I discovered that I made the transition-y vowel thing too audible in the last test so I fixed the oto. I think I might release the reclist and this voicebank once I finish. I'm halfway done anyways. Hopefully, it'll be sometime in the Winter if I decide to release them.

But yeah, this was a really unexpected outcome for me, considering that I tried this about a year and a half ago and it turned out horrible. But now, it sounds like normal vcv... minus the weird faint sound of transitioning in the middle of the previous note. (But when you do hear the transition here, it's really noticeable and gross haha.)

Song - Soleil by Tiara http://lovetrue-emotion.net/studiotiara/index.html

Ust by me

Semi-vcv test thing 2

I saw this video back in December:
http://www.nicovideo.jp/watch/sm24924472

It literally does everything.
Like, just look at all the applets. They're crazy awesome.

This is really short because I was overwhelmed by how awesome NameWave is. I am not exaggerating one single bit, I swear. GO TRY IT RIGHT NOW I'M NOT EVEN KIDDING iT'S SO COOL. IT EVEN DOES A CLICK TRACK THING OMG.

And lol I obviously like this song a lot b/c I keep trying to cover it haha and never finishing... (sadface). I really want to finish this though, so... man, just wish me luck lol. And persistence.

song: http://www.nicovideo.jp/watch/sm24536934
NameWave: http://zsp.6.ql.bz/namewave/index.html

NAMEWAVE IS OHMYGOD SO NEAT

context: https://cohost.org/ArtificialAngel/post/7738675-i-kept-telling-mysel

made in around 2 hours

Still alive, just busy and moved on to other things in life. I'll be around though :)

1am (driving Home From The Party)

oldish thing, possible next mariette release (2 pit so far but like idk man might make more??????), possible reclist release, and what on earth did i do back when i made this to make it sound good

also the part after this sounds hecking lit but it's unfinished so i didn't include it. like i almost pissed my pants i didn' t expect what i heard. i expected absolute garbag like i usually make lol idk how im even gonna make the rest of it sound that good man

also im tired so bye

another wip so i don't accidentally leave this cursed site for two years again

doing some testing for a cool thing i'm doing. basically it's the same concept as the fake vcv except it's a rentan cv w/ fake vc (essentially for tweaking consonant strength + slightly better blending, all the stuff that makes cvvc fun and good, except w/ like maybe 20 more configs than a cv bank lmao...). it's like, very small. i'll release the reclist + other resources (guidebgm + separate oto or smth, idk) w/ the voicebank ;)

Also fair warning, much like the fake vcv (which i'm still working on after all these years don't worry lmao), the 20-something weird configs there are are hell to set up LMAO... i'm so sorry spongebob....

meteor - john
ust - yaenako

meteor (yet another WIP) (also testing a new recording method i made up lmao)

So the semi-cvvc thing (name pending lmao) was a really weird success and honestly turned out PRETTY viable in terms of usage compared to the other thing i was doing (I'm still working on that I just have new ideas don't worry...). Also gonna be my next vb release, after finalizing the reclist + making some adjustments

sry for the wip spam i'm gonna archive + remove them soon lmaoo. it ain't over yet i'm about to experiment w/ english so double sry. i don't want this to get too long-winded, also im tired and going to sleep, good night :p

the au lait - lemm
ust - kibka

__edit (10/10/2019)
i accidentally recorded a 5-pitch utau OOPS fdjsofdjs

more reclist testing, also i accidentally recorded a 3-pitch utau OOPS fdjsofdjs

maybe i'll actually release her someday.....

original - Shibayan
remix - deadman 死人
instrumental - Fobia
vsq - V1c1o3mHM

----Edit 2019-10-24----
Download the semi-untested voicebank... if you dare......
https://trashbees.blogspot.com/2019/10/super-duper-unofficial-urin-sasaki.html

【佐々木ウリン】Omae Wa Mou / Tiny Little Adiantum【UTAUカバー】

there's like so many weird, obvious inconsistencies in this vb but i've already got 5 pitches, too late to turn back at this point....

also i'm working on a cool tool that does vb distribution stuff. i'll be using that for all my releases and i'm so sorry in advance.

meteor - john
ust - yaenako

carmine low range testing or something

i completely forgot to post this on here i've been making jokes about it on twitter for the past few months but here's some highlights:

* no note blending because i'm a coward
* no pitch bends because i'm a coward
* no WORLD data caching because i'm lazy, so it takes 3 years to render
* timing inaccuracy of potentially 5ms per note, with no per-note correction, because i'm lazy
* it blows up if you so much as look at it the wrong way while using it
* the only redeeming quality is that i'm using WORLD, which is 100% dope and I love it

i'm still working on it i'll open source the thing when i feel like it lmao....

i made a singing synthesizer and it sucks

Technically already released Urin but it be like that sometimes. 

https://trashbees.blogspot.com/p/urin-sasaki.html
https://www.dropbox.com/s/7apbs08qs6ovump/stand_alone.zip?dl=0

Stand Alone - Nakyamurya

Urin Sasaki - Stand Alone - UTAU release + UST

i finally did it lads.

It's around like 750-800 recordings, w/ like about 1-3 oto lines each (I'll get the exact metrics when I finish the vb). It's missing a few things, like the "thr" consonant unfortunately, and a lot of ending only consonants that I wanted to add. It has generally very cv-like usage -- [- cv] for starting and [cv] for mid as per this method's general standard. I feel like I've mentioned all this before tbh.

It's got a large number of issues I gotta work out but I think it definitely sounds like something that has been worth working on for like 5 years haha :D

(...Also edit cause I forgot to mention this) but the image is what the usage kinda looks like. I personally thought it would be worth trying cutting out a lot of diphthongs and adding more consonant clusters bc 1) I don't see that as often and thought it was worth trying out, and 2) I personally think it's easier to do vowel transitions than manual consonant clustering... but I understand exactly why most people would disagree (you don't have nearly as many issues with envelopes), but I'm also writing this reclist for use in my synth that doesn't handle manual clustering and doesn't cache the f0 yet (I am a lazy programmer and most of these problems are because of my slapdash note splicing algorithms and lack of a user interface...) so it has to be both small and handle all consonant combinations. But theoretically I could just. Just add stuff so you could do that if you wanted, just for funsies, once I finish testing the base reclist...

vcv-ish utau english reclist test :p

mostly just posting this bc i was really proud of the improvement in the method I've been working on listen to this smooth motherfucker. 3 years of work really has done wonders. i hope you enjoy my hilarious accent that no, i do not plan on ever attempting to remove from any of carmine's voicebanks, we're stuck with it forever boyz....

also no i don't plan on posting here much outside of stuff like this im so sorry.....

more semi-vcv stuff but this time in japanese

i was like "i want MarietteS5 but not sucks" and recorded this in a fit of post nut clarity. anyways im gonna go eat some steamed veggies with creole seasoning haphazardly sprinkled on and disappear for another 6 months when i decide to release this vb or something idk

OPE WAIT I FORGOT SOMETHING: https://trashbees.blogspot.com/2020/07/shhh-its-secret-heres-carmine-mini-1.html

i annoyed my neighbors to record this voicebank in the span of 10 minutes

i've been working on this for the past week and let me tell you.

I made the very unfortunate mistake of "being a programmer" and saying "I've made a singing synth basic text to speech for a video game couldn't possibly be that hard" and i've never been so horrifyingly wrong in all my life.

First off, Lua doesn't really have any kind of WAVE interface so I had to write my own, which I've decided to attempt to expand upon and I've got it here on github: https://github.com/trashbees/LAud

I stumbled upon a VERY annoying and hilariously dumb performance issue that I had to fix to get anything longer than a couple of seconds to render in under 10 minutes (i made it loop over the audio file sample-by-sample and it took a lil table concatenation to make it work in under a second. whoops)

I had to rework the tts part (currently not open sourced yet but i'll do it eventually) like 10 times just to get pauses, voice configs, and diphone parsing to work, make a script to parse and serialize cmudict as a lua table (for getting phonetic representations), i had to summon a demon to encourage me to record a MONOPHONIC vb, every time i open audacity i writhe in pain, and hyphenated words STILL don't work.

anyways, that's how i've spent my week

i sold my soul to make a simple text to speech synth so i could play the navy seals copypasta

* added pauses (variable length! just add extra punctuation!!!)
* can now type out words @F @OW @N @EH @T @IH @K @AH @L @IY (phonetically :))
* diphones work!!! DIPHONES WORK!!!!!!!!!

I have some stuff I wanna work out, like....
* additional stress/auxiliary stuff
* a "pronunciation guesser" so it doesn't try to skip weird words
* documentation so I don't forget how to use it

would you guys still like me if i became emo?

hi im trashbees and i log on once every 2 years to post wips and disappear bc i get nervous :(

*makes a little gay song that pisses you off*

(always @ follow limit sorrie ily :<)

CAPTAIN MARIE

Popular tracks