#inanimate utau

[drawingeveryutau]

Michishirube Kiro! Managed by 94km/h!

Released May 22nd, 2024; Kiro is an inanimate UTAU made from various sounds associated with traffic, from car horns to train whistles. They're a robot created to help direct traffic and protect people, something they dedicate much time to. They struggle with differentiating animals from still objects.

[integratedcane]

Doe (誤伝ドゥ) by FafroMOONs

I adore this creature

[cheatexp]

sosik wanted me to post these eventually..

[electric-dot-png]

new chibis for my website (and artfight) !!!

they are also on my utau blog's pinned post ^u^

yami sorry you are Excluded. Because you dont fit

[staticwire]

[hrrashthing]

what yall know ab kurukuru nejiri..

[krazysquid]

so i've been busy with the wx voicebank

so remember when i said this was going to be a C+V english bank with japanese capabilities? i lied. i actually gave up on the japanese and doubled down on the english. more on that under the cut btw. most of the japanese sounds... weird.

it may not look the part, but this bank is made out of duct tape and hope. i don't know what it is, but something about the 'iy' sample makes it die if i shift some of the otos 2 inches to the left

also. never use moresampler with it. never.

nerd rant under cut

now i have some CV samples. you may be saying 'but squid! since your voicebank is synthetic, wouldn't that have no impact on the naturalness of the bank since you're just stringing sounds together in Audacity or something?'

LOUD INCORRECT BUZZER

before i did cv samples with k, it just. wouldn't make noise. the k sample in the voicebank is actually just the 'k f' sample in a fancy trenchcoat because k on its own makes no sound for some godforsaken (probably frq file related) reason

you may notice. that i also have cv samples for d.

you see. utau has this thing where it will extend your samples so that utau voicebanks can sing long notes without cutting off. however. since consonants are in fact samples, utau will stretch them. this is fine for most consonants.

EXCEPT FOR D

and others, but primarily d sounds weird

so to solve both of these problems i did the sane thing and strung together k/d followed by every sound that could possibly follow k/d.

and since i'm using the C+V phonemizer, i have to manually enter in the CV samples. i really should make my own phonemizer but those things scare me lol

[angellutau]