【JP / ENG DiffSinger Demo】Palette【秋山大智 AI / A.I.CHI】

Hello! I used both English and Japanese in the same track for this cover (super cool that you can do that thanks to his dictionary..!) A lot of progress has been made, I managed to convert A.I.CHI’s ENUNU files to the ENUNUX format, so his Diffsinger version now uses the timing models from his ENUNU ver as well :-) which makes for a much easier time using his model (not as much manual adjusting..)
So far I have trained him myself a little bit more (100k steps or so) HOWEVER I am having trouble getting his model to work with the phonemes (I know how to fix it but it’s a lengthy process of elimination and I don’t feel like doing it orz) so this cover is using his original model with 57k steps in total! He still sounds so so good T_T I am glad he exists..
--
Model Information
No. of Songs: 21
Sample Data: ~66 Minutes
Label Count: 22,133
No. of Steps Trained: 57,000
Model Production: PixPrucer
DiffSinger Port: haru0l
Recorded Range: G2 ~ C#5
Diffsinger: github.com/Moo...
OpenUTAU for Diffsinger: github.com/xun...
--
UST by HaruVampire
Original - Yuyoyuppe www.nicovideo....
Tuning, Mix, Art - peeslubn

[@Cresoul07]

Diffsinger is like synth v but free and with the possibility to create your vb, its too powerful I love it

[@jayzeroey]

ABSOLUTELY GORGEOUS OMG. I am STUNNED,, and the art?? god tier … daichi 4ever

[@peeslubn]

AHHHHHHHHH THANK U

[@xuusynth]

aichi is my favourite littleguy

[@Peachy-Fruit]

he sounds SO good I love him

[@blackstart_v]

HE ALMOST SOUNDS LIKE HDVM SYNTHV /pos THIS IS SO GOOD

[@roganarchi2319]

He's literally so fucking cool

[@SunGuardian524]

prettyyyyyyyyy

[@ryouichi]

if u said he was synthv i would just believe u ..