Help me again 😭😭😭
#✨│ai-help
proven hill
I can’t find the public link thing
manic nebula
If o do this, what should happen?
i*
And I do this after installing it?
Or before?
manic nebula
What should appear after I do this?
rare gobletBOT
Ayo? @manic nebula level 2 !!! 
manic nebula
https://media.discordapp.net/attachments/1159290139609137264/1219666894672367676...
I did it and still nothing
proven hill
you dont have a pc to test?
manic nebula
Nope 😕
Do you know a ai thing that works with iPhones? The RVC guide collab worked well but now it’s not working anymore
proven hill
not free and good ones sadly
proven hill
i dont know, its probably some issue with colab
queen phoenix
where is the colab for rvc?
proven hill
But thank you either way!!!
no problem, maybe try with a different browser
brave garnetBOT
⠀
Google Colabs 
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
queen phoenix
thnkx!!
proven hill
or even better, use this one: https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing
queen phoenix
ight i will
rare gobletBOT
Ayo? @queen phoenix level 1 !!!
queen phoenix
or even better, use this one: https://colab.research.google.com/drive/16LkwvFZeu...
how do i put the models??\
queen phoenix
thx!!
queen phoenix
https://www.youtube.com/watch?v=wh4cLMLLkj0&t=15s
help am i tripping it said this then i refreshed the choose model part and its blank
wait imma try other models
proven hill
help am i tripping it said this then i refreshed the choose model part and its b...
where did you download the model from
queen phoenix
where did you download the model from
from weights, not sure if i did the upload to google drive correct.
what i did was dropped my files to upload to drive
then i copied the google drive link for it
proven hill
is the drive link public?
iron cobalt
Hello. I have used RVC WebUI locally early today and I had no problem with making AI covers. But now I get this traceback like this when trying to make a new AI cover:
I once put the path to the audio file in the wrong field, instead I put this audio file path to the index path field
Traceback (most recent call last):
File "D:\RVC1006Nvidia\runtime\lib\site-packages\gradio\routes.py", line 321, in run_predict
output = await app.blocks.process_api(
File "D:\RVC1006Nvidia\runtime\lib\site-packages\gradio\blocks.py", line 1007, in process_api
data = self.postprocess_data(fn_index, result["prediction"], state)
File "D:\RVC1006Nvidia\runtime\lib\site-packages\gradio\blocks.py", line 953, in postprocess_data
prediction_value = block.postprocess(prediction_value)
File "D:\RVC1006Nvidia\runtime\lib\site-packages\gradio\components.py", line 2076, in postprocess
processing_utils.audio_to_file(sample_rate, data, file.name)
File "D:\RVC1006Nvidia\runtime\lib\site-packages\gradio\processing_utils.py", line 206, in audio_to_file
data = convert_to_16_bit_wav(data)
File "D:\RVC1006Nvidia\runtime\lib\site-packages\gradio\processing_utils.py", line 219, in convert_to_16_bit_wav
if data.dtype in [np.float64, np.float32, np.float16]:
AttributeError: 'NoneType' object has no attribute 'dtype'
rare gobletBOT
Ayo? @grim bay level 8 !!!
past pagoda
How to change text into armaan malik's ai voice
low shard
How to change text into armaan malik's ai voice
for text to speech, tts, check #1212368971127590922 , the section of TTS Mixed With rvc so you can use rvc voices
upbeat kernel
INFO:hank1:[14, 0.0001]
INFO:hank1:loss_disc=3.948, loss_gen=3.619, loss_fm=11.484,loss_mel=22.124, loss_kl=6.070
Process Process-1:
Traceback (most recent call last):
File "multiprocessing\process.py", line 315, in _bootstrap
File "multiprocessing\process.py", line 108, in run
File "C:\Users\vlc2\Downloads\Mangio-RVC-v23.7.0_INFER_TRAIN\Mangio-RVC-v23.7.0\train_nsf_sim_cache_sid_load_pretrain.py", line 225, in run
train_and_evaluate(
File "C:\Users\vlc2\Downloads\Mangio-RVC-v23.7.0_INFER_TRAIN\Mangio-RVC-v23.7.0\train_nsf_sim_cache_sid_load_pretrain.py", line 579, in train_and_evaluate
cursor.execute("SELECT stop FROM stop_train LIMIT 1")
sqlite3.OperationalError: no such table: stop_train```help please
mangio rvc
spiral chasm
@sand iris Hey, I just wanted to let you know that I asked for access to hardware with a decent GPU and everything went smoothly - the model sounds great. Thank you very much, you were right.
hybrid spire
I'm expecting to be offline in a couple minutes, but I forgot that I've got a model in training on RVC Disconnected at the moment
I expect it'll disconnect by the time I'm back, but what do I do then? I'd like to avoid restarting the training, since it's a hefty dataset
grim bay
What's the link for the tuto of making models?
Also how much time is enough for making AI model?
low shard
What's the link for the tuto of making models?
Also how much time is enough for ...
check our docs for a deeper explaination https://docs.aihub.wtf
tawdry shell
anyone got any good vocals splitting apps to recommend?
grim bay
check our docs for a deeper explaination https://docs.aihub.wtf
When I choose "How to Make Voice Models", it shows this
low shard
When I choose "How to Make Voice Models", it shows this
retry clicking this https://docs.aihub.wtf/essentials/how-to-make-voice-models/
azure marshBOT
RVC Guides (How to Make AI Cover)
muted fiber
Hey guys how do I get a RVC again? I got a new Computer
velvet crown
how do i get the index¿. Failed to train index: need at least one array to concatenate
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.
unique blade
wtf
brittle wing
I'm having trouble with the Ilaria-RVC interface. I'm at Download Voice Models tab, I put the URL of the huggingface model in the Huggingface or Google Drive field. I copied the name of the pth file in the Name of the Model without spaces field. I click download and it says ERROR - The download failed. Check if the link is valid.
I've checked, the link is valid.
rare gobletBOT
Ayo? @brittle wing level 1 !!!
unique blade
how to upload the dataset to trmsorboard
tall torrent
guys im getting the error "Traceback (most recent call last):
File "/content/Ilaria-RVC/app.py", line 1, in <module>
import subprocess, torch, os, traceback, sys, warnings, shutil, numpy as np
ModuleNotFoundError: No module named 'torch' " how would i fix this
v new to this haha
upbeat kernel
```INFO:hank1:Train Epoch: 1 [88%]
INFO:hank1:[14, 0.0001]
INFO:hank1:loss_disc=...
???
rare gobletBOT
Ayo? @upbeat kernel level 1 !!!
thin stump
@hidden oracle soo what do you need? RVC is free
iron cobalt
-local
azure marshBOT
brittle wing
Is 1 minute per epoch on colab normal?
rare gobletBOT
Ayo? @brittle wing level 1 !!!
brittle wing
Also, is 500 epochs a reasonable amount?
spare shale
Also, is 500 epochs a reasonable amount?
yep
spare shale
Is 1 minute per epoch on colab normal?
I think no?
brittle wing
Should it be faster?
spare shale
Should it be faster?
Well it depends on your dataset
brittle wing
It's like 38 mins
spare shale
1 file?
spare shale
What do you use?
And Where do you use it?
brittle wing
wdym?
muted fiber
Hey guys I have applio on OneDrive and its the same as my desktop normally, but it's still giving me an error
spare shale
What do you use?
And Where do you use it?
easyGui?, RVC D...
colab or local?
spare shale
I'm using the RVCdisconnected notebook
sowwy idk
brittle wing
-colab
brittle wing
Already looking at tensorboard rn
muted fiber
Hey guys I have applio on OneDrive and its the same as my desktop normally, but ...
Does anyone know
rare gobletBOT
Ayo? @brittle wing level 45 !!!
rare gobletBOT
Ayo? @brittle wing level 2 !!!
brittle wing
bruh, embed failure
muted fiber
Like where I tried to move it to my regular desktop but it says its the same place
should I try moving it to documents or something? lol
lavish wolf
Did someone know how to create a voice model please?
brittle wing
Is 1 minute per epoch on colab normal?
how bout this
lavish wolf
Did someone know how to create a voice model please?
Is there a channel with a tutorial in the server?
brittle wing
and how should I accomodate a 39-minute single-file dataset
muted fiber
ok lol
brittle wing
what are the best options for the mdx colab for making good quality lead only acapellas?
i've just been running it thru kara2, kim vocal 2, and reverb hq in that order but not touching any of the other options
thin stump
Is there a channel with a tutorial in the server?
Read here and download this if you haven't already https://docs.aihub.wtf/essentials/how-to-make-voice-models/ https://www.audacityteam.org/download/
unique blade
how to upload data set to tensor board
thin stump
how to upload data set to tensor board
thats covered in the guide 
unique blade
i cant faind it on it
thin stump
TB or tensorboard
fluid matrix
wdym upload dataset to tensorboard
that doesnt make sense
what are you trying to do?
grim bay
I need to know is it still works to making AI model if I use the vocal that have some part silence?
unique blade
TB or tensorboard
tensorboard
unique blade
tensorboard is for supervising training
what
thin stump
no, i mean you can search for both of them? Are you trying to make a model
unique blade
yes
unique blade
i read it all then the githup
timid creek
i feel completely stupid for asking this as i feel like i should already know it but what program do i use for....making music? using a voice bank i got from here (im using shiho hinomori's one for reference) 😭
i think this goes here (idk)
candid spade
How do I train AI
bleak glen
Hello everyone guys, I have a problem: when I turn on “Beatrice” everything is fine (res 200+), but when I select any RVC then res. drops to 10 and nothing is heard as I say (
What should I do?!?((
coral sandal
Is AMD unsupported for RVC?
sand iris
Is AMD unsupported for RVC?
On windows it doesn't work
brittle wing
If I train 150 epochs, save the model, then continue training the model with another 150 epochs, is that the same as training for 300 epochs from scratch?
fluid matrix
If I train 150 epochs, save the model, then continue training the model with ano...
you are just continuing the training process ig
just make sure you use the exact same settings
and dont preprocess anything again
brittle wing
you are just continuing the training process ig
Will the models be the same? 150 + 150 = the same as 300 in one go?
brittle wing
alright
wintry grotto
does anyone know how to fix this?
timid creek
i feel completely stupid for asking this as i feel like i should already know it...
scratch that, add bronya rand to the mix for reference 😭
rare gobletBOT
Ayo? @timid creek level 1 !!!
quiet charm
i've been looking at all of the different uis and they all seem to separate the instrumental, but I'm trying to just submit raw vocals into it. is there any way to turn that off?
rare gobletBOT
Ayo? @quiet charm level 1 !!!
quiet charm
and then whenever I finally get the final file, the pronounciation is all messed up because it seems to apply a bunch of cleaning fx that aren't necessary for the files and it just messes up the vocals
if not, are there any versions of the uis that are only for vocals? I used to know one but i don't think it works anymore
hot lagoon
Help
plush compass
RVC v2 Disconnected Colab Am I the only one that can't do it right now? The first Install Dependencies doesn't go on for 8 minutes. I've tried turning Colab off and on again, but it's the 5th time and it's the same.
plush compass
RVC v2 Disconnected Colab Am I the only one that can't do it right now? The firs...
Changing the ID solved it. I'm curious as to why. Does it only happen with certain IDs?
half cove
Changing the ID solved it. I'm curious as to why. Does it only happen with certa...
what id did you change 
lime thunder
anyone know how to use rvc v1 models 😦
feral saffron
-Colab
brittle wing
Do I need the G and D files to resume training (in Colab)? Can I just use the normal path file?
brittle wing
hello helpers
patent moth
This isn't a colab or an RVC or anything, this is just an FAQ page and when I click on the "how to" stuff it eventually just ends up leading me to a glossary page of what words mean, it doesn't actually send to anything.
Sorry for pinging and responding to this but I searched this channel for people with the "JSONDecodeError" and found this.
I think I'm probably doing something wrong but how am I supposed to go anywhere from here?
bitter sable
why when i try to download zip i have this issue
ModuleNotFoundError Traceback (most recent call last)
<ipython-input-2-67c0c587a0e8> in <cell line: 4>()
2 #@markdown Link the URL path to the model (Mega, Drive, etc.) and start the code
3
----> 4 from mega import Mega
5 import os
6 import shutil
ModuleNotFoundError: No module named 'mega'
NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.
To view examples of installing some common dependencies, click the
"Open Examples" button below.
cloud spade
-rvc d
teal elm
Why does the second step in the Collab notebook doesn't work?
Do I need to download the files onto my PC or is it just a problem with the code itself?
craggy dagger
Hey guys I'm trying to train a voice model on the Mangio fork and it just queues forever at the process data step. Any ideas?
If this helps.
Oh is Mangio dead? Someone was saying it might be on the github.
little lantern
Why does the start.http bat file close as soon as I open it
I have python, grep, cuda, and pytorch installed?
Do i need to restart pc?
pastel oak
Do I need the G and D files to resume training (in Colab)? Can I just use the no...
on local you always needed the g and d files so i assume you do on colab too
pastel oak
Why does the start.http bat file close as soon as I open it
you talking about wokada?
#🔍│help-w-okada message do this
latent maple
for some reason whenever i talk with the client its always ecohing and i have no idea why, and the voice always sounds so distorted ive followed many steps from many videos any idea of what the problem could be
i even copied the same modules like chunk, extra, gain, tune
but nothings worked
every rvc sound sounds ass
pastel oak
for some reason whenever i talk with the client its always ecohing and i have no...
tick the echo box and sup2
little lantern
you talking about wokada?
https://discord.com/channels/1159260121998827560/1159...
Tysm I will try this now. I got the download off some youtube video posted last august.
pastel oak
Tysm I will try this now. I got the download off some youtube video posted last ...
https://rentry.co/VoiceChangerGuide
everything you need is here also the latest version. all video tutorials have wrong info
little lantern
Thank you!
autumn gust
hey can someone help me make an ai cover, I have not done this in like half a year and I forgot everything (my method does not work anymore too)
storm bobcat
does anyone know why it has a lot of static when its in a discord vc?
grim bay
Should I isolate the vocals first and then cut the character's voices, or vice versa?
thin stump
Should I isolate the vocals first and then cut the character's voices, or vice v...
It's vice versa. But if there's tons of music and sfx then just use MDX 23 or BS Roformer
Bandit plus gets rids of sfx but it can muffle the vocals since you're doing an anime model
grim bay
Is it still works to making AI model if I use the vocal that have some part silence?
brittle wing
quick question, does ogg files work as a dataset?
azure marshBOT
RVC Guides (How to Make AI Cover)
thin stump
Is it still works to making AI model if I use the vocal that have some part sile...
wdym, you could dm me a pic of your dataset
brittle wing
Yes https://discord.com/channels/1159260121998827560/1159289354439626772/1186550...
thx bru
grim bay
wdym, you could dm me a pic of your dataset
Well, I mean when I'm looking for my dataset
There are some silence on the anime
thin stump
There are some silence on the anime
denoising should take care of that usually. You could still dm me and i'll give you a workflow
low shard
This isn't a colab or an RVC or anything, this is just an FAQ page and when I cl...
Just check the rvc folder and the cloud folder
It tells u the colabs
You needa read into it
The jsoj error is bc u use an old one
round plank
anyone know why im just gettin this screen
oh no image perms bruh
i opened go-web.bat and its just doing nothing
anyone know the issue
brittle wing
i opened go-web.bat and its just doing nothing
It will show you a port
Go to your browser and navigate to the address localhost:[the port here, without brackets]
round plank
alr
rare gobletBOT
Ayo? @round plank level 1 !!!
brittle wing
on local you always needed the g and d files so i assume you do on colab too
ugh shit, I didn't save them
just the main one
I assume the D and G files are updated everytime I reach the epoch save interval
@round plank how you doing
brittle wing
just the main one
luckily I "only" did 150 epochs
rare gobletBOT
Ayo? @brittle wing level 3 !!!
round plank
<@1148858750543855648> how you doing
reinstalling the folder ngl i got really annoyed before you replied and deleted it
round plank
Go to your browser and navigate to the address localhost:[the port here, without...
doesnt do anything
pastel oak
doesnt do anything
you have perms to post pics now if you need
round plank
round plank
yea it doesnt connect
round plank
when I run goweb.bat as admin I get this tho
i've been looking for hours i genueinly dont know how to fix it
brittle wing
do you have python installed on your machine
brittle wing
so that's strange how it can't find the directory
round plank
well im pretty sure my python isnt installed in system32
brittle wing
okay, maybe try modifying where it's looking in the bat file, or (re)install python to system32
round plank
what about the runtime thing though
thats not even in my actual python installtion location
and idk how to change where its looking ngl
rare gobletBOT
Ayo? @round plank level 2 !!!
brittle wing
what about the runtime thing though
idk
round plank
ahhhhh so confusing
bleak glen
Hello everyone guys, I have a problem: when I turn on “Beatrice” everything is fine (res 200+), but when I select any RVC then res. drops to 10 and vol: always 0.0000 nothing is heard as I say (
What should I do?!?((
pastel oak
Hello everyone guys, I have a problem: when I turn on “Beatrice” everything is f...
does the cmd give any error message?
sand iris
don't run as admin
is it doing anything whatsoever when running as a normal user ? Check task manager
weird that it would freeze, but it's probably trying to do something
round plank
don't run as admin
what do i check
sand iris
any idea what's wrong?
Just guessing, could be a bazillion things :
- Antivirus interfering ?
- Program still being read from disk ?
- Corrupted files ?
I really don't know, I haven't seen this often
rare gobletBOT
Ayo? @sand iris level 82 !!!
round plank
Just guessing, could be a bazillion things :
- Antivirus interfering ?
- Program...
Antivirus is off, reinstalled like 30 times, wym by read from disk though
stiff goblet
what is wrong with this graph
azure marshBOT
RVC Guides (How to Make AI Cover)
ebon valve
guys what do I need to do if I already pasted my voice model into the models folder but it doesn't show up in the program?
I don't have image perms so I can't send screenshots ://
pastel oak
what do you think?
is nice
pastel oak
guys what do I need to do if I already pasted my voice model into the models fol...
you put it in the weights folder?
brittle wing
-colab
ebon valve
you put it in the weights folder?
there's no weights folder, I followed this tutorial btw in case you need to know https://www.youtube.com/watch?v=0nTfLpqfN1Q
pastel oak
there's no weights folder, I followed this tutorial btw in case you need to know...
i would just use the tutorials used in the pinned messages rather than some random youtube one
proven hill
there's no weights folder, I followed this tutorial btw in case you need to know...
use this https://youtu.be/wh4cLMLLkj0
eager pelican
you are in the system32 folder
NEVER run as admin
run normally
and thats a runtime\python.exe
aka the folder runtime in RVC which has the python.exe file
so it MUST be in the same folder as the goweb.bat file
well i just realized i gave gahdur's answer but more complicated and more un necessary
bitter crest
hey can RVC pass through vanguard ?
rare gobletBOT
Ayo? @bitter crest level 1 !!!
bitter crest
hey can RVC pass through vanguard ?
I mean, isn't vanguard a problem
torn bluff
Thanks to someone taking their time for helping me in the dms last night from this server, I have finally successfully created my first AI vocal model. It's mostly trained with the intention I wanted, however it needs more training. I was told to use tensorboard which I plan on doing for the next session (once my colabs GPU has been restored later tonight). I believe I remember reading once the graph starts going back up, you will start over training. How do you stop your AI from training further and save what you have? Also I trained with 200 Epochs. How many more Epochs do you think I may need to finish this model off?
pastel oak
Thanks to someone taking their time for helping me in the dms last night from th...
Very blunt and simple put:
You will just have to keep training until the graph stays flat or starts rising up. But of course, it might just rise up briefly and start falling down, so its best advised to keep training
It is best to save every 10 epochs (the first slider on step 3).
Its a feel-type thing but if youve gone an additional 100 epochs and the graph hasnt lowered itself, you will know the OT has been reached those 100 epochs before. So then you read how many steps the OT point has (the lowest point in graph), if its lets say 3.2k steps you go into your weights folder and select any model closest but LOWER to 3.2k
I am saying 100 epochs just as a reference, maybe 50 are enough or maybe 200 are needed. I am working with tensorboard for the first time this very moment so cant give you a rough number to target yet
Epochs depend on your dataset, how clean it is, how long etc. so theres not a definite number I can give you how many more epochs you need to train. Could be only 100 more potentially
torn bluff
Okay, that all makes sense. Another question I would have for you is, how do you train the model you had already trained before hand? instead of the WAV file i used the first time, do i now use the pth and index of the one that was trained previously?
pastel oak
You do all the first steps like you did the first time. You process the dataset WAV file you most likely are referring to, then you feature extraction etc.
Then on the last step (picture below, excuse me for phone picture) you put down your latest G and D files of your trained model (I hope you have them saved) and then click on train model
So your D and G path might be called something like
D_5120.pth
G_5120.pth
torn bluff
Yes they are saved. Okay, I get it now. I use the same WAV file i did the first time but to further it ill use the latest G and D files. i guess it's the same pth file for both
haughty wasp
guys how make ai model?
pastel oak
Yes they are saved. Okay, I get it now. I use the same WAV file i did the first ...
Just to make sure regarding the last sentence: The D and G files are separate, but theyll have the same number (steps)
torn bluff
Just to make sure regarding the last sentence: The D and G files are separate, b...
yes but you paste the same pth in both?
pastel oak
No, the D pth is different to the G pth
at least thats how it works locally, not sure if the colab is any different
torn bluff
Okay, Good it looks like other pth files were also saved to Google drive
i guess ill have to figure out which other one I need to use
pastel oak
guys how make ai model?
haughty wasp
https://docs.aihub.wtf/
thanks
torn bluff
I know something called "200e" was the one pth i used
i just don't know what the second one will be
pastel oak
I know something called "200e" was the one pth i used
Uhh, if you mean a file thats called
yourmodel_e300_s34800
(the numbers are examples)
then thats neither the D or G, thats your finalised pth file that youd use for rvc to create AI covers or whatever with
the D and G files really are called simply D with some numbers .pth
They should be saved in logs
torn bluff
oh those letters are in the files?
pastel oak
Should be yes.
sudden tartan
Any other way to make voice models without rvc?
pastel oak
logs - yourmodelname - d and g files, index etc. everything to be found
sudden tartan
Any other way to make voice models without rvc?
thin stump
Any other way to make voice models without rvc?
RVC is from hours of coding + other misc ways to use it. Use kits ai or voicify if you're not okay in the head
sudden tartan
This is confusing idk
thin stump
It's just a link. Ignore that
sudden tartan
Ok
pastel oak
Ok
the step by step tutorial by ai hub is very simple to use
in pinned messages the latest one
sacred finch
Question, what's the best yet affordable microphone that can be used for this?
rare gobletBOT
Ayo? @sacred finch level 1 !!!
pastel oak
Question, what's the best yet affordable microphone that can be used for this?
are you talking about the voice changer?
Any stand microphone should do the trick ngl ☠️
Just google best budget microphones if you dont want to spend much
sacred finch
are you talking about the voice changer?
Any stand microphone should do the tri...
Fair! Ok! Thanks.
upbeat talon
Hello, gentlemen and ladies
I have a problem with RVC v2 In google collab
It simply does not start
that is, when I pressed the run me fiirst button, nothing happened
That is, the download went but no ticks or crosses appeared
latent kettle
Hello, gentlemen and ladies
I have a problem with RVC v2 In google collab
It sim...
I think it is due to mounting drive.
rugged stag
I am trying to create content with AI voices and I am using "Easy GUI (for RVC v2, with crepe) (with improved downloader)" in google colab to do so, I get an error in STEP 2 "DOWNLOAD THE MODEL", I put the link from Drive, and an error appears in the code.
I have installed everything as the first step says and I also tried with other voices but the problem persists, any ideas?
torn bluff
It's usually 12 hours when you get your colabs GPU back right? Mine has been 16
pastel oak
It's usually 12 hours when you get your colabs GPU back right? Mine has been 16
I read somewhere its 14-24 in this channel
torn bluff
ill just check back every hour, thanks
pastel oak
I am trying to create content with AI voices and I am using "Easy GUI (for RVC v...
Tbh just use the assets from this server
Use Ilarias RVC instead of the outdated stuff online in yt vids
check the latest pinned message
pastel oak
ill just check back every hour, thanks
I checked again
"12 to 24, mostly 24"
Hope you get lucky
brittle wing
Hello
I have a long question:
So basically an internet friend requested a voice model and I trained it with basically 2-3 accounts but even if I switched the accounts to finish faster when I switched to the 3rd account the training relapsed from 1 epochs and didn't go past 400 even through I did everything correctly, exported the model, resumed successfully.Afterwards I waited for a week to pass and then started from 1 to 400 again with the same model name, I tried to go past 400 witb a third account but it didn't work it kept relapsing from 1 and now I am starting from scratch
So my real question is : is there a limit of the accounts I use for training was it because of that because afterwards I tried to re-use the account with the main folder yet it relapsed from 1 again.
It takes so much time it's very mentally draining
Or it was cause I used the same model name and it automatically memorized itself as 400 epochs without deleting the folder with the model name & training while all the files were kept
Like events, .pth, index I didn't delete them I just retrained the model from 1 yet it didn't work
Maybe I didn't resume correctly
-guides
rare thorn
grim bay
Does someone have some tips on preparing the dataset of a character in the Anime?
brittle wing
Can someone answer and tell me what's the problem
brittle wing
Does someone have some tips on preparing the dataset of a character in the Anime...
For talking or singing
grim bay
For talking or singing
To make an AI model for singing
brittle wing
Well you extract the voice remove noise and reverb
grim bay
Can you dm me the link?
brittle wing
CAN SOMEONE ANSWER PLEASE
grim bay
Btw, I'm having a hard time looking at some Anime scenes
grim bay
Alright
azure marshBOT
RVC Guides (How to Make AI Cover)
brittle wing
[what~4](https://cdn.discordapp.com/emojis/1206290075101560852.webp?size=48&qual...
Huh
half cove
he just gives it for free?
pastel oak
Hello
I have a long question:
So basically an internet friend requested a voice ...
i dont have the answer to your question, but are you using tensorboard which helps detect when your model is ready and doesnt need training? id be surprised if you actually needed over 400 epochs
half cove
😭
brittle wing
he just gives it for free?
I shouldn't have told that.
half cove
i dont have the answer to your question, but are you using tensorboard which hel...
eh
some of my models overtrain way higher
thats just not true
half cove
I shouldn't have told that.
😭
brittle wing
i dont have the answer to your question, but are you using tensorboard which hel...
Huh wait did the tensorboard just detect that it can't go over 400 epochs so that's why it restarted from 1???
pastel oak
anyway ignore what i said i dont know
brittle wing
Is it a problem if I use more than two accounts
pastel oak
maybe colab limitation 🤷♂️
half cove
wait whats the issue
brittle wing
Was that the issue
half cove
cba reading back
half cove
wait whats the issue
@brittle wing
brittle wing
And by far seems strange if I plan on making a 900 epochs model in future how will I be able too
brittle wing
<@456226577798135808>
IDK honestly
rare gobletBOT
Ayo? @brittle wing level 4 !!!
pastel oak
Hello
I have a long question:
So basically an internet friend requested a voice ...
this @half cove
half cove
Hello
I have a long question:
So basically an internet friend requested a voice ...
prob did something wrong
half cove
there isnt a "colab restriction" for retraining
brittle wing
there isnt a "colab restriction" for retraining
I'm glad
half cove
What exactly.
could be anything tbh
brittle wing
That's confusing
half cove
yep
brittle wing
Is it the number of accounts used for training?
Or that I used numbers in the model name
half cove
Is it the number of accounts used for training?
i mean the files shouldnt change
half cove
Or that I used numbers in the model name
numbers in model name is fine
just no spaces or non ascii letters
brittle wing
numbers in model name is fine
So it isn't that
Well I'm starting with a brand new account.
IDK
brittle wing
just no spaces or non ascii letters
Ascci?
brittle wing
I didn't use any characters the name I set was RYUSERALOVER500e
half cove
should be fine
brittle wing
I'm still wondering if it was to avoid overtraining
half cove
no
brittle wing
And 400 epochs was the Target
How do I resume properly also where's the guide
It got deleted...
half cove
How do I resume properly also where's the guide
https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/ bottom of this
just save preprocessed file and model then import it in
brittle wing
Wait ...
half cove
with using another account just download folder and put in rvc disconnected in the other account
brittle wing
AH THANK YOU
brittle wing
with using another account just download folder and put in rvc disconnected in t...
Nahh it doesn't work like that I just share them as editor and it works
Also they take time to download and upload
half cove
Nahh it doesn't work like that I just share them as editor and it works
sharing as editor works?
if i remember correctly you needed to transfer ownershit
brittle wing
I have tried already this method
half cove
i might be wrong here
brittle wing
sharing as editor works?
if i remember correctly you needed to transfer ownershi...
No I just share them to the new email add to my drive and continue training.
brittle wing
But when I switched to third account the problems began
half cove
But when I switched to third account the problems began
you did the same thing as before?
exact same thing
half cove
and it had no errors?
brittle wing
No but it didn't relapse from 400 it restarted from 1 epoch as if I didn't train anything
And I got nervous cause I just waste my time.
I had no issues before
sacred finch
Question, is there anyway to stop the voice from reverbing to a whisper? Like I say a phrase, and it repeats it two more extra times, devolvling into a whisper.
I know this is probably due to it trying to "render" what I am saying
round plank
you are in the system32 folder
NEVER run as admin
run normally
and thats a runti...
Yeah its in the same folder it just stays on that one single line and does nothing
eager pelican
Yeah its in the same folder it just stays on that one single line and does nothi...
no no like
when you open CMD in admin its always at
C:\Windows\System32> as it said in the window
you need to either directly link that python.exe or simply open as non admin (as you should)
never open as admin
specially python stuff
eager pelican
It just does nothing
does it say that the python file wasnt found?
or that it just doesnt show anything
round plank
it just shows the command that runs it and then nothing else
brittle wing
it sounds so robotic...
eager pelican
it just shows the command that runs it and then nothing else
doesnt it say smth like localhost:8080 or 0.0.0.0:8080?
brittle wing
how does training the voice work?
rare gobletBOT
Ayo? @brittle wing level 1 !!!
brittle wing
does it gradually get better
eager pelican
I get the port number at the end yea
then ye you need to open that website thing
round plank
It doesnt connect
eager pelican
It doesnt connect
can you show me a screenshot rq?
like when you run what it shows
round plank
yeah
vagrant cypress
how do I make a voice model from my own voice
round plank
pastel oak
how do I make a voice model from my own voice
vagrant cypress
yes ik but Idk how to make the pth file
eager pelican
how long does it stay like that
round plank
forever
pastel oak
yes ik but Idk how to make the pth file
eager pelican
forever
like it usually should take at worst 3 minuets to open up unless you got a slow storage device
but ye it should be like this
2024-03-20 22:01:47 | INFO | configs.config | Found GPU NVIDIA GeForce RTX 3080 Ti
is_half:True, device:cuda:0
2024-03-20 22:01:47 | INFO | __main__ | Use Language: en_US
Running on local URL: http://0.0.0.0:7897```ive had it up for like 30 minutes
rare gobletBOT
Ayo? @round plank level 3 !!!
eager pelican
ive had it up for like 30 minutes
ok ye thats weird
did you download the prezip which is like 8GB? or did you download the repo which is like 11MB or smth
vagrant cypress
I did this a long time ago I'm on the webui for like 10 minutes trying to figure out what do I do
the tutorial wants a pth file
round plank
8gb the complete pack thing
pastel oak
I did this a long time ago I'm on the webui for like 10 minutes trying to figure...
Local or colab?
vagrant cypress
local
pastel oak
do you use mainline?
vagrant cypress
wdym
eager pelican
8gb the complete pack thing
whats your SSD or HDD?
cause even HDD should be fast enough NOT 30 minutes
vagrant cypress
main
rare gobletBOT
Ayo? @vagrant cypress level 1 !!!
pastel oak
main
https://docs.aihub.wtf/rvc/local/mainline/#training-
Follow all the steps starting from "Training"
round plank
whats your SSD or HDD?
980 pro and a 870 evo as a backup
pastel oak
thats how you train the voice to become a pth file
eager pelican
980 pro and a 870 evo as a backup
ok ye thats super weird
idfk how that would even happen
which RVC zip did you install? maybe its an old or smth?
round plank
ok ye thats *super* weird
idfk how that would even happen
which RVC zip did you ...
october 6th one newest i saw
vagrant cypress
thats how you train the voice to become a pth file
thx
eager pelican
october 6th one newest i saw
how much RAM you got? and whats your GPU and CPU cause maybe CPU is the problem
maybe tho cause its rare for those to be causing a delay of 30 minutes
round plank
3090 5950x 64gb of ram
eager pelican
thats even more weird now
i got no idea how that would happen
you got good hardware and nothing seems to be wrong
but why would it take 30 minutes..
eager pelican
@half cove since you're the only person ik 
help me out
why would someone with a good system
and everything else like RVC prezip properly downloaded
why would someone like him, get a 30 minute delay on opening python cause like he says after 30 minutes and still nothing shows up
well how about you try and move the files to somewhere else
like dont have it in that many folders
cause thats like 5 folders just for the files
maybe the path are too long for python to handle
pastel oak
Question, is there anyway to stop the voice from reverbing to a whisper? Like I...
for wokada realtime voice changer questions ask in #🔍│help-w-okada for the future
but for yours tick sup2 and echo it should fix it
half cove
@round plank u downloaded prezip right?
what did u use to unzip?
round plank
7zip
eager pelican
<@1148858750543855648> u downloaded prezip right?
what did u use to unzip?
ye the 8GB one
sacred finch
for wokada realtime voice changer questions ask in <#1159290161683767298> for th...
Ok! Thanks man!
eager pelican
he downloaded the correct one
half cove
weird 
eager pelican
ye duh
half cove
where is it saved
eager pelican
in his desktop
wet cave
Setting up rvc (or anything like that) with an AMD gpu is a pain
round plank
desktop
round plank
i tried taking out of one still didnt work
eager pelican
cause python hates paths that are longer than 260 characters iirc
eager pelican
i tried taking out of one still didnt work
try putting it in your a folder in desktop itself
instead of under so many folders
or might as well try putting in the C drive itself
like not desktop on C: itself
thin stump
does it gradually get better
RVC will pick up on
- background / residual noise
- mouth clicking etc.
It's going to sound robotic if you don't clean it properly, or overtrain/undertrain
So just read this https://docs.aihub.wtf/essentials/how-to-make-voice-models/
but um theree's no way to fix robotic breathing. That's how it is
round plank
try putting it in your a folder in desktop itself
instead of under so many fold...
already tried all of that
warped folio
website thing not working for me
eager pelican
already tried all of that
ok then im out of ideas
thin stump
website thing not working for me
click the links inside the website since it's still using the old url
warped folio
click the links inside the website since it's still using the old url
which link?
thin stump
which link?
if it says 404, then the directory to the newer website isn't working (the aihub.wtf doc)
unless you're talking about colab
warped folio
no its when i put in the link for the ai voice it doesnt work
round plank
ok then im out of ideas
yeah all good bro thanks for trying
eager pelican
yeah all good bro thanks for trying
ye sorry i couldnt figure it out
python can be goofy at times
but never seen it THIS goofy
thin stump
no its when i put in the link for the ai voice it doesnt work
use #1192011222023950368 so you could post an image? You have to say what colab or hugginface link so others can help you
warped folio
idk what colab is i just followed a yt video
rare gobletBOT
Ayo? @warped folio level 1 !!!
half cove
ye sorry i couldnt figure it out
python can be goofy at times
but never seen it ...
its not showing any errors and i dont see anything wrong 
python moment
round plank
its so weird cause I use the rvc real time voice changer and stablediffusion but this randomly doesnt work
eager pelican
its not showing any errors and i dont see anything wrong <:distress:115957741324...
maybe launch python with debug stuff? i think they DEFIANTLY HAVE THAT RIGHT? lemme check
oook i have a crazy idea lemme test it
godammit
-d : turn on parser debugging output (for experts only, only works on
debug builds)
debug builds only
welp i got no other ways
gray rover
Şebnem ferah sesi yok mu?
pastel oak
Şebnem ferah sesi yok mu?
pastel oak
@vagrant cypress you solved the issue? i was looking up the issue in previous chats
brisk yew
I cannot understand how to do anything brah
Idk how to get a voice model thingy majigy working
I installed Applio but it just
doesn't work
I run the run-applio.bat and nothing
brittle wing
do I only need to train the index once? (Unlike the model which is a long process?) And does it matter if I do it before or after training the model?
fickle elk
I probably have a very specific problem since it is a particular voice I want to use.
From the tutorial I was following there is supposed to be a link popping up after Step 3 but it doesn't do it.
And something seems already amiss at Step two since it throws errors.
I just want to make an AI Cover with Lucifer Morningstar's voice.
I can send screenshots if that helps examining the issue.
pastel oak
do I only need to train the index once? (Unlike the model which is a long proces...
doesnt matter when and only once yea
pastel oak
I probably have a very specific problem since it is a particular voice I want to...
the only relevant tutorial is this one, anything else has failed for me too ngl
im not sure what kind of link is supposed to pop up after step 3?
brittle wing
can someone help? my games and programs pick up my microphone instead of the voice changer
fickle elk
im not sure what kind of link is supposed to pop up after step 3?
Probably a outdated method/tutorial. Video is from last year so it probably has changed in the meantime.
brittle wing
Does anyone have an indication of how long Updating and installing pip packages... (in colab) is supposed to take? I refuse to believe it is supposed to take 10 minutes
(now more than 10)
does anyone know how to make voice models without pitch guidance/pitch correction?
fluid matrix
can someone help? my games and programs pick up my microphone instead of the voi...
did you change the input setting now?
fluid matrix
does anyone know how to make voice models without pitch guidance/pitch correctio...
what do you mean, you want to train a model without it?
brittle wing
did you change the input setting now?
fixed thankyou works fine now
fluid matrix
just dont select it while training ig
brittle wing
what do you mean, you want to train a model without it?
context: i have seen models without pitch guidance on weights.gg and i thought to myself that i wanna try making one like that.
fluid matrix
just select it or deselect it when you are training a model
but why would you want that
brittle wing
y not
fluid matrix
you should enable it for models that are supposed to sing
and speaking models benefit from it too
brittle wing
well for my hunter1s1k screaming model...
fluid matrix
its generally better to just have it on
idk, i guess you can try that
just deselect
but thats probably not the issue
brittle wing
Does anyone have an indication of how long `Updating and installing pip packages...
.
brittle wing
but thats probably not the issue
fair. but what would be the issue?
fluid matrix
fair. but what would be the issue?
your dataset or selected sample rate for example
brittle wing
your dataset or selected sample rate for example
ah
brittle wing
what colab are you using?
RVC Disconnected
brittle wing
and are you sure you are still connected?
yes
nvm
I restarted it and it did it in 5m32s
Looks like it is RNG driven
wispy kettle
I don't have any checkpoint files in the weights folder, anyone know why?
pastel oak
I don't have any checkpoint files in the weights folder, anyone know why?
have to turn this on when training
wispy kettle
have to turn this on when training
Awesome, cheers!
rare gobletBOT
Ayo? @wispy kettle level 1 !!!
pastel oak
if its too late and you need a specific step as pth file, you can go into your logs folder, delete every D and G path up until the one you need (or move them into a different folder, because it always checks for the most recent no matter what you do in the step after), then do the train thing and select the D and G paths, hit training quickly, let the weight get created and close again
so like here i only have two, you delete the 24960 ones and only keep 17920 then rvc actually uses the 17920 and creates the weight
wispy kettle
if its too late and you need a specific step as pth file, you can go into your l...
You're a mind reader aren't you? I was gonna ask about this but couldn't figure out how to put it into words ❤️
plucky flare
i wanna w-okada client in google collab
but didnt find one
the Hina Modified Realtime Voice Changer is so laggy and the original one is disconnecting
rare gobletBOT
Ayo? @plucky flare level 1 !!!
rare gobletBOT
Ayo? @brittle wing level 4 !!!
fluid matrix
what's the difference between added and trained index files
you just need the added index file, dont bother with it
wild adder
What setting can i change that allows the model to follow my volume/pitch changes more accurately rather than being monotone
brittle wing
What setting can i change that allows the model to follow my volume/pitch change...
I think temperature
wild adder
I think temperature
I dont see that?
fluid matrix
What setting can i change that allows the model to follow my volume/pitch change...
are you using tts or what?
wild adder
No its me
wild adder
Is it possible its capping my volume
vagrant cypress
<@1145737587248595014> you solved the issue? i was looking up the issue in previ...
yes I solved it
fluid matrix
Is it possible its capping my volume
wdym, are you using a voice changer?
i thought you wanted to do a cover
with colab
wild adder
wdym, are you using a voice changer?
No I mean in the original recording there are some parts where I am louder, but in Ai its like the volume/pitch is the same the whole time
Holy the search feauture helped so much thanks lol
pastel oak
yes I solved it
what was the fix?
wild adder
In the tutorial it said put it to 0 lmao
wild adder
lmao
rare gobletBOT
Ayo? @wild adder level 2 !!!
fluid matrix
dont put it too high either tho, it will introduce artifacting if you do
that settings basically controlling how strong the "accent" of the model is
vagrant cypress
what was the fix?
when I chose the voice this thing would generate without a value
but it was a waste of time cuz my model sounds like robot
I was so desperate I did a 1000 epochs lol
wild adder
that settings basically controlling how strong the "accent" of the model is
thanks
fluid matrix
I was so desperate I did a 1000 epochs lol
dont train a model to 1000 epochs just because you can, that might hurt the model more than it helps
always check with tensorboard for overtraining
and maybe you didnt select your index file when converting lol
vagrant cypress
ik but I had nothing to do so I just did it
rare gobletBOT
Ayo? @vagrant cypress level 2 !!!
vagrant cypress
and maybe you didnt select your index file when converting lol
I did
fluid matrix
well either way, training it that much wouldnt fix the issue
vagrant cypress
btw what can I do to make the voice sound normal
fluid matrix
your dataset is probably at fault
vagrant cypress
yea probably cuz it's only 6 minutes of my friend talking on discord and he is pretty loud so it sounds awful
pastel oak
you can send me the dataset i can check if i can help tomorrow
vagrant cypress
you mean like the audio files?
pastel oak
ye
vagrant cypress
dm's?
pastel oak
yas
wispy kettle
Btw what exactly determines a good quality model besides not overtraining and using HQ dataset
fluid matrix
thats pretty much it tbh
if your datasets bad
nothing can fix it
thats the most important thing
pastel oak
Btw what exactly determines a good quality model besides not overtraining and us...
for the person to have clear pronounciation, vowels, range etc if you mean that
fluid matrix
i would count that as having a good dataset
wispy kettle
Yeah so a flawless dataset basically
fluid matrix
and maybe your training settings, but not much you can do wrong there tbh
wispy kettle
Has anyone ever tried using sound "upscalers" on shit quality dataset and seen an improvement on the final model?
wispy kettle
What's wrong with PM
fluid matrix
that wont work on rvc anyways
you can use that on your dataset and it might help you getting better extracts but while you are training it will downgrade it anyways
brittle wing
How do I fix extremely slow download speed on Colab
fluid matrix
upsampling like that wont really improve much, its not meant for that
wispy kettle
you can use that on your dataset and it might help you getting better extracts b...
Right, then there really isn't any point waste time on it
fluid matrix
How do I fix extremely slow download speed on Colab
not much you can do tbh, just wait
brittle wing
How do I fix extremely slow download speed on Colab
I want to download my D and G files every 10 epochs (from colab). So I have it saved and can continue training when colab disconnects. But that's obviously not possible
fluid matrix
Right, then there really isn't any point waste time on it
you could use it to upsample really bad audio first so you can use that to extract
like remove reverb for example
but upsampling a finished dataset wont help at all
wispy kettle
Right that's what I meant, upsampling the audio files before turning it into a complete dataset
rare gobletBOT
Ayo? @wispy kettle level 2 !!!
brittle wing
I apologize for perhaps a stupid question, but what does the "Target sample rate:" parameter in Mangio RVC Web mean?
there are options to choose 40K 48K 32K
what is better to choose?
fluid matrix
most of the time you will be fine with 32k. if you want to make sure, check for what you should select by analysing your dataset with a spectrogram
for example here it peaks at 20k, so you would select 40k
that image isnt ideal tho
hot lagoon
is there pertrains for 32k?
fluid matrix
you mean for ov2?
brittle wing
I want to download my D and G files every 10 epochs (from colab). So I have it s...
how can I do this then?
fluid matrix
if you select 32k, it will select it automatically
fluid matrix
how can I do this then?
why would you want to do that
oh
this colab can save stuff too
brittle wing
why would you want to do that
because I want to have the most recent save of a model when colab inevitably disconnects
fluid matrix
this might be better for you
half cove
https://colab.research.google.com/github/blaise-tk/rvc_cli/blob/master/RVC_CLI.i...
wait why am i credited for this
😭
fluid matrix
also your stuff gets saved on google drive too
downloading from there should be easier
fluid matrix
wait why am i credited for this
probably because of the pretrain fix idfk 💀
hot lagoon
you mean for ov2?
yeah and that other one fogate the name tho
fluid matrix
yeah and that other one fogate the name tho
yes there are
brittle wing
https://colab.research.google.com/github/blaise-tk/rvc_cli/blob/master/RVC_CLI.i...
can we resume training in this colab as well?
fluid matrix
yeah and that other one fogate the name tho
there is one for rin too if you need that
hot lagoon
Ov2 is for small datset?
fluid matrix
can we resume training in this colab as well?
yes
fluid matrix
Ov2 is for small datset?
generally yes
hot lagoon
What is rin use for the most generally?
half cove
https://huggingface.co/poiqazwsx/Ov2Super32kfix/tree/main
you can use normal if ur using applio btw
fluid matrix
there is a fix for preventing disconnection on the other colab too
rare gobletBOT
Ayo? @fluid matrix level 80 !!!
fluid matrix
you can use normal if ur using applio btw
yeah but he asked for the specific pretrains
brittle wing
yo guys
fluid matrix
What is rin use for the most generally?
anime stuff i think, never used it
fluid matrix
its based on english tho
fluid matrix
why does the voicechanger eat the end of the words
your mic not picking up the words most likely
threshold issue
or your settigns are not optimal, most likely since you just joined
hot lagoon
is there a way to train my own pertrains?
fluid matrix
did you follow this guide or a outdated video?
fluid matrix
is there a way to train my own pertrains?
there is, but dont bother with that
fluid matrix
you really have to know what you are doing
hot lagoon
I can try to.
fluid matrix
you have to train it with a loooooooooooooooooooot of data and have good enough specs for that
not to mention making sure that everything is perfectly clean
fluid matrix
there is a checkbox for autobackups
brittle wing
I'm not talm bout autobackups
fluid matrix
the thing you sent is just for selecting custom pretrains like it says
brittle wing
yes
fluid matrix
thats why you select that box
brittle wing
yes
brittle wing
from drive?
fluid matrix
yeah
brittle wing
oh
I'm not used to this luxury
I thought I needed to specify the path myself
mb
fluid matrix
you are using the same google drive that you mount so no need
ripe flicker
Sorry, how many epochs do I have to put in to have a perfect model? With a colab?
brittle wing
Sorry, how many epochs do I have to put in to have a perfect model? With a colab...
depends on a lot of things, there is no perfect amount of epochs
best thing you can do is look at tensorboard
ripe flicker
I would like to understand how much should I put in general, and how long should the dataset be?
brittle wing
dataset 10-50 min
brittle wing
I would like to understand how much should I put in general, and how long should...
start off with 100-500 epochs
ripe flicker
Great, thanks
Does a dataset made of sing vocal audio help in driving?
Training *
outer raptor
guys, help
my pc doesnt seem to extract thezip file of power (chainsaw man) pt br
it says error 0x80004005 unespecified
tribal oar
this happens when starting to train the model in mangio
it stops and gives me this error
hot lagoon
just to make sure but BVE 2 in uvr is not the best?
thin stump
just to make sure but BVE 2 in uvr is not the best?
The version 2 is called BVE V2 on x-minus btw
uvr hasn't been updated in a while
hot lagoon
But would we rate it on a 1 to 10?
flat mirage
hey i was reading the "How to Make AI Covers using Ilaria RVC" and it says " 12 for male to female and -12 for female to male".. what does that mean? 😭
brittle wing
your mic not picking up the words most likely
Do you use okada
fluid matrix
Do you use okada
didnt you read this already? https://rentry.co/VoiceChangerGuide
should resolve your issue
proven hill
hey i was reading the "How to Make AI Covers using Ilaria RVC" and it says " 12 ...
Examples:
Original singer/voice is male --> if the model is male the pitch will be 0, otherwise if its female its 12
other way around for female
ashen lynx
So Kanye posted the vocals for his new song, Carnival, but it still has the reverb and whatnot. Could that be a problem with the AI covers and is there a way to remove the effects? Also if this doesn't belong here, I'm sorry. I posted the vocals to show what effects are on the song.
brittle wing
It's broken
fluid matrix
It's broken
pretty sure you are using a old colab
use the ones listed here instead
that ones for training, use ilaria rvc for inference for example
-colab
hot lagoon
how to make a diffsinger voice?
golden creek
rvc disconnected colab is getting slow, used to be 2 mins on first cell
now it's taking forever
been 10 mins and nothing happening, my gpu is wasted lmao
@rigid mango sorry for the ping, any update tho?
hollow nymph
for some reason i keep getting an error when i paste a link for models in rvc and i have no idea why
plucky flare
im using Hina Modified Realtime Voice Changer on Colab so why the res so high ??
tough vigil
hi is theres any other models website but not weights.gg
dusky mist
Guys this is my first time ever doing this I need help, long story short I want to make Travis Scott ai songs on fl studio and am having trouble on to even figure out how to use any of this please and thank you 🙏
regal timber
does anyone know why when i try to download the model in colab it does not work. Sorry this is my first time using AI so im just trying to figure things out.
ocean mulch
Hi, im using RVC Disconnect and i got a runtime error. How do i export my training data out to my google drive?
grim bay
Does someone have some tips on preparing the dataset of a character in the Anime?
rotund prawn
A LOT
hollow nymph
for some reason i keep getting an error when i paste a link for models in rvc an...
nevermind figured it out
pastel oak
Guys this is my first time ever doing this I need help, long story short I want ...
check latest pinned message and on that page go to how to make ai covers
pastel oak
Does someone have some tips on preparing the dataset of a character in the Anime...
Idk what tips youre looking for, the process is always the same except the anime studio will have the vocals more cleaned so maybe even easier
If its a side character or short anime you might not find enough dataset
rotund prawn
Like what?
Making dataset
rotund prawn
Idk what tips youre looking for, the process is always the same except the anime...
Not easier
rotund prawn
Make sure
The tone of voice doesn't change
Same voice style
grim bay
Also, in the anime, there are some parts they keep silence for a second then continue to say. Should I cut those silence parts?
rotund prawn
Yes yes
grim bay
I really don't know how to do on this Anime
grim bay
Thanks
rotund prawn
I did
orchid palm
Hey everyone, just a quick question. I am trying to convert a monologue into phone guy from fnaf. I am pretty sure that i set everything up correctly ( i have the .pth file in weight folder, and the .index and npy files in the logs folder). For some reason when i attempt to convert the voice it just gives me an audio file with no sound in it. What could I be doing wrong/how to fix?
hot lagoon
How do I make back vocals with diffsinger
wild adder
Is there a good J Cole model???
real beacon
Where is the Ilaria RVC new collab?
azure marshBOT
RVC Guides (How to Make AI Cover)
winged jolt
On macbook, I installed the v.1.5.3.18a after unzipping and stuff what should I click to open it? when i click start_http it says no app found to open it
sly wind
How to update it... Do I need to reinstall hole stuff again
glad zealot
How to update it... Do I need to reinstall hole stuff again
Yes, unless it's applio
sand iris
On macbook, I installed the v.1.5.3.18a after unzipping and stuff what should I ...
.18a on mac ?
Don't think they released it for mac chief
Check the file you downloaded, if there's no mention of mac, it's a windows build
sly wind
Yes, unless it's applio
😩😩
hot lagoon
Does anyone know how to use voice sympathizer for background vocals
rare rapids
why when i change to uploaded voice model, the voice is gone?
novel kiln
guys how do u change the profile picture of the voice model?
pastel oak
guys how do u change the profile picture of the voice model?
on "Edit" where you upload models, instead of upload you click on the "blank" on the left side where the picture would be and upload
novel kiln
thanks bro
rotund prawn
Which is important
Epoch or stepcount?
pastel oak
My OT is at 18.4k, the closest before is 17.9k and after 18.5k
Guide says use closest before, would there be any benefit/does it matter using the after model with this small gap to the OT?
I saw someone in this chat once say use the one after so me wondering if that was specific to that case or universal
half cove
My OT is at 18.4k, the closest before is 17.9k and after 18.5k
Guide says use c...
just test both
half cove
Which is important
Epoch or stepcount?
what matters is your model not over training and the quality of the dataset
rotund prawn
I asked for the training to be at its final point
Thanks
My dataset is clear
half cove
I asked for the training to be at its final point
Thanks
My dataset is clear
i mean tensorboard uses steps
so more important to look at that ig
pastel oak
ig = I guess
haughty ice
I want to training my voice for rvc, but I'm not sure. How many minutes of data should I use, specifically how many minutes of data do singing and talking data typically use each?
pastel oak
I want to training my voice for rvc, but I'm not sure. How many minutes of data ...
check the latest pinned message and go to dataset, everything you need is explained. generally 10-20 is fine
brittle wing
Is this supposed to be going for 9 minutes
I refuse to believe so
(This is RVC-CLI)
Nvm, it moved on to some other things so it is working
RVC Disconnected did it in 5m30s. Kind of an L for this notebook
trim star
I hate to bother people but I've been researching for some files all day and night.
I git cloned some forked repos, found some useful various ressources (Custom Pretrained G .pth and D.pth) f0_RIN E3, Snowie.. they are as big in file size similarly to the super G and D models. Found also "Applio-Recode" and a runtime folder that seemed good to have but outdated.
I had major issues with conda and python as I had py312 installed while these repos are based on runtimes, only one was "DLM", I absolutely must find a way to create one fucking site-packages environment and stick to it in my env path.
I got one to work today, Applio 3.0.9 (released today) as it is a precompiled archive file, ezclap. But the entire fucking reason I spent so much time on all of this is to discover RMVPE+ new inferring method which can set frequencies or midi notes that will not be exceeded. I was thankful for your help last time, code0. I do crazy trained models and go completely berzerk sometimes, I fucking love this project and you made me discover Mangio-Crepe "Hops" function, very accurate when you know what you're doing, you were right (monophonic).
But I was never able to make my new forked RVC_v3 HF and my new Aesthetic RVC Inference HF to work, which they both contain the new hybrid version and RNMVE+
I'm very tired, can I have someone to carry me at this point 😅
I tried the new pretrained models and I haven't found any information that indicates what they do more or less than others
haughty ice
When training an RVC, is it better if the data is a sining voice or just a talking voice? Or is it better to have both?
rare gobletBOT
Ayo? @haughty ice level 1 !!!
trim star
Both for singing
Code mentioned somethinf about this
At the end of thr day, leave the pitch shift on
Make sure you remove emptiness from your dataset as well hehe
Anyways, I really love this Applio forked version, it contains autotune, split audio into segments for inferences, compression along with voice cleaner/denoiser directly from the inferring menu
abstract kestrel
Someone please separate this😭😭(lead,back,Inst,highnote&adlibs)
trim star
brittle wing
huh
trim star
No idea bro
rare gobletBOT
Ayo? @trim star level 3 !!!
pastel oak
No idea bro
hes asking why you copy pasted his screenshot without any comment to it
swift lotus
Can someone help me with training a model? The Google Colab is a pain since my laptop doesnt use GPU so there are constant errors
rare gobletBOT
Ayo? @swift lotus level 1 !!!
grim bay
Can someone show me an example of a dataset?
thick dagger
How to fix echo sound after train ?
glad zealot
Can someone show me an example of a dataset?
Dataset is just the audio you are using to train
glad zealot
How to fix echo sound after train ?
Echo?
When you're using the model it sounds like there's an echo?
azure marshBOT
RVC Guides (How to Make AI Cover)
thick dagger
nope
glad zealot
What is it then
thick dagger
It sound like lower tone after sound voice ai
glad zealot
Did you set the pitch/tune?
thick dagger
i set pitch -12
rare gobletBOT
Ayo? @thick dagger level 2 !!!
thick dagger
Do you have another colab not CoverGen_NO_UI_v2_en.ipynb ?
glad zealot
There's the ui version and Ilaria rvc
But on Ilaria rvc you need it to have no background music just pure vocal for the audio you want to change the voice
thick dagger
can i get a link and guide
glad zealot
-colab
brittle wing
@fluid matrix how can I continue training with a preprocessed dataset?
in RVC CLI
the new notebook you sent to me
next belfry
what does RVC mean
brittle wing
what does RVC mean
retrival based cloning(?)
rare gobletBOT
Ayo? @brittle wing level 5 !!!
glad zealot
retrival based cloning(?)
Close, it's retrieval based voice cloning
next belfry
Close, it's retrieval based voice cloning
what does that mean
glad zealot
what does that mean
It just tries to copy the voice by using the pitch and tones of the voice
brittle wing
My dumbass thought it was real voice changer 😭💀
glad zealot
My dumbass thought it was real voice changer 😭💀
Can be lol
next belfry
It just tries to copy the voice by using the pitch and tones of the voice
Oh then how do u use it
rare gobletBOT
Ayo? @next belfry level 1 !!!
next belfry
How do i use the audio I downloaded?
glad zealot
How do i use the audio I downloaded?
You need to remove background noises first, and remove echoes, remove voices that you don't want, etc.
brittle wing
Can be lol
How so
glad zealot
Oh then how do u use it
You put the new audio to the program and train it there
next belfry
You need to remove background noises first, and remove echoes, remove voices tha...
What app do i use thou
Oh
glad zealot
@half cove what's the command for guides?
brittle wing
<#1159290161683767298>
Thanks for not making me feel like a dumb person haha
half cove
<@936673139419664414> what's the command for guides?
which ones?
glad zealot
which ones?
Training
half cove
-links
azure marshBOT
glad zealot
Im trying voice.ai but i cant figure it out
We don't do support for voice.ai
glad zealot
The 3 main atm is applio, mainline, mangio
glad zealot
Links here
next belfry
are they free ?
glad zealot
how do u download that
Which one are you gonna use?
Also more applio stuff in #📰│dev-updates
glad zealot
Just 1
next belfry
the last one?
rare gobletBOT
Ayo? @next belfry level 2 !!!
next belfry
3.0.9 one?
glad zealot
Ye the latest
brittle wing
Feature extract
can you elaborate on this[
next belfry
5 GB bruh 😭
glad zealot
can you elaborate on this[
There should be a button or something called feature extract on the program you are using
glad zealot
5 GB bruh 😭
Ahh you also need a decent gpu to run it
next belfry
Ahh you also need a decent gpu to run it
which one do u use?
glad zealot
-colab
glad zealot
You can use rvc disconnected or applio on colab to train
next belfry
what does that mean now 😭
glad zealot
@half cove halp
glad zealot
what does that mean now 😭
Google colab is basically just an online way of running programs
next belfry
bro this voice mod is so much more diffcult than i thought 😭
i just wanna be harbor in valo yo 😭
glad zealot
i just wanna be harbor in valo yo 😭
Voice changer?
next belfry
yeah
glad zealot
But not sure what it will like
As long as you still have the preprocessed audio, the feature extract should run fine
brittle wing
There should be a button or something called feature extract on the program you ...
but what is it supposed to do? I want to use a dataset that's already processed. And I want to continue training the D and G files
next belfry
Still needs gpu
how do u check it
glad zealot
but what is it supposed to do? I want to use a dataset that's already processed....
If it's just preprocess, it won't do anything, you need to feature extract it first
grim bay
As long as you still have the preprocessed audio, the feature extract should run...
Could you elaborate?
I'm not sure I understand
glad zealot
@half cove this too
glad zealot
Could you elaborate?
You said you have the pre processed audio right?
half cove
I'm not sure I understand
what you need help with?
glad zealot
They have preprocessed audio but don't know what to do next from what I'm reading
half cove
if they preprocessed as in by rvc or the dataset being cleaned
if the files are pre-processed by rvc next step is training
brittle wing
im here now
finally, poopmaster is here
brittle wing
<@456226577798135808> what stage are you at rn and what you usin
RVC CLI. I have the processed dataset and a D and G file with 120 epochs
(RVC CLI on colab)
grim bay
You said you have the pre processed audio right?
To be honest, now I need to cut and find which parts of the voices to put into the dataset
half cove
RVC CLI. I have the processed dataset and a D and G file with 120 epochs
d and g means you trained it
i think on rvc cli the main pth is in logs
grim bay
what you need help with?
I don't know when I choose the parts in the Anime, I need to know what is the result for a dataset
Because idk which voice parts is good to put into the dataset
half cove
I don't know when I choose the parts in the Anime, I need to know what is the re...
so you want to know how to make a dataset?
grim bay
so you want to know how to make a dataset?
Yeah
half cove
try reading this ig https://docs.aihub.wtf/rvc/resources/datasets/
if you have any questions then ask
half cove
just make sure its the voice with nothing else such as noise sfx etc
brittle wing
d and g means you trained it
yes
how do I point it do my processed dataset
and how do I point it to my d and g files
did you look at the notebook I posted/
grim bay
But I want to know which part should I cut
brittle wing
?
grim bay
Because in the Anime, sometimes they keeps silence for a few seconds
So I need to know should I remove it?
half cove
and how do I point it to my d and g files
you want to retrain?
half cove
So I need to know should I remove it?
yes remove silence