AI music mastering -- now good enough?

There is so much a mastering engineer does that involves “taste”. Even the simplest things like how much to push the limiter and to decide if/how much the true peaks are allowed to go over 0 requires a trained ear and a lot of experience. An AI can’t listen and decide if a certain amount of distortion is acceptable for a certain song. I don’t even think the AI can do the very basics that are just a part of prepping the master like removing any clicks and pops or other artifacts. And what about deciding on track spacing or doing crossfades between songs for a gapless album? And what about the million other things a mastering engineer does lol?

I wish the process of mastering wasn’t such a black box for a lot of people so then they could actually make an informed decision about this kind of thing and they would not be oblivious to what they are missing out on having an AI program do it (including having your audio go through some nice analog gear and all the mix feedback you could be getting from an actual human listening to your track in a properly treated environment). Sure the AI might be useful for giving suggestions or maybe pointing out something you missed while mastering, but it is not going to master a song by itself any better than it could write the song by itself.

1 Like

Art is still very much a human thing. Sure AI can make a cool picture, or put together a copy paste tune, but NUANCE is human. For now at least. It should remain that way IMO, though honestly, I’d rather AIs were attempting art than killing us. :smiley:

2 Likes

I used the free Bandlab mastering tool two weeks ago. I don’t know whether it’s AI based.

I have no knowledge about mastering, but what this free tool did, sounded a ton better than my unmastered version. I didn’t expect that.

Upon more detailed listening, I noted that some melodies sunk from the master. They weren’t as prominent as they should be. I guess that if I’d want them louder, I’d have to change the mix.

I didn’t continue that route at all, but I actually enjoyed being able to listen to my tracks in this mastered way.

1 Like

Would be very interessting for me if AI can soon clean up one muddy recorded audio file from my digitone :wink:

Speaking about SoundCloud, has anyone tried the Dolby mastering option they have been offering since well before the “AI” word was used as a wildcard for everything?

I use CloudBounce to boost up client and my own demos after studio sessions. Regarding clients, they do not pay for mastering services. Therefore, I am not going to put a lot of effort into the task. Usually, I can quickly send them a file link before leaving studio. With that said, it did eliminate the occasional complaint regarding demos sounding much lower than mastered releases, which should be common-sense.

I also tried eMastered and LANDR in the past as well. Neither service was bad per se, but CloudBounce tended to provide the “best” initial results in most situations. YMMV.

Would I use CloudBounce or another AI mastering service for my own finalized projects? Probably, if I released music hyper frequently or did not have the resources or skills for other options. However, an iterative process is necessary for the best results in my opinion:

  1. Create an AI master.
  2. Compare it to reference tracks.
  3. Adjust mix settings or AI mastering parameters as needed.
  4. Repeat until happy with the outcome.

image

1 Like

Dang. I missed that memo, since I use the desktop version and never access the website. I bought the lifetime subscription years ago when it was on sale for $99. :frowning:

It appears CloudBounce was bought by Image-Line and the mastering technology will be integrated into FL Studio: