Gal received a copy of the newest Vocalo engine 2 days ago and we're already making a collab lmao
- - - - -
GETCHA by: KIRA and Giga
MIDI/mix: yours truly
@Galanist : Ken tuning
@pix-prucer : Chris tuning

V5s are pain

I just recieved her today and am making a cover already gosh I love Lily so much 💛
Finished cover: https://www.youtube.com/watch?v=9VarNV1-s-o
- - - - -
Song: Reversible campaign by DECO*27 & Naoko Itai
MIDI/Tuning/Mix: your truly

I got Lily now

Anyone that rembers him deserves a PixPrucer servant discount

I completely rerecorded Konryuu(紺竜) and gave him a shitton of extra samples (not used here) and omg does he sound good

If you were asking why the previous bank is privated, well, I didn't quite like it, the F#3 pitch was a bit scuffed, it was recorded on an old mic and it also didn't cover the whole potential range I could've recorded him on.

I'll release him as soon as a cover song idea comes to my mind 💔
- - - - - - - - - -
Creditsss
Song: "Logic agent" by niki
UST/Mix: yours truly
Art: @hq_png (twitter)

Resurrecting dragon from the dead

Here are 2 examples of a w.i.p. AI voicebank for PIX trained on 15.3 minutes od data
1st demo: The day I became a monster by Rerulili
Vocal manipulation and mixing: PixPrucer
2nd demo: Pop star by hirai ken
Vocal manipulation and mixing: haru0l

PIX NNSVS short tests

Just received a newly trained ENUNU model still on 25:14 minutes
Didn't even bother to correct the timing LMAO
- - - - -
Song: Torinoko City by 40mP
UST: UtaUtaUtau
Tuning/Mix: Yours truly

More PIX AI testing

Yay a new cover I'm working on rn
I decided to use Cangqiong here since I've been sleeping on her for way too long now
- - - - - - - - - - - -
Credits: 
Song: "僕が夢を捨てて大人になるまで" by Kasamuura Toota
MIDI/Tuning/Mix: yours truly

Cangqiong Japanese real

A w.i.p. of an upcoming "Why does a heart break" cover
GOSH I LOVE HIM SO MUCH HIS VOICE IS JUST SO UGHHHHHH vry good,,,,,,
- - - - -
Credits:
Song: by KIRA
VPR/Mix: Yours truly
Art: https://twitter.com/yen_mi1/status/1269364206563885056/photo/1

Finished cover! -->https://www.youtube.com/watch?v=pibxmfWVcbM

WDAHB w.i.p. // Chris V5

that's what I'm doing in exchange for robux


yeah
- - - - -
Song by Ryuusei-P
UST/Mix: Your truly

Magnet wip

Finally snatched her box from amazon
I've been wanting to buy flower for a long time now and I'm so glad I did, she sounds awesome ❤

For curious I snatched the breath files from her CD and added manually in the mix
Finished cover: https://www.youtube.com/watch?v=1kyukjkcCaw
- - - - - -
Unknown mother goose by wowaka
MIDI/tuning/mix: yours truly

FLOWER IS MINE NOW // Unknown mother goose w.i.p.

alternative title: cishet male sings fairytale

treat that as a shitpost ok thank u
- - - - -
Fairytale by byzzg
UST: Haru0l
Mix: Yours truly

Renzo recorded a fucking vcv

Yeah

It's a big ass w.i.p. and it still misses 4 pitches so

VCV Lite; C3 D#3 F#3 A3 C4 D#4
- - - - -
Song: Nandemonaiya by Radwimps
UST/Mix: yours truly

New PIX VB ???

QingSu my beloved
Also! I used cross language synthesis to make her sing in english here

Credits:
Song: POP/STARS by K/DA
SVP+Mix: PixPrucer
Instrumental: TAEB

Youtube upload: https://www.youtube.com/watch?v=0IkBvEfLF48

She's a goddes with a blade

A wip of a requested cover!
- - - - -
Credits:
Song : Biegnij przed siebie - Anna Wyszkoni
Instrumental: https://www.youtube.com/watch?v=Ec538Lqkkt8
SVP/Mix: your struly

QingSu singing in Polish

I think the title is self-explanatory enough ahahah
I've been attending voice lessons recently and it helped me a lot with my singing and voice overall
This time I aimed for a husky tone, it's my first time recording it and I'm pretty satisfied with the quality of it
Might make it a 4 append VB in the end we'll see
- - - - -
なないろの朝 by Wato & Chouya
SVP by UtaUtaUtau
Art by hq and nekudotexe
Tuning+Mix: yours truly

New CVVC VB because my voice developed

UST by Yhandere
Song by kzlivetune

It's 2:30AM I should be sleeping but NAOKY

Last Night Good Night // Amano Naoky AIβ wip.

RYO IS THE BEST SYNTHV AI NO ARGUING WITH ME
I just used auto-tuning on him he sounds SO GOOD

Youtube upload (tuning shown): https://www.youtube.com/watch?v=4XMOBv_YOQ0
- - - - - - - - - -
Credits:
なんでもないや by Radwimps
SVP/MIX - yours truly

Silence, Vocaloid

I love this song okay
Full cover: https://www.youtube.com/watch?v=0IJkxqhR2GM
- - - - - - - - - -
"Naming you" by thommofonia
SVP/Mix: yours truly

Ryo JP --> EN

I was doing a cover recently with some lower parts and I've ecountered an issue where the voices sound really robotic/raspy in their low range. I decided to investigate and here are raw SynthV renders of Ryo singing a descending scale on 2nd octave. You can clearly hear how more broken he sounds as he goes lower.

Hopefully it's just a matter of updating the vocoder or the voice itself
The other cause could be a lack of 2nd octave singing in the dataset

Male SynthV AI voices die on 2nd octave

Why did I actually put effort in tuning this post help
- - - - - - - - - - -
Credits:

Original:
Singer:  Mo Si Man 莫斯满 Lao Mao 老猫
Composer:  Lao Mao 老猫
Lyrics:  Dong Fang Jun 东方骏

SVP and mixing: yours truly

Cangqiong // Ye Hua Xiang // SynthV カバー

Her Cool timbre is everything I ever wanted, bless M5

Full cover: https://youtu.be/mTTNtwYjono

Very Cool, Stardust

Satellite Mind by Yunosuke (cover w.i.p.)
Also, a new CVVC Powerscale voicebank test

Full cover here: https://www.youtube.com/watch?v=Nr5hC6dNoVo

VB/Tuning/Mix: PixPrucer
UST: ヴィラン
Art: vieri

Satellite Mind // PIX Powerscale CVVC test

Yeah, It's useful, very
You can already download the bank btw

- - - - -
Logic Agent by Niki
UST/Mix: your truly
Art: rev

So PIX Powerscale has an F5 pitch

This is what happens when a tenor takes UTAU seriously

Full cover --> https://www.youtube.com/watch?v=NdzYjjz3wlQ
- - - - -
Song: Nounai Disco by r-906
UST: Pifuyuu
Tuning/Mix: Yours Truly
Art: rev

More high range slay // PIX Powerscale w.i.p.

Kanaria songs are fun to cover

Full upload: https://youtu.be/JbOGIdBg6ag
- - - - -
MIRA by Kanaria
UST/mix: yours truly
Art: hq

Back at my 6 harmonies era

Getting my ass ready for Wei Na, sorry ChiYu but you won't end up in the final cover </3
- - - - -
PoKeR : composition, arrange
St : lyrics
SVP/Mix: yours truly

Tuning practice thing // Chivalrous Beauty w.i.p.

hi piiiiiiiiiiiiiiiix
- - - - -
Song - 傘村トータ
UST - 千秋
Tuning/Mix: yours truly

あなたの夜が明けるまで w.i.p.

Qi Fu is my new fav okay
Also the mix is quick don't mind the loudness
- - - - -
Song: Kemu
MIDI/Tuning/Mix: Yours truly

I LOVE YOU SILLY GREEN DUDE FROM ACE // Haiboku no Shounen w.i.p.

MY QUEEN I'VE DARED TO SLEEP ON YOU FOR SO LONG
Will finish this tomorrow. <-- didn't
- - - - - - - - - -
Credits:
"Phony" by Tsumiki
Lyrics - SINAY
VPR/Mix: yours truly

I FORGOR I OWN MAIKA

Yayyyy it worked just like I wanted it to ~

This voice model is a combination of a CVCC PL bank I recorded back in 2020 (never ended up releasing it) as well as string-based recordings. Originally, the idea behind the model was to make a pre-trained base for other Polish models to train on, but just out of spite I decided to test it on regular songs and it works surprisingly well! There are also 3 other banks of mine I'd like to add to this model, eventually totalling up to about 3 hours of data

I might release this model eventually with actual song data once I get some time to record, yahooo

Oh and also, I just trained the acoustic model and stole timings from Stephan
- - - - - - - - - -
Credits:
'Euphoria' - Loreen
Translyrics/USTX/Tuning/Mix: yours truly

50min UTAU2NNSVS test // PIX PL AI ???

He sounds a tad better with a postfilter :]
Also the acoustics are smarter now, more layers helped bring that detail out hehe

FULL https://youtu.be/Zl9qdy3J4kg
- - - - - - - - - -
Credits
"Donut Hole" by Hachi / Yu Suto / Masaki Hori / Yasuhisa Kataoka
UST - EchoBlossom123
Tuning/Mix - yours truly

Getting A.I.chi to that 1.1.0 update

Used Solaria from my cover as reference audio
- - - - - - - - - -
Song: Someone Like You by Adele

So I trained Stephan on Diff-SVC and... woe

flopper but make him more realistic

full cover: https://youtu.be/V-upYcvxa-M

I definitely need to let the model cook more cus not only I trained him from scratch but also for just 84k steps (took me 6 days). I combined all his Polish UTAU recordings for the pre-trained base (heard in this cover) together with my singing. I guess I'll have to fine-tune him on singing from now on
- - - - -
"Nic dwa razy" by Wisława Szymborska
Music & Melody by sanah
USTX + tuning : yours truly

PIX AI 2 - DiffSinger test

Yuh he can reach those notes fine so far but he defo needs more data nonetheless
I'll try to record this exact song so it can reach those highs just fine, the current data for head voice is scarce unfortunately but NNSVS still handles it fine
- - - - -
UST: Riten
Lyrics & Music: Omoi
Illust: rev
Tuning: yours truly

Messing around with PIX AI 2 (ENUNU)

.      ! THIS MODEL IS STILL A WIP !
The quality will get better in the future 

DiffSinger is truly a wonderful engine
This is his ENUNU Japanese database in action labelled with Polish phonemes that represent the closest sounds to Japanese pronunciation ! Alongside with the help of 15 other singers to make him sing almost flawlessly in the language

Model specs:
trained on refactor-v2 branch
16 singers divided into multispeaker embeddings
10h Polish singing + 1.5h other singing (including karasu) data
80k/200k steps
- - - - -
"Ale Jazz!" - by sanah
USTX - yours truly

I gaslighted Karasu AI to sing in Polish

pfp by @Tweek_Twix on twitter

Welcome to the wip hell

Download the floppa here: https://pixproducer.wixsite.com/voicebanks/pix

Old profile: @pix-prucer

PixPrucer (2)

Popular tracks