Raw LLM Responses

What are your thoughts on how Vocaloids have started to use AI recently? To explain how they're used, first they do it in a way that avoids all the theft because for each voicebank, they retrain the model using consensual recordings from their voice providers. (And I imagine that the base model they train to first learn how to copy voices likely uses consensual recordings too tho no info about that as for now) Said voice providers are then compensated for not only their time recording, but most of them get a cut of all revenue made by people buying their voicebank. Meaning they also basically get passive income from people buying their voice. (There are exceptions but for the most part in these exceptions the voice provider is instead paid a massive amount of money upfront to make up for it.) It cannot generate a song or lyrics, and you still need to write a melody for them to sing, and make lyrics. Then you need to "tune" it, which is basically adding all pitch expressions (since singers don't just stay flat on the note 24/7 like a robot), "parameter tuning" which is basically telling the voice to go a bit breathier during this section, be a bit more tense and loud during this one, etc, to mimic how a human singer would sing different parts differently. And of course if it mispronounces certain parts you have to fix the pronunciation. Though, something the AI can do is automatically do the "tuning" for you, just making the pitch expressions based on what it "feels" is natural, in which case the way that the singer expresses themselves is kind of not yours. I personally prefer to do that part all myself though there are a lot of people who don't so in a way that's a bit more like just letting the machine do it for you. I'm curious on what thoughts on all of this are considering it's a bit of a grey area -- it *does* replace the human voice, albeit it's becoming more of its own art form with its own workflow, like people taking portrait photography vs drawing portraits. But although you can of course do the entire process of tuning and stuff yourself, the AI *can* automate a lot of it. And these voices have become harder to distinguish from actual humans for a lot of people.