This is actually very interesting, I need to try it out. Nailing some famous but not obvious signature synth lead sounds, for example. Perhaps, if given audio examples.

Explaining what you want is probably the trickiest part, it’s like that with non-music stuff as well.

 

seems less complicated to just do it yourself , , ,

 

Yeah ChatGPT creates diva h2p-preset code if you ask it, but in my experience the code does not work.

ChatGPT does not provide the ends binary-part of the preset-data and the format of the code is totally different from real diva's h2p-code.

Maybe you can train it to do that better but free version's restrictions make it uncomfortable task ("It seems like I can’t do more advanced data analysis right now. Please try again later.").

 

Once in a interview i heard Jaycen Joshua telling one pretty interesting thing: when he was a starter, he took all the notes from Larrabee Studios made by other producers; Notes that had chains and patches for processing. And he basically discovered that people were using the same thing on everything. It would change the piece of gear, but not it's place on the chain or general schematics. People were abusing from 1073 going on Tube-Tech compressors when recording vocals, and it became a 'standard' formula. The only difference in between producers is that some would use a Manley Voxbox instead of the 1073 or a Summit compressor instead of Tube-Tech, for example. But the basic chains were barely the same.

So using AI to approach those standard and extensively used studio presets can be a hand on the wheel. Also to ask for variations. At least one thing that AI does a great job is documenting and organizing those kind of info without the need to steal chains from a random producer at a MWTM masterclass where no single concept is explained behind the processing. Also, no ads to bias your decision. It's a win win for me.