Are A.I. Generated DX7 Sounds Good? Digital God or RoboBozo?

There's a lot of hype and anxiety over the development of A.I. The A.I. generated DX7 sounds expose the limitations. Much of the hype is fueled by a childish want for new toys without a maturity of developing skills of tools one already has. In the terms of Steve Jobs, it's creation of B and C players.
Off The Matrix DX7 patches can be found at: www.offthematrix.net/product-...
Stock footage came from PexBell, Digital_Expert and spacetrip of pixabay.

[@GEOSynths]

Thanks, I enjoyed this. This sort of Patch Generation has been around for a while in some Synths, it's the Nord Lead A1, even the Arturia Polybrute has some form of Patch Genration in the analog world. How it differs though from "Machine Learning" isn't for me to say or understand, but the results are very random...so far anyway.

[@mastercylinder1939]

A noble ambition, I hope you are successful.

[@DH-fu7bx]

It's a shortcut to succes. It's a real bummer that people are going to lose more braincels in the future now.

[@audioartisan]

That's exactly what's happening. The human race seems to be obsessed with not doing anything, and having others do the work for them. Good luck with that ppl. I grew up in the 70's thank goodness, so I can see the deception and idiocracy pretty clear compared to those born into it.

[@rbus]

Fun bit of trivia: The Kurzweil K250 from 1983 was hyped as using "AI techniques" to generate it's amazing (for it's day) piano sounds. In one sense, it's true but not in any way that involves neural nets. The K250 was a sampler built at a time when memory was horribly, horribly expensive -- and to capture all the nuances & variations of tones of a grand piano in an under 1MB sample set, they employed a BRILLIANT trick. Hire analog synthesizer guru Robert Moog to build a powerful per-voice filter circuit and use software to "train" the modulation of filter parameters to approximate the real tone captured from a real piano with various levels of note velocity, dampening and other expressive characteristics. This technique apparently succeeded. As for neural nets creating amazing DX7 patches? Of course you can. Actually given software & fully MIDI-controlled patch editing, you could very easily build a neural net that can approximate any sound to the best that given synth can provide. Hell, I've created some amazing FM sounds entirely with AI and no synthesizer tools.

[@offthematrix5310]

I thought about the possibility of making a board with new samples for the vintage Kurzweil synths. But it would be challenge if the data used compression. You are very knowledgeable, I wonder if it could be done?

[@rbus]

@@offthematrix5310 There is a lot of exciting things happening right now on the K250 mailing list with a new addon board that adds memory and USB port to transfer samples quickly from PC. Not too familiar with the later Kurzweil synths though the 90s had a few PC soundcards using the Kurzweil Mass soundchip like Turtle Beach & AVM Apex. They supplied an editor that gave you a lot of control over the chip.

[@museonfilm8919]

Very thought provoking , thanks. AI has crept in at an alarming rate, and just like how the internet trashed many small businesses - AI could end up being the death of practical skills, if we are not careful.

[@tenalock]

quality

[@tommy2nes]

so bad