The title says it I guess there seem to be a lot of Gimmicks floating around to do with AI in music production but I think they must be at least a handful useful applications!
For splitting stems, I’m using htdemucs. I’m not sure if that is old already?
Track-splitting is good. I tried some AI reverbs which just seemed to tax cpu but no better than stock reverb or Ableton hybrid reverb.
Udio is freaking people out. It makes full songs that actually sound decent. A bit controversial tho apparently it is trained on copyrighted data…
Which reverb plugin was that?
Will have to check out udio but idk feel like even when the full song AI becomes quite good it’ll never over take people. There will always be something missing without the human element? maybe I will be proven wrong, that would be a shame.
Reverb plugin was the reverb from the sonible AI plugins!
Reverb plugins are cool but not that special in the bigger picture! AI can get way more creative, we’re talking AI adaptive plugins like; EQ’s, Multiband compressors, Modulators etc.
A good example I found last year was Gulfoss by Soundtheory. It’s a really cool intelligent EQ.
1 Like
The realtime analysis of Gullfoss uses Soundtheory’s computational auditory perception model to understand which audible elements are competing for your attention.
Found on the plugins homepage, would be interested to know exactly what they are doing in the background but I guess probably would not get a chance to see the algorithm even if you asked 
The songs by udio are wild… lots of them about sh*tting pants too:
@harvoo I tried gulfoss (dont tell anyone I did not pay) and I liked it but found it almost didnt have enough knobs to tweak? maybe need to give it more of a chance…
Yeah I know, if you go through some of the example ones they are quality! It worries me, makes me think we’re gonna have quite a bit of competition going forward. If be with you thinking that humans still have the edge at the moment. I said that from the start when these tools were first appearing but I’m shocked at how fast this has come along already - the most surprising example I’ve seen is udio, there could be more out there 
My favourite so far 
Warning - it’s rude.
I saw a video where James Hype (the DJ) is making a track, he takes the vocal from an old track, then gets AI to change it so its not a cover and uses that 
For the stem separation I found that these is a lot out there to choose from and its really difficult to find anywhere giving an unbiased opinion… so far htdemucs has been good for me. I also tried gaudiolab.io and found that quite good (and free)
also on the topic of track separation, I’m messing with this one called MVSEP more info here. It’s free again you run it in google colab…
Things could start getting interesting with copyright if authorities change laws to recognise “ideas” beyond after they get messed with by AI. (maybe we get to the point the copyright still counts if you mutated the original by using AI)??
Thanks for MVSEP will check that!!
Went into a bit of a rabbithole last night and found “Ultimate Vocal Remover” on github. Looks very promising contender for a “best” one… ifs free too 
Yeah I think that algo might even be in UVR - a lot of these sharing the same free algorithms or combinations under the hood!
1 Like
Talking of so copyright - did you see that so diss track drake made… featuring all sorts of people he did via ai…
Very well for drake to make one but he is the first to lawyer up someone creates an AI version of him🤭
1 Like
Chat GPT for answers to mixing questions. Example: “explain how the attack on my compressor will effect the sound of my sub bass, and what the commonly used range is” that sort of stuff I’ve found is gold!!
Other uses of AI I’m not so sure. Not totally convinced by the udio stuff either…