. These are systems that are given basic algorithms and then trained on a library of data. The algorithm allows the system to synthesize new results based on partially matching the training data using the specified parameters.
While sometimes machine learning systems can produce very human-seeming results, they are in many ways quite limited. The limits are the result of the training data sets and the parameters extracted from the training data.
Actual intelligence can synthesize new parameters when presented with training data that is different from the other data. Also actual intelligence can dynamically ignore some data in a data set by judging it to be specious or irrelevant.
In short, computers still cannot reproduce the sensitivity and judgment of trained and experienced humans. Those automated mastering services are mainly cash grabs, trying to find unusual ways to monetize machine learning.
-
"Everything that is called “AI” is actually just machine learning." or sometimes, just rule-based :) I'd be curious if anyone A/B tested AI vs. humans for mastering though. Feb 8 at 6:11
-
It's like thinking Alexa is a smart AI, because she can tell you who won the Battle of Hastings. Really, she's just reiterating a human post on the subject. I suppose you still learned who won it, though.– n00dles ♦Feb 8 at 23:53
-
1…or you could ask ChatGPT, which, if you're lucky, might give you the right answer.. or alternatively a completely wrong answer, wrapped in convincing prose ;))– TetsujinFeb 9 at 10:48
-
1If 99.9% of people could not tell the difference between an AI and an Experienced Human product is not the subjective distinction between good and best rather a mute point. One would need to see the comparison of quality as judged by a general audience before asserting "computers still cannot reproduce the sensitivity and judgment of trained and experienced humans" Feb 9 at 19:37
-
3@Blindman67 If 99% of people could not tell the difference, then I would agree that AI/ML can master as well as a human. I don't think we're even close to 99% at this time. Feb 9 at 20:04
You be the judge.
Will AI be as good? Well one should ask ones target audience. However that is not a thing that can be done preemptively. Only you can judge how much an AI can enhance your work.
Mastering is a subjective art where quality is as much kudos as professionalism. Can you pass of an AI's work as professionally mastered (implied)?
Here is "...so you can be the judge of whether AI beats real ears in an actual mastering studio." It compares 4 online (AI) mastering services against an experienced professional human, with examples for each product.
Listen to the samples and "You be the judge".
-
1@FranckDernoncourt The earliest comment on the blogpost is November 14, 2021 so I would guess last quarter of 2021. Feb 9 at 23:28
-
1My favorite one is actually the pre-master. The preferred mastering engineer of the mixer/producer made the best master, but could have done more to fix problems that shouldn’t have been left in the pre-master but were. Feb 10 at 16:16
-
1I actually prefer the MajorDecibel master, but if I was mastering that, it would be different to all of them. I don't like Zac's master, tbth. None are ideal to me, though. But that's the point, this is all very subjective and all that really matters is what the client thinks, and if it fits with the genre and the format. (and also how good the pre-master is, and here, it could be better)– n00dles ♦Feb 12 at 18:00
-
1@n00dles I listened to the pre-master last. I had listened to all the masters three or four times before I started wondering why they all had some similar things I didn’t like and then realized the pre-master was there and checked that out. Of course the pre-master is the source of the common problems but at least it isn’t over processed like all of the masters are. I would have done some light EQ and also automated in some more dynamic range and that’s it. The masters all seem to have added unnecessary processing that just took away and didn’t add. Feb 12 at 21:31
-
1The experiment could be repeated with several human mastering engineers - some people would still prefer one over the other. You could then throw them all [human & AI] in a pot & pick out anonymous examples at random, with no labelling, double-blind. You'd likely end up with results similar to when they did the first mp3/analog blind tests, in which they discovered "most people haven't a clue".– TetsujinFeb 14 at 9:34