AV-nnightworks

@vadriann.bsky.social

🇨🇺 🇺🇸 🏳️‍⚧️ 🏴‍☠️ he/they, 30s. Sample-based concatenating singing synthesis hobbyist. Language enthusiast. Voice amateur. Burnt-out illustrator. Creator & voice provider of UTAU voicebank [Adrian]. 🚫NoAI🚫 🔞 Preferred if followers were over 18, thanks!

You know what I want to see more of? Honest to goodness covers or originals featuring a duet of the user’s UTAU & a pro voice synth of their choice.. I think that is cute & want to see it happen..

Parallel project I'm currently chipping away at for a friend also happens to be UTAU-relevant, so y'all can take a peek lol. It's not going to be perfect but I've already learned so much from it for the next one!

Eclipsed Sounds & a handful of other Synth V voices are my nuance re: ethical language models… … I know nobody asked & I do have strong opinions about it in general so I will not digress

Absolutely fascinating to see voice over artists passionately advocating against working on projects using genAI, all the while their profile photo is AI generated. I am very aware nuance exists but I still find it so fascinating. Be consistent, or something.

I just realized I could have used the [ehll yuw] sample bc “you” is encountered so often in lyrics I decided to do VCV-style samples for all vowels+single consonant phonemes... clearly already forgot I did that

AV-nnightworks@vadriann.bsky.social · 4w ago

Finally able to put my experiment to the test... seems to be working as intended...! (I still kind of hate the typical CV-VC sound of the singular [V C]'s going into the [CV]'s but girl VCV was going to be impossible for this)

CVVC was my first real trudge through the trenches of UTAU voice bank configuration, but I ended up preferring the sound of VCV simply bc it produces overall a more consistent sound… I’m hoping with the experiment I’m doing w/ Mystic, it will help the sounds concatenate a little better. We’ll see!

Really fascinating side effect of having commissioned the Adrian_Medium phonemizer- seeing which of my vowels Cadlaxa chose as equivalents to their ARPAbet counterparts helped me actually HEAR the difference between two vowels I otherwise couldn’t hear before bc I have a merger with those.

It’s so ironically amusing to me how offended some music artists get when asked if they use genA/I to make their music when they were perfectly fine using it for their album art “Why would you accuse me-“ you should KNOW why, silly…