Forums

I have a particular use case for AI that I'm interested in getting others' opinions on. I'm working on revising and remastering a piece for resubmission. I perhaps too ambitiously decided to feature a choir singing Irish. In my original mix I tried very hard to finagle EW Hollywood Choirs' Wordbuilder system syllable by syllable based on online audio recordings of each individual word, but after hours of work I'm not very happy with the result. It's not a huge deal, since epic choirs are not usually expected to be understandable, but I worked hard on the lyrics and translation and I'd love to hear them more genuinely represented in the final version. Lately I've been eyeing AI voice transformers. It appears possible to record myself singing Irish, transform my voice into a handful of AI-generated voices, and layer those with Hollywood Choir's samples to create something that sounds like a choir actually singing in Irish. In this case nothing would be "generated by feeding a prompt into AI-software." The AI would instead be fed a recording of my voice and would duplicate the notes in a new, generated voice. My creative activity would not be too different from what I currently do with samples of choirs and instruments: every note and expression would be my creative choice, but the actual sound would not be created by me.  I'm curious what others think about this. AI as a whole generally feels icky to me, not the least because AI companies are notoriously unscrupulous about using others' work for training their models, without compensation. On the other hand, some companies at least claim to be ethical in this regard; kits.ai for instance says that their models are trained using audio recorded by contracted and compensated session vocalists. This seems no different from how EastWest obtained the audio which they sampled for their choirs. My other option would be to record my wife and try to duplicate her voice with enough differentiation using a variety of filters and effects. My wife is doubtful about this working well, however! What do you all think? Acceptable use case? Not acceptable?

HEAVY! But tastefully so! This does an excellent job of punctuating the heavy guitars -- which sound great, but would get tiring -- with lighter sections. Dark and creepy throughout, exactly what the original was begging for. Well done!

eh, providing that AI will get good fast at writing at least very serviceable remixes...isn't it all based on good faith at this point, anyway? the essay or info bit would be more of an outlet for the artist to be honest and simultaneously highlight why they think this creative M.O. was interesting in their mind. i think if you allow the use of AI, it's good to at the very least encourage honest disclosure about the particular use.   even if someone *hacks* the judges panel with an almost purely A.I. created remix (and maybe an equally A.I. created faux essay detailing the use of A.I.), they could still disclose it later on in the remix comments... "hhahaha gotcha!" then it's funny atleast.   i think we oughta be focused on people who care about creativity period...setting up some punkbuster-like anti-cheat system seems futile. (of course, you could employ an A.I. to punkbust mixes that are likely created by A.I., but that would seem a very elon musk solution. can't beat em, join em :D)

That background video is cool as fuck