Lil Peep - (64 Min Dataset)

Demo: https://youtu.be/8dkNwRP09hQ Literally just downloaded this as audio and ran it through inference directly

Not as big as my dataset 😜😜

how big 🗿

180 minutes 😜😜

wtffff

thas gonna take a while to train

Which is why I haven't made the RVC model yet 😜😜

well train it to 400 or 300 epochs

since theres 180 minutes of dataset <a:3682skullswishy2:1100290651079319592>

All clean

What? <:droidsobpixel:900162905402654740> <:droidsobpixel:900162905402654740>

Stems yeah

There's a lot of stems lmao

Found the screenshot

Nearly 3 hours

Pretty sure that's all doubles and lots of silences which I intentionally went through to remove. My dataset was originally over 6 hours long.

I trained it on 64 minutes of just varied vocals with no silence at all

There's definitely still silences but not 2 hours worth, and I'd say for doubles it's probably still worth using them as long as yk it's not identical

Since it's still a different like

Way the person would speak

They were identical, most of the pro tools projects had duplicate stems - I assume it is because the engineer used a feature similar to 'make unique' in FL

I made sure to remove these

I suppose I still have half what you said you had, maybe we just used different stuff 🤷🤷

Idk though

It's likely we have the exact same dataset, I dug deep and to my understanding I have every pro tools session that was ever leaked publicly of his.

Yep I assumed so

Zach Bryan - Hymnal (lil peep AI cover) by <@269224879604105227> using this model https://youtu.be/f4W_8PPCcyU

Amazing work 👏

Thank you ❤️

like how no one talking about this bra <:skullsob:1093200088261734420>

dont tell me its that vid bro 😭

Великолепно

wowww thank you