loxstep wrote:Wonder how that would work. If it was anything like other AI projects I've encountered, an AI for audio would be trained on a corpus of existing songs. You could put in thousands of songs and it would look for patterns in how songs tend to be mixed / mastered. In the best case scenario, an AI could compare your track with similar commercial examples, and point out differences in the dynamic range, frequencies, etc. I bet a lot of people would find that useful. But I would be hesitant to use such a tool. I don't want my mixes to sound like everyone else's.
I think mixing is too subjective to be automated. There is no right level something should be at. It depends on what you want your audience to pay attention to.
Yep I fully agree but remember I said for us bedroom engineers? I was thinking the AI would on initial set up be able to analyze your mic, then you play a song you think is perfectly mixed at the max volume you can in your room.
Then the AI will analyze the results of the environment so it will let you know during mixing if something is too hot or needs boosting etc.
That avoids the frequent 'car test'.
. I'm on the 14th floor and my car is 2 levels underground. Just when I'd think I have the perfect mix, I head down to the car just to find out the bottom end is very weak in the mix.
With this AI, it will have all that taken care of with suggestions.
Once final mix is ready, the AI will analyze the final mix for mastering frequencies and suggest or automatically master the song like LANDR.
Love your idea of the AI analyzing many different genre of music for learning.
Wish I could code or program or whatever these geniuses do that create plugins, etc.