Hello, New here, I have a question i like to ask
Ive managed to gather some voice lines from a character in a game I play and I was hoping if I can do an AI cover using that voice
How should I start and what should I use to make it?
#New here, wanting to try creating an AI cover
eager lotus
eager lotus
anyone?
surreal kindleBOT
Ayo? @eager lotus level 1 !!! 
zinc bramble
Hello, New here, I have a question i like to ask
Ive managed to gather some voi...
so, training a model and making a cover?
https://docs.aihub.wtf/ has everything you need to get started! if you have any issues let us know
eager lotus
Okay I'll take a look at this later today
eager lotus
hmm whats wrong with this one?
i followed the installation guide for the voice separation
and this is what it says
zinc bramble
not sure about the voice separation colabs tbh
you could just use https://www.mvsep.com imo
eager lotus
aaaaand im stuck in queue :/ hmm are there any other softwares that works?\
does this usually take this long to do?
surreal kindleBOT
Ayo? @eager lotus level 2 !!!
eager lotus
hmm what can you tell me about this UVR 6? It says I need high end specs for it to work
burnt rampart
hmm what can you tell me about this UVR 6? It says I need high end specs for it ...
cpu mode could work, only slower but not too slow
eager lotus
so like blender?
burnt rampart
if the RAM is limited, I'd suggest processing up to 10 min audio each
eager lotus
okay ive managed to download the files of the song
Now next part is using the voice of the character I want to use
wait for the character's voice, does it have to be 10 mins long? Or individual voice lines?
eager lotus
anyone still on? if not I'll check later as im about to go to sleep
anyway i'll try tomorrow
i hope someone will be there to guide me
eager lotus
hello im back anyone here?
eager lotus
as ive said last night i got most of the recordings i needed
what next?
zinc bramble
https://docs.aihub.wtf/rvc/resources/datasets/ ---> dataset editing part
https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/ ---> train the model via Colab
tawny pineBOT
eager lotus
whats this error? am i missing something?
surreal kindleBOT
Ayo? @eager lotus level 3 !!!
eager lotus
earlier I was somewhere around 200 when training the data set, however it was already night time and I needed to close my PC to save power and do the chores
so I stopped it and closed it
what should I do now?
anyone?
zinc bramble
earlier I was somewhere around 200 when training the data set, however it was al...
ya had to export the model before leaving
if you didn't, ya probably lost the model
eager lotus
damn, so I have to re do everything from scratch?
zinc bramble
whats this error? am i missing something?
and about this, did you run Set Training Variables? weird error
zinc bramble
damn, so I have to re do everything from scratch?
yep... i guess
eager lotus
but it keeps showing that error
zinc bramble
and about this, did you run Set Training Variables? weird error
_ _
eager lotus
alright
eager lotus
Update: so I entered 700 epochs and stopped at 537
I tried to export this and another error showed up
zinc bramble
oh yeah... you had to stop training after like 1 or 2 hours and then export it
it disconnected...and you lost the model again probably
zinc bramble
is leaving the epoch default 500 okay
yeah ig, but just know that you're probably not hitting that in 2 hours
zinc bramble
no
surreal kindleBOT
Ayo? @eager lotus level 4 !!!
zinc bramble
https://tenor.com/view/vegeta-dragon-ball-super-defeat-jiren-gif-18521730
😭
itll work out i promise
eager lotus
I just wanted to hear her voice again, and hear her sing 
well i guess i'll rest now...
I dont know what else to do now with this, im worried that if I try it again tomorrow the same thing happens again and again and again
eager lotus
Alright going to try it again if this doesnt work I give up
stuck again
i already set the training variables
what else did i miss?
hurry before i lose my progress again
eager lotus
update it finally finished
but it says i need to download the file thats closest to before the over training point?
so far i downloaded 4 files
e500 s4500
e80 s720 (2nd lowest the lowest being e60 s540)
e100 s900
I set training as
save_frequency 20
total epoch 500
batch size 8
which one of the 4 i downloaded should i use?
zinc bramble
try the latest one
meager solar
If you want the easy way to make AI Covers I recommended the Astra Labs bot
But when Astra Labs doesn't work I recommend this Colab https://colab.research.google.com/github/SociallyIneptWeeb/AICoverGen/blob/main/AICoverGen_colab.ipynb?authuser=4#scrollTo=odzpJHpr_PaF
eager lotus
If you want the easy way to make AI Covers I recommended the Astra Labs bot
Does it work offline/cloud?
meager solar
No it's just a bot and check the 'Astra Labs' channel
eager lotus
ok
eager lotus
try the latest one
e500 s4500?
meager solar
Here's the tutorial how to use Astra Labs and how to make an AI Covers there, they'll DM'ed you when it's done
You should register the model you desire there if your desire models aren't exists yet there, type the '/register' command if you want to put your desire models there
If you don't need another option other than pitch you should just use the '/cover' or 'cover-file' command because that '/advance -cover' or '/advance-cover-file' are for those who want to pitch the instrument etc. too other than just pitch the models
meager solar
You should join on Astra Labs' server if you want to train/make your own models and you can still do AI Covers there too and you can do AI Covers on 'bot' channel/forum too bc it's similar with 'Astra Labs' forum here in AI Hub but you can still upload the mp3/WAV etc. on cover-file or advance-cover-file there bc you can't add any files on 'Astra Labs' channel/forum from here in AI Hub anymore
