cant forget the samples neva neva neva

O.O

im ngl this took 2 FULL DAYS to cook

wrong screenshot (this is the right one)

mommy

did u like marisas magic show

ya

very nice

ight who's next

hmmm

leaning towards reimu

but idk

ya reimu

the universe wants reimu so reimu it is

Kirisame Marisa (Touhou: Azure Reflections) (RVC v2) 320 Epochs

this model gotta be one of the best ive made (tho ive only made 3 in total lmfao)

Bro have to be Model Master Right now

I want tips about how to make those Fire AI Voice for sure if it possible that you can tell me

alot of voiceclips and like um

U um

Clean voiceclips

Ya

Voiceclips then like

More voiceclips

So

Like

You got 30 min of voiceline clip

And you train them?

More voiceclips

So 30 min of voiceline and it like each one 1 min

So would be 30 voiceclip in total

Like that?

Bring out the voiceclips

Yea

Heh voiceclips

A file similar to a video clip that contains a short audio segment that a user records to send to a contact

Heh

Then more voiceclips

Literally LostWord give me 6-10 Voicelines

Oke now I know where to get more

Rip the voiceclips

There will be spell card long ass speaking

Tell em bring out the voiceclips

🔥🔥🔥

Tell ‘‘em to bring out the voiceclips

Where my long ass spell card thingy

It only get 2-3 line for spellcard 😭

When it not in 2x speed it look slow as fk

i was gonna try touhou lost word but their voicelines are too compressed the audio sounds like dookie as a result of being part of a mobile game

bums me out since they got way more characters than azure reflections

tbh i dont blame them

Really? I use lost word and they sound perfectly fine

I could send you a sample if some of my lost word models

If you want

yea sure

Okay!

Here is Reimu

And here is Marisa

nah that compression is coming thru the model

main reason why im avoiding lost word

Huh I see

i talked to codename about this and we both agreed that ogg/vorbis compressed audio files (the compression used for lost word's audio) was very much not ideal for models

Okay. So there's no way around this?

Just saying

ogg/vorbis is a lossy compression method meaning that the audio loses its quality as the method is applied and once its lost you can never perfectly restore it back to its original state

Darn. That's not good

And that really bumps me out because I really like voices from lost word compared to other games

same here

i want a good quality suika model 🥺

There are some type of voice acting from fanime but

I don’t know how to remove the intrusment that good to having even only vocal

With no lose of it

UVR5?

have you try it?

i don't really know Those UVR setting that much

Literally i like this Marisa Vocal from Azure Reflection a bit now because of Lost Word is kinda weird

i don't actually know how to read Epochs and it really make me wondering how am i suppose to know i am overfitting or under

"Lost World is kinda weird" That's debatable

You could try changing to Demucs htdemucs

It works pretty great in my opinion

Oke I will try it

demucs is superior uvr screws up audio quality

TRUE…

Well fuck