Naysan♤

弘法トロン 声明シンセのウェブアプリです。Web MIDI APIに対応。 グリーンバックにして楽しもう！AU/VSTも鋭意開発中！ hugelton.com/apps/KoboTron/

Finalizing some stuff

Soon I think

This is the newer engine with tension! (The vocoder used is so undertrained- but acoustic still needs a lot of care but hey! progress!) UST: MEKARI

That one low quality oto visual by yours truly

Chopped chud resampler test Ust: りゅー

Neural Singing Voice Synthesis Engine (old) Test :DDD (engine in progress)

@annamaeblythe This only provide range in the vocoder, not acoustic model. Acoustic range will always be limited by the data it has, unless you use tone shift with it which is a big hassle especially if your data doesnt have a lot of pitches

@NaysanMLo7 This is fixed by using this vocoder :) ! github.com/openvpi/vocode… Don't use it on human recorded models with a decent pitch range, though! It will flatten it completely 😅

Idk if ppl know this or not but if youre doing utau to diffsinger please make sure your render or the final audio you are going to use for training, please make sure you cover LOTS of pitches. Its to prevent your model not able to sing in certain range thats not in the data range

It could be why the model turn into whisper after a certain range and "coughin" issue lmao

Fun fact: hifisampler that everyone love is an AI resampler. The data used (for pc-nsf-hifigan) were all consensual so nothing to worry about :D

I'm sorry, but I don't get people who stopped listening to vsynths because the pipeline moves to AI and think it's bad just because the vocals are AI generated 😭

Headlock Nine0 practice mix svp by novawari

Upsampled defoko

Defoko appends leak /j

Why the fock does Twitter keeps suggesting me apologie tweets 😭 i swear every time I got a notification from this account its some random person saying sorry to something

Len kills Miku yay

sillysampler's silly flag

This is what I wish would happen to miku

vb used: yokuatsu takuto

SillySampler wip 7/24/2025 I love (hate) writing a resampler