Alright ty
#✨│ai-help
candid crescent
minor zodiac
HELP
HOW DO I KEEP TRAINING A MODEL?
I had a power outage while training the model
rare gobletBOT
Ayo? @minor zodiac level 2 !!! 
odd shale
<@628603171735994379> Help?
Don't ping staffers unnecessarily, please.
But okay, what's the issue?
minor zodiac
I had a power outage while training the model
-I had a power outage while training the model
odd shale
-I had a power outage while training the model
Don't repeat sentences.
minor zodiac
So, how do I keep training the model with the saved epochs?
odd shale
So, how do I keep training the model with the saved epochs?
Which samplerate you selected when training for first time your model?
minor zodiac
40k, v2
minor zodiac
Alright, What's next?
odd shale
Alright, What's next?
Then, on "experiment name" Write the exact name of your model folder (you can find it on logs folder)
minor zodiac
Done
low shard
mmm i will see about it, meanwhile if you want you can use the hugging face space of ilaria rvc, which runs on cpu, so its slower, but atleast for the moment it works using the gdrive links here: https://huggingface.co/spaces/TheStinger/Ilaria_RVC (remember to click the top right 3 dots, duplicate the space and wait for it to finish, then you will be able to use it)
odd shale
Done
Set the same batch size you used previously and same save frequency
You can change the total epoch amount if you want
minor zodiac
Already done, same for the Total Epochs and CPU process
odd shale
Already done, same for the Total Epochs and CPU process
Don't press on preprocess neither feature extraction
Just press on "Train model"
And that's it
candid crescent
does this mean it's done? cause it says different on the guide and it's completely stopped moving
minor zodiac
What about the Pre-trained section?
odd shale
What about the Pre-trained section?
Why the pretrain section is blank?
minor zodiac
Cuz I wanna resume the model training
odd shale
That thing should NOT be on blank
candid crescent
does this mean it's done? cause it says different on the guide and it's complete...
is this uh, an error?
rare gobletBOT
Ayo? @candid crescent level 5 !!!
candid crescent
k, I am stuck. I'm not able to train feature index and my Feature extraction gets stuck.. can sm1 help?
half cove
please move it out of onedrive 😭
ashen sleet
So I run go-web.bat and get this warning (but the web interface successfully opens)
C:\Users\jhump\AppData\Roaming\Python\Python39\site-packages\scipy\__init__.py:146: UserWarning: A NumPy version >=1.16.5 and <1.23.0 is required for this version of SciPy (detected version 1.23.5
warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"
Found GPU NVIDIA GeForce RTX 3070
Set fp16_run to true in 32k.json
Set fp16_run to true in 40k.json
Set fp16_run to true in 48k.json
Use Language: en_US
Running on local URL: http://0.0.0.0:7897```Ayo? @ashen sleet level 1 !!!
candid crescent
please move it out of onedrive 😭
ah is that the issue?
ashen sleet
And then when I click convert I get this
future: <Task finished name='v8ub4gt7lcn_11' coro=<Queue.process_events() done, defined at C:\Users\jhump\Documents\Inform\Projects\Mangio-RVC-v23.7.0\runtime\lib\site-packages\gradio\queueing.py:343> exception=1 validation error for PredictBody
event_id
Field required [type=missing, input_value={'data': [0, 'C:/Users/jh...on_hash': 'v8ub4gt7lcn'}, input_type=dict]
For further information visit https://errors.pydantic.dev/2.6/v/missing>
Traceback (most recent call last):
File "C:\Users\jhump\Documents\Inform\Projects\Mangio-RVC-v23.7.0\runtime\lib\site-packages\gradio\queueing.py", line 347, in process_events
client_awake = await self.gather_event_data(event)
File "C:\Users\jhump\Documents\Inform\Projects\Mangio-RVC-v23.7.0\runtime\lib\site-packages\gradio\queueing.py", line 220, in gather_event_data
data, client_awake = await self.get_message(event, timeout=receive_timeout)
File "C:\Users\jhump\Documents\Inform\Projects\Mangio-RVC-v23.7.0\runtime\lib\site-packages\gradio\queueing.py", line 456, in get_message
return PredictBody(**data), True
File "C:\Users\jhump\Documents\Inform\Projects\Mangio-RVC-v23.7.0\runtime\lib\site-packages\pydantic\main.py", line 171, in __init__
self.__pydantic_validator__.validate_python(data, self_instance=self)
pydantic_core._pydantic_core.ValidationError: 1 validation error for PredictBody
event_id
Field required [type=missing, input_value={'data': [0, 'C:/Users/jh...on_hash': 'v8ub4gt7lcn'}, input_type=dict]
For further information visit https://errors.pydantic.dev/2.6/v/missing```
azure marshBOT
RVC Guides (How to Make AI Cover)
ashen sleet
So I run `go-web.bat` and get this warning (but the web interface successfully o...
So I guess I need a differnet version of something called NumPy...
Can anyone help me explore that?
I'm not sure how to update to a lower version of something
candid crescent
please move it out of onedrive 😭
It didn't help. still unable to make a index file
half cove
It didn't help. still unable to make a index file
prob didnt preprocess properly
candid crescent
half cove
still in OneDrive
half cove
Am I doing something wrong?
are you sure thats the right file path?
half cove
should be right file path
candid crescent
Should I restart RVC as a whole and try again?
half cove
where is your rvc folder
candid crescent
ah ffs
doesn't it have to be in the same space as the w-okada though?
should I move the whole thing?
minor zodiac
That thing should NOT be on blank
Thank you, it's working now 👍
celest kettle
guys whatever model i use it no worky
like it dont download
the cell js stops the second i start it
odd shale
the cell js stops the second i start it
What colab are you trying to use?
candid crescent
move the rvc folder out of onedrive
still don't fix
half cove
prob isnt able to find it
half cove
oh fuck im dumb
candid crescent
no you aren't 🤦
half cove
@candid crescent put it in a folde then put file path of the folder
candid crescent
<@444843948830097408> put it in a folde then put file path of the folder
alrighty
Every folder I make is just turning into onedrive folders 😡
half cove
candid crescent
I LEGIT DO NOT HAVE ONEDRIVE
HOW IN GOD'S GREEN EARTH IS MY FILE A ONEDRIVE FILE
half cove
average windows experience
summer mirage
It's very strange but every time I want to convert I get an error like this. I didn't do anything different than normal and it was all of a sudden. can someone please help?
candid crescent
average windows experience
after struggling for an hour to get it to not be onedrive, it still doesn't work in a file non-onedrive
half cove
how ar you creating the folders
sinful nexus
-colab
candid crescent
Right clicking in my Local disk C area, and creating a new file.
half cove
that should just make a normal folder
candid crescent
yes sir
fading drum
thank you
half cove
Do without the /voice.flac
minor zodiac
I have a question, the more epochs the model has, the better it will sound?
half cove
I have a question, the more epochs the model has, the better it will sound?
No
Overtraining exists
noble heath
can someone give me a link to the new rvc
candid crescent
You need folder with audio
❤️
that was the issue
I was linking the file not the folder
half cove
can someone give me a link to the new rvc
-links
azure marshBOT
Suggestions for @noble heath
noble heath
which one is the rvc website
sinful nexus
-colab
candid crescent
is it just.. stuck?
candid crescent
how can I tell when overtraining occurs?
half cove
check cmd
candid crescent
--it'll tell me?
candid crescent
oh and I went to stop the training and it didnt stop
The training model area just didnt change
half cove
just close cmd to stop training
brittle wing
hi, for some reason the "choose the model" panel doesn´t show any model, i put a drive link
candid crescent
just close cmd to stop training
Wait do you mean TB will tell me when its overtrained? or the cmd itself
thin stump
hi, for some reason the "choose the model" panel doesn´t show any model, i put a...
okada? refresh the page
half cove
Wait do you mean TB will tell me when its overtrained? or the cmd itself
tb will show when its overtrained then u get the closest pth before overtraining
brittle wing
okada? refresh the page
ok
rare gobletBOT
Ayo? @brittle wing level 1 !!!
half cove
then open the localhost link
candid crescent
and... just nothing happened
rare gobletBOT
Ayo? @candid crescent level 6 !!!
candid crescent
oh is it that easy?
half cove
screenshot
brittle wing
okada? refresh the page
i did it and it keep doesn´t work
candid crescent
candid crescent
half cove
try deleting the tensorvenv file and running it again
candid crescent
it's from the official guide page
candid crescent
try deleting the tensorvenv file and running it again
oh it's in the rcv folder
is it not supposed to be here 👀
half cove
no put it in the rvc1006 nvidia folder
candid crescent
oh it's doing different things now
wait okay I have TB open now.
How is this gonna help me see the overtrain?
half cove
also make sure you click this while training
half cove
How is this gonna help me see the overtrain?
ok so
candid crescent
half cove
open loss/g/total
candid crescent
open loss/g/total
a file?
candid crescent
yes
half cove
why can i not spell
frosty night
-rvc
half cove
now when you train this will have a line
candid crescent
now when you train this will have a line
and it will tell me when it's overtrained?
with what signal
thin stump
i did it and it keep doesn´t work
you can manually input the files then refresh if it doesn't upload. that's what i meant
candid crescent
when shit goes up
like when the line goes up
half cove
read guide
brittle wing
you can manually input the files then refresh if it doesn't upload. that's what ...
ok
candid crescent
also make sure you click this while training
this on Yes or no?
half cove
this on Yes or no?
on yes
brittle wing
you can manually input the files then refresh if it doesn't upload. that's what ...
how i do that?
thin stump
I don't have okada downloaded since I only use colab for it. It should be near the top center-left of your screen
lol im way off hold on
thin stump
oh, i see
brittle wing
idk what to do
thin stump
oh yeah that is definitely important
https://discord.com/channels/1159260121998827560/1213509354343637065 TDLR; you have to look at 4 graphs: g/total, d/total, kl, mel. If one of them goes up it's done. If it's constant keep training
candid crescent
https://discord.com/channels/1159260121998827560/1213509354343637065 TDLR; you h...
4 graphs???
this is gonna be a long staring contest
thin stump
you're mainly looking at g/total but if you feel like something gone wrong like g dipping so much then look at your other graphs
candid crescent
so if any of the 4 go up, it's done. no matter the positions of the others?
thin stump
i'm talking a huge climb. Small ones you can wait and see if they go back down like kl
brittle wing
well, how i share a hugging face link?
thin stump
well, how i share a hugging face link?
I can't really help sorry, I never used ilaria
brittle wing
ok
barren axle
hey guys! my step 2 on the google collab is not working after I paste the zip link. Do you know what might be the problem?
rare gobletBOT
Ayo? @barren axle level 1 !!!
half cove
https://discord.com/channels/1159260121998827560/1213509354343637065 TDLR; you h...
for new users lookin and only loss g and testing checkpoints should be good enough
half cove
well, how i share a hugging face link?
barren axle
It says: ModuleNotFoundError: No module named 'mega'
thin stump
so if any of the 4 go up, it's done. no matter the positions of the others?
but yeah it's impossible to overtrain since you have your checkpoints
half cove
It says: ModuleNotFoundError: No module named 'mega'
what colab u usin?
barren axle
https://huggingface.co/RVC123/RVC/resolve/main/shakira_crepe.zip this is the one
half cove
"colab"
proven hill
I can't really help sorry, I never used ilaria
you never used me?
barren axle
rugged stirrup
What's the difference between normal rmvpe VS rmvpe_gpu?
half cove
https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sh...
please use a newer colab
half cove
What's the difference between normal rmvpe VS rmvpe_gpu?
one uses gpu other dont
barren axle
ohh sorry! thank you!
half cove
ohh sorry! thank you!
-colab
rugged stirrup
one uses gpu other dont
Oh ok
candid crescent
holy, how long does it take for 1 epoch?? and I have to wait for this to level?
candid crescent
does this only move when it hits the save point?
candid crescent
how long ur dataset
18 mins long
proven hill
https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sh...
thsi is super old
half cove
18 mins long
and what batch size
proven hill
use mine
half cove
also show what cmd says
proven hill
proper shale
does this only move when it hits the save point?
yeah, also refresh TensorBoard with the refresh button within the orange bar above the graphs
*when that happens
half cove
it does take time
proper shale
https://youtu.be/wh4cLMLLkj0
mj in the thumbnail? based
proven hill
mj in the thumbnail? based
i did it
unironically
for you
thats it, the only reason its there
proven hill
i swear to my mother
half cove
also @candid crescent on tensorboard go on the settings icon top right and turn on auto refresh
proven hill
i thought of you lol
proper shale
based
half cove
makes life easy
candid crescent
also <@444843948830097408> on tensorboard go on the settings icon top right and ...
Already on every 30s
proven hill
hope the tutorial is useful btw
half cove
Already on every 30s
good
rugged stirrup
whats this mean? ValueError: 40000 SR doesn't match target 32000 SR
proper shale
ya changed sample rate in the middle of training?
rugged stirrup
I didnt
proper shale
or after you resumed
rugged stirrup
I hit train model and it did that
proper shale
try selecting 32k and then training... 
rugged stirrup
oh ok
candid crescent
Do I just press train model again when it overtrains?
rugged stirrup
wait no I did select 32k
proven hill
whats this mean? ValueError: 40000 SR doesn't match target 32000 SR
what are you using?
rugged stirrup
what are you using?
newest RVC out
proven hill
explain it to me in details
proven hill
newest RVC out
mainline?
proven hill
youre trying to train, what settings did you used?
proven hill
why did you use 16 for the CPU?
rugged stirrup
it defaulted to that
proven hill
what CPU do you have?
rugged stirrup
proven hill
okay you can leave at 16 then
rugged stirrup
I used a flac file format if that matters
proven hill
whats this mean? ValueError: 40000 SR doesn't match target 32000 SR
and this error appears when you press train?
rugged stirrup
and this error appears when you press train?
yeah
proven hill
I used a flac file format if that matters
its good
go to logs and delete the folder of the model youre making
rugged stirrup
alright
proven hill
then start again everything, use 32k from the beginning
rugged stirrup
alright ill try this again
proven hill
always use 32k
proven hill
alright ill try this again
let me know
rugged stirrup
so never use 40k? or 48 cause that would cause artifacts?
proven hill
yup
rugged stirrup
I thought it was the quality of the voice or something
magic hawk
help?
rugged stirrup
alright training the feature index worked much faster this time
half cove
I thought it was the quality of the voice or something
90% of datasets should be used with 32k
proven hill
no its not, voices usually never go above 16k (for one side, so for two sides ((stereo)) its 16x2 = 32k)
magic hawk
can you guys give the link for google research drive
rugged stirrup
alright training the model now
half cove
can you guys give the link for google research drive
-colab
proven hill
alright training the model now
problem solved?
magic hawk
yess
rugged stirrup
problem solved?
yeah now its working thanks
magic hawk
thx
proven hill
yeah now its working thanks
no prob!
proven hill
can you guys give the link for google research drive
you can find it here + video tutorial https://youtu.be/wh4cLMLLkj0
rugged stirrup
no prob!
oh and also how do I see my model training on tensorboard? I have it installed and everything I just dont know how to view the training it shows me like a ton of graphs
proven hill
oh and also how do I see my model training on tensorboard? I have it installed a...
i never use tensorboard, wait for someone who does
rugged stirrup
alright
proven hill
pretty sure there was a guide on it tho
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 💖
proven hill
apparently not here
half cove
oh and also how do I see my model training on tensorboard? I have it installed a...
Let me get on pc and send the bat file
1 sec
rugged stirrup
Oh I already have it
half cove
Nice
rugged stirrup
I followed the guide thing and it installed it this is what it looks like I cant see my model training
half cove
I selected it and refreshed
what epoch is it at
half cove
it should be showing
rugged stirrup
refreshed and all graphs are still blank
well if that doesnt work what should I look for to get the best model?
candid crescent
Uh.. how do I stop the training
😭
wont stop!!
MY EPOCHES WON'T STOP!!
im gat ola
Okay my first model is... realistic-ish, not enough for what I want but
fading drum
Im now having an issue compressing the voice folder thing down to under 2gb so i can use it with replay, the folder size if 3.9gb, and the lowest i can comrpess it is 2.7gb, any advice to get it lower?
fading drum
Uh.. how do I stop the training
use task manager to terminate python
or press ctrl c in the cmd
thin stump
Is there a way to change the image submission in #1175430844685484042 or do I do that manually on weights?
sinful nexus
-colab
normal vigil
why i cant open the rvc after i close the tab
azure marshBOT
RVC Guides (How to Make AI Cover)
odd shale
MY EPOCHES WON'T STOP!!
Epochessy
odd shale
MY EPOCHES WON'T STOP!!
It's because you did set 1000 epochs as goal
solemn dagger
Does anyone know the best model in UVR to get rid of echo?
candid crescent
It's because you did set 1000 epochs as goal
oops
viscid moss
Does anyone know the best model in UVR to get rid of echo?
De-Echo-De-Reverb
steady ferry
I'm having an error installing AICoverGen on my PC with a new install of windows 10.
it says "Fairseq failed to build wheel" and has an error with Yvals.h(20) fatal error c1083 Cannot open include file 'crtdbg.h' and cl.exe fails with exit code 2
Its the same error every time.
i'm using Python 3.9 with Pip version 24 on Windows 10 22H2 (build 19045.4046)
I already installed the newest ffmpeg, Sox 14.4.2, Visual Studio 2022 and VSBuildTools 2022 and added them to the User and System Variable Paths (my windows account has admin privilege's)
I've tried using VS 2022 and VS 2019 with the matching version of BuildTools and re-installing them multiple times.
I made sure Desktop Development with C++ and Universal Windows Platform Development were selected with the latest Windows 11 and Windows 10 SDK (also restarted the PC after each try)
Any ideas on what could be happening?
odd shale
I'm having an error installing AICoverGen on my PC with a new install of windows...
If you talk english then use #✨│ai-help
Altho you can ask Eddy about your error on that same channel
dire vapor
I have a 22 minute dataset, how many batch sizes should I use?
brittle wing
fading drum
Im now having an issue compressing the voice folder thing down to under 2gb so i...
yeah and me too
Solution: Delete other pths
proven hill
I have a 22 minute dataset, how many batch sizes should I use?
whats your gpu?
or are you using colab?
odd shale
tends =/= it is always
Okay, i was wrong.
rare gobletBOT
Ayo? @golden basin level 1 !!!
odd shale
Anyone who wants to use W-Okada, must install VB Cable
limpid frigate
what is epochs?
spiral osprey
When using deEcho-deReverb do I need to check a box? in the vocal isolation tutorial it doesn't say anything.
thin stump
When using deEcho-deReverb do I need to check a box? in the vocal isolation tuto...
No reverb only
you only leave the box blank when you are using BVE or HP_5 & 6 models. And you can get model maker if you read this https://aihubdocs.github.io/en/extra/model-maker-role/#model-maker-role
spiral osprey
No reverb only
and UVR-DeNoise do I need to check a box?
thin stump
no noise only ofc
spiral osprey
no noise only ofc
ok,thanks.
thin stump
@gusty merlin Don't spam general. Read everything here https://aihubdocs.github.io/en/essentials/how-to-make-voice-models/
gusty merlin
<@1207312334989819908> Don't spam general. Read everything here https://aihubdoc...
my english bad ;-;
rare gobletBOT
Ayo? @gusty merlin level 3 !!!
gusty merlin
okk
thin stump
Use DeepL for translations
drowsy steppe
I'm having a bit of a problem with my AI covers
The AI Voice like strains when it tries to hit high notes, is there a way to fix this? (Also yes I'm new to this)
gusty merlin
why my tensorboard is "No dashboards are active for the current data set."
is it normal?
azure marshBOT
RVC Guides (How to Make AI Cover)
upbeat coral
I trained my voice for a model, and used RVC for a cover of morgan wallen's last night.
at some parts there is a bit of faint robotic/static (1:44 is perfect example) also some of the lyrics are getting a bit slurred or mixed up and not correct so any advice? for the most part it sounds acceptable to me
zinc mesa
One message removed from a suspended account.
tacit badger
low shard
what is .index for
added.index is the file of the voice model that contains the accent of the ovicr
low shard
One message removed from a suspended account.
there are some tts, look in the edge tts section, there is a part mixed with rvc where you can use rvc models #1212368971127590922
low shard
i have that file in my drive what step should i skip
do you wanna make a model or use a model
gusty merlin
do you wanna make a model or use a model
make
low shard
make
in rvc disconnected, run the dipendencies, put the same local training variables that you used for that model and run the cell
then run this
and this
gusty merlin
in rvc disconnected, run the dipendencies, put the same local training variables...
ooooooooo
rare gobletBOT
Ayo? @gusty merlin level 4 !!!
low shard
is there a tips to not disconected
If its for inactivity, either click randomly every 5 mins or follow this guide #1213586188091523075
gusty merlin
noooo usage limits ;-;
low shard
noooo usage limits ;-;
if you cant connect to gpu backend, you can just wait 12-24 hours till tmr to resume training
gusty merlin
can i use other acc?
low shard
yea but it wont resume from it
gusty merlin
ooo
fathom carbon
is there a colab where you can upload the model instead of giving a hugginface file? :/
rare gobletBOT
Ayo? @fathom carbon level 1 !!!
upbeat coral
is there a colab where you can upload the model instead of giving a hugginface f...
the huggingface files are models.
What do u mean
fathom carbon
ye ik but i have my own model that i am too stipid to upload to huggingface
upbeat coral
what do you need to do with it
fathom carbon
just a cover
upbeat coral
did you merge the song yet
fathom carbon
ye
upbeat coral
you trained the model, did you do it over a song for cloning?
ok
so you can do youtube
post it on youtube and send the link
or soundcloud
make a soundcloud
low shard
is there a colab where you can upload the model instead of giving a hugginface f...
btw, the applio colab does
river canopy
This option should only be ticked if you don't want your model listed on the public tracker. I'm level 1😭
rugged stirrup
is this fine that it had a spike down will the model still sound good?
gusty merlin
is in rvc disconected auto save?
low shard
This option should only be ticked if you don't want your model listed on the pub...
public tracker? What colaba re you using
low shard
is in rvc disconected auto save?
nope
carmine hearth
Hi, I have a question: is applio colab not available on the free tier?
When I tried to run applio colab, it didn't warn me that it contains code that is not allowed on the free tier, but after training for about 20 epochs, it cut off my runtime with a notification that it is not allowed on the free tier (not inactivity or max duration), and the second time I tried it, it cut off my runtime at 100 epochs.
Of course I know that I can use rvcdisconnected to train on the free tier, but I want to be sure about this.
Sorry if my English grammar is weird, I'm using a machine translator.
vivid wraith
I’m so confused, how do I make an ai voice cover? 😭
gusty merlin
-colab
thin stump
is this fine that it had a spike down will the model still sound good?
What happened during the g dip? You can check the other graphs
rugged stirrup
I dont really know what happened
thin stump
I dont really know what happened
they share the same step value so you can check
rare gobletBOT
Ayo? @thin stump level 16 !!!
thin stump
kl, mel, d
rugged stirrup
This?
thin stump
lol i can't read that. I meant the graphs
rugged stirrup
here is a bunch of loss graphs
thin stump
Ohh, that's interesting now. You're d/total peaked at 30k so 35k on g/total is overtraining
rugged stirrup
So its getting overtrained?
gusty merlin
how to continue my training but disconected. now im 20ep?`
thin stump
So its getting overtrained?
yup, I can tell thats a suspicious dip
rugged stirrup
Well here is all the models with epochs I didnt save a ton because space but which one would likely be the best? if you could somehow tell me
thin stump
unless i'm reading the wrong one? Why is there so many d totals
rugged stirrup
Dataset is about 37 minutes long
rugged stirrup
unless i'm reading the wrong one? Why is there so many d totals
Idk
gusty merlin
how to continue my training but disconected. now im 20ep?`
help
rugged stirrup
Loss d total?
thin stump
just copy and paste what i wrote so it pops up. Yeah
rugged stirrup
thin stump
oh, it's all good. How is g/total?
rugged stirrup
rare gobletBOT
Ayo? @rugged stirrup level 10 !!!
thin stump
Not so sure why its dipping and going up like that
it's a collapse. Can't prevent it from happening 
rugged stirrup
Well does that mean the model is trashed?
thin stump
I'm not the judge of that so just use this and see how it sounds
rugged stirrup
I thought model collapse means the model is like awful or something
thin stump
How do I stop training cause there isnt any stop button in RVC
Same way you started training right? I only use colab
thin stump
I’m so confused, how do I make an ai voice cover? 😭
https://aihubdocs.github.io/en/essentials/how-to-make-ai-cover/ . I'm not trying to dono wall anyone. All of it is explained in that guide 
plush compass
I'm really sorry to ask this question, but where can I get the dataset for anime characters? youtube?
hot lagoon
hopeing to find someone who use the waves harmony
dusky ferry
I trained my voice for a model, and used RVC for a cover of morgan wallen's last...
how do I train my voice?
vivid wraith
https://aihubdocs.github.io/en/essentials/how-to-make-ai-cover/ . I'm not trying...
Where do I paste the link of the model?
thin stump
Where do I paste the link of the model?
It's a huggingface link so like this. You right click and copy the link of the download button
molten flare
how do i upload model on easygui?
thin stump
Where do I paste the link of the model?
or weights.gg to download models
since i'm not sure which one you are using for ai covers
and you can also look in #1175430844685484042
brave sierra
im having difficulties launching the app, it worked the first time but after a couple minutes it stopped working, then it closed, and now it wont re-open (couldnt find another tech help channel)
it wont let me send a screenshot of the problem im having which is funny
rare gobletBOT
Ayo? @brave sierra level 1 !!!
thin stump
@vivid wraith dm'd you
undone valve
So finally colab let me back in, but I can't resume the training :/
it tells me:
Exception: No features exist for this model yet. Did you run Feature Extraction?
I guess it has to do with that:
2024-03-09 10:20:52 | INFO | fairseq.tasks.hubert_pretraining | current directory is /content/Mangio-RVC-Fork
2024-03-09 10:20:52 | INFO | fairseq.tasks.hubert_pretraining | HubertPretrainingTask Config {'_name': 'hubert_pretraining', 'data': 'metadata', 'fine_tuning': False, 'labels': ['km'], 'label_dir': 'label', 'label_rate': 50.0, 'sample_rate': 16000, 'normalize': False, 'enable_padding': False, 'max_keep_size': None, 'max_sample_size': 250000, 'min_sample_size': 32000, 'single_target': False, 'random_crop': True, 'pad_audio': False}
2024-03-09 10:20:52 | INFO | fairseq.models.hubert.hubert | HubertModel Config: {'_name': 'hubert', 'label_rate': 50.0, 'extractor_mode': default, 'encoder_layers': 12, 'encoder_embed_dim': 768, 'encoder_ffn_embed_dim': 3072, 'encoder_attention_heads': 12, 'activation_fn': gelu, 'layer_type': transformer, 'dropout': 0.1, 'attention_dropout': 0.1, 'activation_dropout': 0.0, 'encoder_layerdrop': 0.05, 'dropout_input': 0.1, 'dropout_features': 0.1, 'final_dim': 256, 'untie_final_proj': True, 'layer_norm_first': False, 'conv_feature_layers': '[(512,10,5)] + [(512,3,2)] * 4 + [(512,2,2)] * 2', 'conv_bias': False, 'logit_temp': 0.1, 'target_glu': False, 'feature_grad_mult':
rare gobletBOT
Ayo? @undone valve level 1 !!!
undone valve
0.1, 'mask_length': 10, 'mask_prob': 0.8, 'mask_selection': static, 'mask_other': 0.0, 'no_mask_overlap': False, 'mask_min_space': 1, 'mask_channel_length': 10, 'mask_channel_prob': 0.0, 'mask_channel_selection': static, 'mask_channel_other': 0.0, 'no_mask_channel_overlap': False, 'mask_channel_min_space': 1, 'conv_pos': 128, 'conv_pos_groups': 16, 'latent_temp': [2.0, 0.5, 0.999995], 'skip_masked': False, 'skip_nomask': False, 'checkpoint_activations': False, 'required_seq_len_multiple': 2, 'depthwise_conv_kernel_size': 31, 'attn_type': '', 'pos_enc_type': 'abs', 'fp16': False}
/usr/local/lib/python3.10/dist-packages/torch/nn/utils/weight_norm.py:30: UserWarning: torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm.
warnings.warn("torch.nn.utils.weight_norm is deprecated in favor of torch.nn.utils.parametrizations.weight_norm.")
move model to cuda
no-feature-todo
I've been using the same dataset i used the 1st time when I first trained my model (it was a zip of a lot of wav files)
I've also tried a new dataset and i have the same error (a zip with only one wav file)
Any idea why it does not work?
fossil rover
hey
i have a problem
so i have a song but when i use it the ai model sounds static and high pitched
any idea why?
undone valve
> 0.1, 'mask_length': 10, 'mask_prob': 0.8, 'mask_selection': static, 'mask_othe...
so i used another mail adress, and it works.
I hope google won't permanently ban for this 😄 at least not before i finish my model 😄
fossil rover
anyone?
latent kettle
so i have a song but when i use it the ai model sounds static and high pitched
low pitch. +12 B/W -12
fossil rover
low pitch. +12 B/W -12
how do i do that again?
latent kettle
how do i do that again?
are you making a cover? or training a model
latent kettle
its a duet
male and female?
fossil rover
it is male
rare gobletBOT
Ayo? @fossil rover level 3 !!!
latent kettle
it was from this server
share me the link
latent kettle
in the voice models tab
which RVC are you using
fossil rover
im trying out appolio
latent kettle
im trying out appolio
in applio just find pitch. and adjust it according to you
vivid wraith
Okay tysm
vivid wraith
I got a ping saying they dmed me idk where it went tho 😭
rare gobletBOT
Ayo? @vivid wraith level 1 !!!
latent kettle
I was saying ty to that :]
aww.. not a problem! yw : )
low shard
Check #📰│dev-updates message
whole violetBOT
half cove
🤔
fossil rover
low pitch. +12 B/W -12
it worked you beautiful genius
latent kettle
it worked you beautiful genius
yw : )
half cove
i can try different shit
might take a while
UI COLABS ARE NOT WORKING RIGHT NOW CHECK #📰│dev-updates message FOR MORE UPDATES
azure marshBOT
RVC Guides (How to Make AI Cover)
spare robin
Greetings Friends. For Format shift inference audio what do Quefrency and Timbre adjust in relation to the audio?
molten timber
Hi. Is there a chance somebody could help me make a cover?
latent kettle
Hi. Is there a chance somebody could help me make a cover?
what happened
latent kettle
Hi. Is there a chance somebody could help me make a cover?
I got it.
but please don't dm me
actually Google colab has banned some Ui based RVC Forks
molten timber
ups, sorry for dms
latent kettle
ups, sorry for dms
not a problem
latent kettle
ups, sorry for dms
molten timber
Thanks, will check it now
dusky ferry
Hi guys, I'm wondering how many samples would I required to train a good voice?
latent kettle
Hi guys, I'm wondering how many samples would I required to train a good voice?
minimum 10 minutes maximum 30-40 minutes
you can also train on 1-5 minutes of dataset. also you can train on some seconds (if quality of your sample is too high ) by reducing batch size
dusky ferry
you can also train on 1-5 minutes of dataset. also you can train on some seconds...
Okay. Thanks a bunch
rare gobletBOT
Ayo? @dusky ferry level 1 !!!
latent kettle
Okay. Thanks a bunch
your welcome : )
glad zealot
https://discord.com/channels/1159260121998827560/1159380240271953940/12159816957...
pretty much all of them
tranquil aurora
hi. For some reason, rvc does not work for me, although I do everything according to the guides. I tried to reinstall more than once and looked for a solution that I couldn’t find. can someone help me please
reef crest
It's very hard to find your way around the various different softwares. How do these models compare to e.g. voice.ai and others? I have yet to find a good comparison of all the different tech, what's state of the art (for realtime usage)?
placid solstice
Hello! When I use the voice changer I can hear myself talking, is there any way that I can disable that?
spiral osprey
I made it in video format because the audio is too long and I can't send it, so can someone please help me? I cleaned the audio following the vocal isolation tutorial, and even so I still think it will sound robotic or very broken, do you know any way to solve it?
tawny cypress
Hello there ! I need some help. I trying to make ai cover with rvc model, but Collab doesnt want to download model... What can I do in this situation ?
split parcel
Can you introduce me the best female model?
half cove
hi. For some reason, rvc does not work for me, although I do everything accordi...
does not work in what way?
half cove
It's very hard to find your way around the various different softwares. How do t...
w-okada is the best for realtime
half cove
Hello! When I use the voice changer I can hear myself talking, is there any way ...
what is your monitor option set to?
reef crest
w-okada is the best for realtime
is it not just a frontend for the various RVC etc. models?
half cove
I made it in video format because the audio is too long and I can't send it, so ...
could be model issue
half cove
is it not just a frontend for the various RVC etc. models?
yes
half cove
Hello there ! I need some help. I trying to make ai cover with rvc model, but Co...
show error
reef crest
well then which type of models is state of the art? RVC or something else?
tranquil aurora
does not work in what way?
When I speak nothing happens
half cove
When I speak nothing happens
realtime?
placid solstice
what is your monitor option set to?
"Default Speakers (Logitech PRO X Gaming Headset) (046d:0aaa)"
rare gobletBOT
Ayo? @placid solstice level 1 !!!
half cove
"Default Speakers (Logitech PRO X Gaming Headset) (046d:0aaa)"
monitor option is to hear yourself
change that and boom
placid solstice
Thank you!
tranquil aurora
show settings
Now for some reason I can’t start it at all, it just says "wait web server..."
rare gobletBOT
Ayo? @tranquil aurora level 1 !!!
half cove
Now for some reason I can’t start it at all, it just says "wait web server..."
try deleting stored_settings.json and restart
keen silo
uhhh
tranquil aurora
show settings
half cove
your using cpu
brittle wing
Rvc still working in mobile?
tranquil aurora
your using cpu
stll not work
half cove
stll not work
what does it say in cmd
tranquil aurora
i switched it to the gpu but it doesnt work anyway
tranquil aurora
and why is ur gain so high
bc i love when everyone's ears hurt from the volume of my microphone
spring orchid
half cove
i thought it wasnt workin on discord 😭
tranquil aurora
i thought it wasnt workin on discord 😭
it didnt working on discrod early
half cove
tranquil aurora
cuz the output was the cable u use that to connect the audio and program
i used cable in discord lol
jaunty cosmos
Hello there ! I need some help. I trying to make ai cover with rvc model, but Co...
Hey ! Do you get this error message too : " # File exists, proceed with creation of creds and client" ? I have the same problem. Last time I used it everything worked perfectly but now I have this message each time I try to upload my model..
half cove
Hey ! Do you get this error message too : " # File exists, proceed with creation...
send colab link
im guessing its the old colab
jaunty cosmos
half cove
yea dont
jaunty cosmos
Arf.. Could you give me the new one please ?
jaunty cosmos
OK I'll try this one ! Thanks 🙂
rare gobletBOT
Ayo? @jaunty cosmos level 1 !!!
fossil yoke
Hi, my RVC works just fine but my friend still can hear my voice along with the changed one. do you know what might be causing this?
iron cobalt
Hello. Can anyone recommend me any RVC for training AI voices, not Google Colab? I mean creating my own AI voice models from a single audio file.
rare gobletBOT
Ayo? @iron cobalt level 2 !!!
iron cobalt
I have an audio file with a voice of one fictional character
gaunt nebula
https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing#scrollTo=OVQoLQJXS7WX in Step 2 i have JSONDECODE Error, how to fix it?
doesn't matter what voice i try to use :/
latent kettle
Hello. Can anyone recommend me any RVC for training AI voices, not Google Colab?...
rvcDisconnected
on Google colab
low shard
https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sh...
outdated colab
check the new ones in https://aihubdocs.github.io/en
gaunt nebula
thank you
rare gobletBOT
Ayo? @gaunt nebula level 1 !!!
low shard
yw
low shard
on Google colab
he said not on colab tho
spiral osprey
could be model issue
I haven't trained yet, I'm saying is the dataset clean?
half cove
oh
latent kettle
he said not on colab tho
I didn't read that. sorry. that's my fault
half cove
i would need to check the audio to say
half cove
use newer colab
gaunt nebula
i have cpu, and no graphic card, so... xD
gaunt nebula
@half cove Applio will work on cpu without graphic card? ;p
half cove
it would work but slow
brittle wing
sup bros, im looking for a ""FREE"" vst of Supertone Clear, does anyone has it?
rare gobletBOT
Ayo? @brittle wing level 2 !!!
half cove
i recommend the colab
gaunt nebula
but on colab i have this jsondecode error :/
half cove
what colab
gaunt nebula
half cove
thats not the applio colab....
gaunt nebula
i know xD i just tell that i was have error and now i'm trying applio
half cove
applio one should work
gaunt nebula
@half cove how to get voice model and index file for Applio?
half cove
<@936673139419664414> how to get voice model and index file for Applio?
training or infering?
quaint heron
Guys which TTS is best for anime characters? So many choices but which one is considered the most consistent?
ivory patio
rare gobletBOT
Ayo? @quaint heron level 1 !!!
ivory patio
Bro come on still same error
half cove
there is a download tab
on the gui
ivory patio
What the link ?
gaunt nebula
sorry i'm noob and new here
azure marshBOT
RVC Guides (How to Make AI Cover)
half cove
What the link ?
-colab
quaint heron
Guys which TTS is best for anime characters? So many choices but which one is considered the most consistent?
its confusing
ivory patio
Ohhh
gaunt nebula
maybe stupid question but... someone knew if there is AI making instrumental based on vocals only?
ivory patio
maybe stupid question but... someone knew if there is AI making instrumental bas...
Can ai make anolog sound?
gaunt nebula
i don't know 😄
ivory patio
stiff crystal
I made a dataset for a voice model, but I can't train in voicemy.ai anymore. Can someone help me train it locally?
stiff crystal
<@1129179890218962965> hey bro what link to make ai cover?
I don't understand your question sorry.
rare gobletBOT
Ayo? @stiff crystal level 7 !!!
ivory patio
I don't understand your question sorry.
What link some guy say that old collab rvc 2 not work any more He told me to switch to a new link but I tried every link but it didn't work
What the new link collab?
stiff crystal
What link some guy say that old collab rvc 2 not work any more He told me to swi...
Sorry I don't know
ivory patio
You dont know the link?
stiff crystal
no
ivory patio
But how to use collab ?
stiff crystal
I can't answer questions rn sorry.
anyways
I made a dataset for a voice model, but I can't train in voicemy.ai anymore. Can someone help me train it locally? please reply to this message if so.
ivory patio
You make voice but how you use it?
Like what link to covert mp3 file to voice and like cover it
Hey bro how can i use collab what link?
rare gobletBOT
Ayo? @ivory patio level 3 !!!
tranquil cliff
Hey bro how can i use collab what link?
which colab?
coarse falcon
Same error 🥺
tranquil cliff
Hey bro how can i use collab what link?
to create model =https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS#scrollTo=ZodNcumpg-JM
coarse falcon
Please some one help
half cove
try applio
coarse falcon
How to use it
tranquil cliff
Hey bro how can i use collab what link?
to create covers = https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb#scrollTo=-7cQtXouqpQi
coarse falcon
I m new i don't know
rare gobletBOT
Ayo? @coarse falcon level 1 !!!
coarse falcon
Yes thx
half cove
try the applio one
spare shale
https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw
https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing
spark jetty
hey
nice to meet y'all
i keep getting a "restart session" error
i have no clue on how to fix it
half cove
https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw
https:...
recommend the applio one for now
others are disconnecting
half cove
i keep getting a "restart session" error
what colab
spark jetty
what colab
so-vits-svc 4.0 inference FIXED thanks goat YeezyBeaver.ipynb
spare shale
???
half cove
so-vits-svc 4.0 inference FIXED thanks goat YeezyBeaver.ipynb
rvc better
also thats super old
spark jetty
rvc better
do you have a link?
rare gobletBOT
Ayo? @spark jetty level 1 !!!
spark jetty
-colab
spark jetty
is it one of these? @half cove
half cove
guides are here
https://aihubdocs.github.io/en
rotund prawn
please use a newer colab
okeyy
turbid pagoda
I'm having problems downloading models off of applio. it will say model downloaded successfully but when I refresh nothing happens? Is anyone else having this problem?
coarse falcon
But it is not opening when i clicking on it
Can anyone please help 🥺
@half cove please help
When i am trying to upload my audio by clicking tap to upload it's giving me two options but that media is not opening
half cove
prob messing up cuz mobile or somethin
coarse falcon
Everything is fine
When i click on Mike for recording it's working
What can I do
half cove
if it works click convert
odd shale
Anyother Collab?
Colab for inference or training?
odd shale
To make ai cover
Use Ilaria RVC.
ivory patio
Why made ai cover so hard
odd shale
Why made ai cover so hard
It ain't that hard.
xd
The only issues you could run into would be isolating songs and getting the acapella
coarse falcon
Ok @odd shale
odd shale
Ok <@628603171735994379>
Please, don't tag staffers unnecessarily
I was already replying to you.
coarse falcon
Will it work for RVC V2 models too
coarse falcon
Use Ilaria RVC.
Can you please give me it's link
rare gobletBOT
Ayo? @coarse falcon level 2 !!!
coarse falcon
And guide too
Same error 🥺something is wrong with my device i think
Thank you for your help
molten pecan
And guide too
https://aihubdocs.github.io/en/ check the AI cover page to learn
Have fun!
molten pecan
Hola Leo
modern canopy
why does it say for me no gpu not connected
ocean socket
-colab
amber geode
do i need to download cuba?
stiff crystal
I made a dataset for a voice model, but I can't train in voicemy.ai anymore. Can someone help me train it locally?
woeful star
Im Using Applio And When I Try To Use It It Logs An ImportError ImportError: cannot import name 'emformer_rnnt_base' from 'torchaudio.models.rnnt' (C:\Users\Cassidy\Downloads\Compressed\Applio-3.0.8\Applio-3.0.8\env\lib\site-packages\torchaudio\models\rnnt.py)
And Doesnt Output Anything
-rvc
foggy folio
guys for bve do you select vocals or instrumental
uneven elm
rare gobletBOT
Ayo? @uneven elm level 1 !!!
unique lance
What should my settings be for having a lower male voice with tryna to use a medium/higher female voice? I'm getting breaks and crackles a lot
indigo garden
RuntimeError: CUDA error: an illegal instruction was encountered what is this error?
woeful star
Im Using Applio And When I Try To Use It It Logs An ImportError
ImportError: cannot import name 'emformer_rnnt_base' from 'torchaudio.models.rnnt' (C:\Users\Cassidy\Downloads\Compressed\Applio-3.0.8\Applio-3.0.8\env\lib\site-packages\torchaudio\models\rnnt.py)
And Doesnt Output Anything
terse siren
how 2 add index file to magio-rvc
terse siren
What should my settings be for having a lower male voice with tryna to use a med...
like, if theres backup vocals in the background of your seperation or?
molten pecan
What should my settings be for having a lower male voice with tryna to use a med...
for realtime voice changer ask in #🔍│help-w-okada
molten pecan
RuntimeError: CUDA error: an illegal instruction was encountered what is this er...
what's your GPU?
molten pecan
Im Using Applio And When I Try To Use It It Logs An ImportError
```ImportError:...
what step are you on?
and which tool?
you need to give context
molten pecan
how 2 add index file to magio-rvc
INDEX goes to logs folder
terse siren
INDEX goes to logs folder
ty
rare gobletBOT
Ayo? @terse siren level 1 !!!
molten pecan
yw
terse siren
like, when i open the folder it says mute do i put it in there or like
NVm
found it out
azure marshBOT
RVC Guides (How to Make AI Cover)
frozen ledge
Why did the new easyGUI update remove batch rendering?
Is there a notebook that does batch rendering at relatively the same speed?
molten pecan
Is there a notebook that does batch rendering at relatively the same speed?
the Applio Colab
https://aihubdocs.github.io/en/rvc/cloud/applio-colab/ here's a guide for it =3
low cliff
easyGUI sucks now
barren rampart
How can i make ai covers on mobile?
I used to make some on my pc but it broke and it's been a while
Oh yeah I'm stupid
Theres the guides in here
low cliff
easygui is not working anymore
spark jetty
do you guys record and put autotune on vocals, then isolate them and swap voices, or do it the other way around?
ruby kelp
low cliff
you guys know about easyGUI
brittle wing
easyGUI sucks now
Wait what happened to easyGUI?
rugged stirrup
should audio be mono or stereo for the dataset?
spare shale
you guys know about easyGUI
yep
Now easyGUI There is an error in checking pth and index
But for creating models it works well
molten pecan
But for creating models it works well
have you tried?
spare shale
(I mean Without Interface)
spare shale
have you tried?
Yes, is working
molten pecan
should audio be mono or stereo for the dataset?
both are fine. but I recommend converting to mono after cleaning it
rugged stirrup
oh ok
molten pecan
for a smaller file size. as RVC will mono it eitherway
rugged stirrup
is 44k sample rate for the audio fine or does it need be to converted to 32k to work with the 32k on RVC
molten pecan
is 44k sample rate for the audio fine or does it need be to converted to 32k to ...
I also get 44k so it's best to leave it like that and use a targe sample rate of 40k
rugged stirrup
I heard that 40k sample rate is not the best and that 32k is the best
brittle wing
easygui is not working anymore
Only on easyGUI make my own models because it's easier to make or train models there than any other colabs so why it's not working anymore? 😭
molten pecan
I heard that 40k sample rate is not the best and that 32k is the best
iirc that's if there's a lot of audio compression. you unsample it to 32k and use that target sample rate
but that's more advanced stuff
@red kayak knows more about it
molten pecan
Only on easyGUI make my own models because it's easier to make or train models t...
https://aihubdocs.github.io/en/rvc/cloud/rvc-disconnected/ check this one out =3
rugged stirrup
Someone said that 32k works with 90 percent datasets
molten pecan
Someone said that 32k works with 90 percent datasets
well yeah it technically does work for any dataset of any sample rate hehe
brittle wing
https://aihubdocs.github.io/en/rvc/cloud/rvc-disconnected/ check this one out =3
I already used that and it's not working to me when Infer the models that I made from there 😑🥱
molten pecan
I already used that and it's not working to me when Infer the models that I made...
huh 😭 it works perfectly fine for me & plenty of people. likely you didn't follow the right procedure
also you can't infer on it... only train
seems like you are a lil lost
rugged stirrup
well yeah it technically does work for any dataset of any sample rate hehe
hmm I just want the one that can sound as natural as possible 48k makes the voices sound robotic
40k is less and 32k is the least
molten pecan
hmm I just want the one that can sound as natural as possible 48k makes the voic...
you are supposed to input a target sample rate that matches your dataset's
if you input a higher value then yeah, it will be very artifacty
molten pecan
40k is less and 32k is the least
hm wait. when you determine the sample rate, do you check the spectrogram?
brittle wing
also you can't infer on it... only train
Yeah unfortunately I can't infer the models that I made from there and I can't download the pth files because the MB is too large and only on easyGUI I have a chance because the models that you made from there already saved on Gradio there or on Google Drive
rugged stirrup
if you input a higher value then yeah, it will be very artifacty
Yeah learned that yesterday
rugged stirrup
hm wait. when you determine the sample rate, do you check the spectrogram?
I don't
molten pecan
Yeah unfortunately I can't infer the models that I made from there and I can't d...
in RVC Disconnected you can simply drag the .PTH from Colab to your Google Drive folder
which takes seconds
molten pecan
I don't
here's the thing, the right procedure should be looking at the spectrogram and see at where it peaks
then multiply the highest number by two
brittle wing
in RVC Disconnected you can simply drag the .PTH from Colab to your Google Drive...
I know but when I downloaded the pth from Google Drive I still can't download and if your want save your storage you don't need to download that so I only have a chance with easyGUI and hopefully will fix that colab
Let's wait that easyGUI to be fix instead because I think that RVC Disconnected aren't good for us mobile users so that's probably why it's not working to me for infer the models that I made from there 😕
grave brook
ight
silent willow
Anyone else hainvg trouble downloading models in easyGUI?
brittle wing
My model is training too fast, something is wrong. That is not normal 🤡
I did everything right, that's weird.
Does anyone know what this "problem" could be?
low cliff
i'm having the same problem
spare shale
If your Dataset is small (2-10 Files that last 10 seconds) This is Normal
I think so
thin stump
anything similar to the melband reformer on uvr 5?
There's a new model called BS reformer 2 that you can use after MDX23? And yes you can still use melband reformer
tranquil trail
There's a new model called BS reformer 2 that you can use after MDX23? And yes y...
but i mean on uvr 5
besides using the website
mvsep
thin stump
Nope
thin stump
Use them both
tranquil trail
Use them both
well i was told vitlarge v2 was gonna take care of mostly the hi hat bleeds
usually not needing the reformer
but i see the reformer does better than the vitlarge itself
thin stump
I haven't tried yet. Just wait until you get model maker chat unlocked since I can't link the discussion
tranquil trail
I haven't tried yet. Just wait until you get model maker chat unlocked since I c...
yeah its bc appolio and some other inferencing stuff are fucked rn
and my datasets are clean but the ai is sounding dog shit
maybe its poor sample choices
thin stump
and my datasets are clean but the ai is sounding dog shit
I think you or some other person tried juice world. Start small with a simple dataeet
That's not a musician
tranquil trail
I think you or some other person tried juice world. Start small with a simple da...
i uploaded a gunna model
the reason it got rejected is bc of the hi hats where artifacting with my model
like the bleeds
it also learned from the hi hat bleeds lmao
which made my model sound ok but bad
thin stump
Nvm idk, as long as the models remove it
brittle wing
Guys, someone help me, I want to train Michael Jackson's voice on the Bad Tour in 1988, but I had some problems with the colab, does anyone know how I can train???
low cliff
im going to scream
thin stump
im going to scream
easygui hasn't been working since yesterday
woeful star
what step are you on?
I Gave A Good bit of context however I shall further elaborate I Am Using APPLIO It Is Installed And Open In The Browser When I Try To convert Audio To Inference It Outputs A Import Error Along With Errors Suggesting It Is Not Writing Properly To My Device | Below I Am Uploading an image showcasing said errors in the terminal | My CPU is An Intel i7 3700 with 8 processors, My GPU Is A Radeon RX550
Screenshot Of Directory
I Am On The Last Step Aka Using It
young halo
Does someone know how to fix this?
woeful star
Does someone know how to fix this?
This Is A Huggingface issue not a rvc issue go away 💀
young halo
This Is A Huggingface issue not a rvc issue go away 💀
Ok???
young halo
Not even the admins or mods have any problem with this, so why would you? Whatever
rugged stirrup
Whats it mean when your model goes up?
young halo
Whats it mean when your model goes up?
You should look at g/total
There you can see if your model is overtraining or not
lament eagle
is this overtrain
rugged stirrup
Oh I used epoch 90 it sounded decent
lament eagle
bruh
thin stump
bruh
collapse lol. I had that before and it didn't sound good because I damaged the audio when I was cleaning it
delicate kettle
Is there a site I can use to use these voice models
Kits.ai doesn’t let you download the audio after you convert
brittle wing
easygui hasn't been working since yesterday
I tried that colab but it's still working tho so what do you guys mean that easyGUI isn't working anymore? But it's about Gradio not the training the models right?
delicate kettle
Is there any site I can use on mobile
ivory patio
is this overtrain
Add more EQ bro cut the lower bass
And add more delay with phaser 🔥🔥
lament eagle
das funny dawd
astral sierra
have colab fixed yet 🥲?
azure marshBOT
RVC Guides (How to Make AI Cover)
chrome shadow
-colab
fading drum
so im training a voice, but it wont make a .index file when its finished, i did 180 epochs, with a bunch of (about 120) wav files for the dataset
red kayak
hmm I just want the one that can sound as natural as possible 48k makes the voic...
With 32k u just filter out those trashier frequencies (not all frequencies are Trash but most ppl rip from YouTube which does remove a lot of quality)
And u don't really hear a major difference between 32k and 40k
48k u can't even hear
Humans cannot hear above 44.1k
Even trained ears and audiophiles
subtle cedar
Can we tag model or credit the model makers/creators on promo? I mean we can tag or credit the model makers on promo when we promoting our AI Covers whether the models made by ours and someone else?
red kayak
it looks like it overtrain at somewhere 5k
Ur smoothing is too high, making hard to spot the global minimum
lament eagle
That was 0.9 iir
red kayak
Can we tag model or credit the model makers/creators on promo? I mean we can tag...
Anything in promo is fine just make sure it applies to rules
red kayak
That was 0.9 iir
Yeah too low, do 0.987 and ignore outliners om
subtle cedar
Anything in promo is fine just make sure it applies to rules
Oh thank you
lament eagle
is this overtrain
This is 0.987 and I always have ignore outliers turned on
red kayak
But for u to get the generator tota to mode collapse isn't a good sign
red kayak
is this overtrain
This is the overtraining yet but the model is a lost cause
Isn't*
Fug
I can't type
lament eagle
You made me jump
zenith swift
Guys, set up rvc normally, the programme changes the voice normally. In valorant also everything is fine, but in discord there is some extra noise when I speak
lament eagle
Yeah anyway I figured it out and find that 140 epoch is the best
red kayak
Yeah anyway I figured it out and find that 140 epoch is the best
Your best approach is to remake the dataset
lament eagle
I damn sure made it clean
red kayak
What did u use?
lament eagle
RX
zenith swift
In discord do I have to turn something off in the voice settings?
lament eagle
No room reverb, just little noise
red kayak
Which modules
lament eagle
then I truncated it
lament eagle
Which modules
Modules?
red kayak
Yeah
lament eagle
Uhh