i was a vocal synth nerd before i was a fedi/foss nerd. we’ve been doing ai since before the ai bubble, and i think vocal synths are a good example of ethical ai.

vocal synths are still a creative tool where you compose the music, lyrics and expression yourself, but the ai engine makes the voice more realistic sounding. you purchase “voice banks” which are effectively training data for a single voice and this voice bank comes from a “voice provider” who is a paid singer that will record samples for the vocal synth engine. a lot of voice providers request to have the voice bank “characterized” to sound different from themselves, and the vocal synth company will do so. compare KAF to KAFU CEVIO.

this is a process based entirely on consent, something openai and the rest of them lack, they just send out an army of scrapers to take anything and everything they can get their hands on, consent be damned.

actually speaking of KAF, i was excited because KAFU was coming to synth v, since i don’t have CEVIO (and don’t speak japanese). but unfortunately, KAFU SV was cancelled because the synth v ai engine made her sound too much like herself, and most likely they couldn’t modify the voice bank to sound differently enough and they cancelled it. at least, that’s the prevailing theory.