any ideas?
#✨│ai-help
jaunty shale
potent sequoia
Hey, idk why but i uploaded the model google drive link and it supposingly uploaded succesfully but when i go to the inference tab the model doesnt appear even after refreshing multiple times, any idea why this may be happening?
brittle wing
How to infer models on local?
glad zealot
How to infer models on local?
you can just use your trainer to infer, there should be a tab called infer
just put the audio you want to covert
brittle wing
you can just use your trainer to infer, there should be a tab called infer
What's that "local" mean it's a inference website other than Gradio right?
hot lagoon
Wich is a good fork for infers ?
glad zealot
What's that "local" mean it's a inference website other than Gradio right?
local means its running on your computer instead of online, meaning it will using your gpu and cpu
will still use gradio
glad zealot
Wich is a good fork for infers ?
ngl they all sound the same to me
just use which you want
hot lagoon
ngl they all sound the same to me
Not going to lie og rvc is gota for kind of fast training.
steel forge
Yeah I'm switching to mainline for training
Applio has built in TTS conversion, if that's worth anything to you
But general census is applio is more bloated
And I can personally say it's buggy as fuck
hot lagoon
how buggy we talking?
potent sequoia
any idea why im not being able to upload models? they keep saying i dont have it set as "anyone with link" when i did set it
brittle wing
local means its running on your computer instead of online, meaning it will usin...
Pls screenrecord of the tutorial how to use that local inference
rare gobletBOT
Ayo? @potent sequoia level 2 !!! 
hot lagoon
glad zealot
Pls screenrecord of the tutorial how to use that local inference
same thing with online
just not using google colab
also i cant record.. i dont have a gpu :D
glad zealot
ye true
potent sequoia
hey RVC is unable to retrieve the model files from my google drive, it keeps saying i dont have it set to "anyone with link" even though i did set it, i even tried with models that have worked in the past and now they dont work anymore, any idea of why this might be happening?
azure marshBOT
RVC Guides (How to Make AI Cover)
jaunty shale
can anyone help me with this? I tried to move it to C drive and its still doing it
odd shale
I think you shouldn't haven't put the Applio on the D Drive
jaunty shale
I think you shouldn't haven't put the Applio on the D Drive
it didn't do that before
plus I tried putting it in C drive it still does it
odd shale
plus I tried putting it in C drive it still does it
That issue is weird.
jaunty shale
lemme try precompiled one
proper shale
hey RVC is unable to retrieve the model files from my google drive, it keeps say...
google drive links r broken, dont use em, instead upload to Huggingface
half cove
google drive links r broken, dont use em, instead upload to Huggingface
again?
i thought gdown fixed it
proper shale
always has been
trolley
literally just use Huggingface
it's less prone to stuff like this
half cove
brittle wing
also i cant record.. i dont have a gpu :D
Okay
finite galleon
google drive links r broken, dont use em, instead upload to Huggingface
I am legit using gdown rn
proper shale
I am legit using gdown rn
id just refrain from doing so, huggingface works always so yeah
half cove
we dont talk about the huggingface incident
proper shale
we dont talk about the huggingface incident
yeah
finite galleon
we dont talk about the huggingface incident
jaunty shale
lemme try precompiled one
still nothing, it still says FileNotFoundError: [Errno 2] No such file or directory:
odd shale
literally just use Huggingface
it's less prone to stuff like this
HF >>>>> Drive
steady ruin
How do I sort models by alphabetical order in the GUI?
hexed cave
I just somehow
browsed through my uvr5 files then found the uvr5.exe
then it opened
now its processing my uh separation
half cove
then it opened
weird
half cove
hexed cave
¯_(ツ)_/¯
brittle wing
Type shit
proper shale
how do I create an URL for a voice model?
upload to huggingface
check https://aihubdocs.github.io/en for more info
hexed cave
if it works it works
yoo poopmaster
half cove
yoo poopmaster
yea?
hexed cave
I tried to separate the vocals from my cupid song
I removed the instruemtnals
the reverb
but when I tried to get the main
uvr5 just got the background voclas as the main
half cove
it should output both if you left the settings the same
hexed cave
wdym?
hexed cave
odd shale
No. The BVE model is not made to be used like that.
hexed cave
uwa?
half cove
i think they did that
hexed cave
ye
odd shale
You must select first an isolation model like MDX23C or Voc FT
Then right click and go to advanced insert model architecture options
half cove
also i meant the output folder
hexed cave
odd shale
i think they did that
No he didn't.
hexed cave
I'm following this guide
odd shale
Poop.
half cove
No he didn't.
what
odd shale
what
He didn't select an vocal isolation model.
BVE is meant for lead/backing vocal isolation
hexed cave
got it
I mena
I did mdx23c first
then the reverb with the vocals I isolated
then I used the vocal isolated to remove the reverb. Then used the vocal reverb isolated file to remove the backing vocals
I'll just restart
odd shale
Yeah, like that
odd shale
then I just process with that?
Yes
odd shale
is there anything else I have to do after processing?
Nope.
half cove
Then right click and go to advanced *insert model architecture* options
weird
i dont really use uvr so ask leo if you have problems with it
hexed cave
ah
half cove
i do most things in rx10
hexed cave
ohhhh :0
hexed cave
LITERALLY
im extracting vocals from the song I'm extracting it from
my gpu is going fking crazy
sour mason
im trying to train and when I do feature extraction, I get this
I was expecting it to say all-feature-done but it quickly aborts. what am I doing wrong?
brittle wing
goof morning, i was having some problems with ilaria's Colab, does anyone know if it's back?
severe sand
What is wrong when the voice is weirdly supressed? Sometimes I get a clear sound and sometimes the voice just vanishes.
Nvm my noise gate was just too intensive
modest isle
is there a video tutorial to download the ai voice programm?
thorny cargo
theres like....this background static whenever i use the model I made
is that an issue with the dataset?
do I need to clean it better?
gray garden
If using weights, and the links produce "zip" in name will this create error or just need to extract somewhere specific?
kindred raft
hi, i did one convertion and when I try to do another it just shows me the one i did
split latch
working on my first Male model
let me know what do u think of the data set
I am making a hamza model
program used:
- Adobe Audition
- Batch Music Splitter
rare gobletBOT
Ayo? @split latch level 4 !!!
split latch
yeah I have 30 min of dataset these are just split using Batch Music Splitter
sour mason
getting this error when processing data
ArrayMemoryError: Unable to allocate 12.9 GiB for an array with shape (2400, 721506) and data type float64
rare gobletBOT
Vote Booster: Vote now for a 10% boost. https://arcane.bot/vote
gray garden
if you download a model and reupoload through a google drive account
supple ibex
can someone rate my ai vocals? I added bass and reverb
gray garden
you just have to use copy command under sharing from that google drive account? so illaria rvc can reference it?
molten pecan
can someone rate my ai vocals? I added bass and reverb
sounds great. only thing is, it sounds low quality
supple ibex
thats the thing I hate about the model
molten pecan
you did a good job with the reverb too =3
supple ibex
its soo hard to clean up cassete audios
supple ibex
he has some newer ones (2004,2005) but theyre potato mic so not much difference
hexed cave
wait casette?
I used uvr5 to do some separation for a cover song I used for separation. I need some opinions on what I should use for when I use an AI model to convert the vocals from
supple ibex
wait casette?
yep
hexed cave
wdym by casette-?
supple ibex
from the river to the sea
🍉 will be free <3
supple ibex
wdym by casette-?
cassete tapes
supple ibex
and some newer songs but with potato quality
hexed cave
interesting
hexed cave
yeh
the first one has background vocals
the last one has its lead vocals- but has the instrumentals in it
supple ibex
run the first one through BVE
molten pecan
the first one has background vocals
remember to remove reverb & backing vocals
you remember the usage guide, right?
the docs link
hexed cave
rare gobletBOT
Ayo? @hexed cave level 9 !!!
hexed cave
this?
molten pecan
https://docs.google.com/document/d/1wTJ_wutDqEtsA99BJOXDDGax25pPIDE84O5E2Rio5Qk/...
try using this one
it's more updated
molten pecan
my gpu went crazy last time
you can use UVR through the cloud though
hexed cave
true
molten pecan
won't use your computers resources
hexed cave
so what do I do?
run the extracted vocals
through the cloud
by making it remove the reverb n backing vocals?
ykw I'll do just that
proud elbow
MVSEP is the cloud way, besides the UVR/MDX colab
haughty schooner
my tensorboard graph looks a bit weird. goes from left to right, than back to middle, and again to right. consistently goes down, bin in a zigzig style? sorry, im from europe so its a bit hard to describe it in english
rare gobletBOT
Ayo? @haughty schooner level 2 !!!
haughty schooner
heres another one 😮
yes, i continued training after stopping. i use magio-rvc-web 23.7.0
*mangio
proud elbow
you resumed training with different batch size from before
haughty schooner
yes, it was reeeally slow and someone recommended other batchsize. now it goes way faster
2mins/epoch
800 epochs, with 2:25hrs dataset
proud elbow
just dont, or if so start over the training
azure marshBOT
RVC Guides (How to Make AI Cover)
haughty schooner
okay, thats good to know. you think it would be wiser to canceln that training and just start over?
sour mason
so after going through Remove reverb -> Extract main vocals -> Remove noise with UVR then removing silence & distortion with Audacity as per the guide, my dataset is still sounding not isolated enough. what else can I do to improve the dataset?
proud elbow
okay, thats good to know. you think it would be wiser to canceln that training a...
yeah, for different batch size and parameters, except if you want to go more epochs
haughty schooner
yeah, for different batch size and parameters, except if you want to go more epo...
thank you so much, i wouldve never find that out!
sour mason
I accidentally closed rvc and ran it again. Now the graph looks like this. Is this ok?
dawn nebula
I accidentally closed rvc and ran it again. Now the graph looks like this. Is th...
its fine
dawn nebula
so after going through Remove reverb -> Extract main vocals -> Remove noise wit...
wait which models did you use exactly
sour mason
is it starting over or picking up from where it left off?
dawn nebula
is it starting over or picking up from where it left off?
starting over
sour mason
ah damn do i need to find the old one and delete it
rare gobletBOT
Ayo? @sour mason level 2 !!!
sour mason
wait which models did you use exactly
lemme search this rq
sour mason
oh what I did was process it with uvr for each song one by one
and then combined all of them into one file with audacity
dawn nebula
oh what I did was process it with uvr for each song one by one
yes, ik uvr, but which models exactly
uvr is a software
there are different models to choose from
sour mason
mdx23c-instvoc HQ->uvr deecho-dereverb->uvr bve 4b sn 44100-1
used for inst separation, reverb removal, backup vocal removal
dawn nebula
i think for now
choose the best 20 minutes out of 90 minutes
then retrain
should be better
sour mason
ah man is that the only way
I want to know how to clean it better for the songs to be covered too
dawn nebula
yeah
dawn nebula
I want to know how to clean it better for the songs to be covered too
what else can you hear apart from the main vocal
in your dataset
sour mason
sometimes id hear the backup voice not removed
dawn nebula
hold up give me 5 mins
proud elbow
mdx23c-instvoc HQ->uvr deecho-dereverb->uvr bve 4b sn 44100-1
use the BVE in this way #✨│ai-help message
sour mason
use the BVE in this way https://discord.com/channels/1159260121998827560/1159290...
how do I get tht option menu to show?
proud elbow
click the wrench icon for settings
rare gobletBOT
Ayo? @proud elbow level 23 !!!
dawn nebula
sometimes id hear the backup voice not removed
do what desuwa said, plus i have an alternative
ensemble mode : vocal/instrumental, avg/avg, vr arc 5 hp karaoke uvr, mdx net karaoke, mdx net karaoke 2
sour mason
use the BVE in this way https://discord.com/channels/1159260121998827560/1159290...
dang this spit out 4 resulting files. how can I just get the main vocal as output?
sour mason
ensemble mode : vocal/instrumental, avg/avg, **vr arc 5 hp karaoke uvr**, **mdx ...
not sure what to do with this sorry
wait nvm
sour mason
ensemble mode : vocal/instrumental, avg/avg, **vr arc 5 hp karaoke uvr**, **mdx ...
so I run these 4 models instead of the 3 I did?
rare gobletBOT
Ayo? @sour mason level 3 !!!
dawn nebula
so I run these 4 models instead of the 3 I did?
theres only 3
sour mason
oops meant 3
dawn nebula
you run it in ensemble mode
with vocal/instrumental settings and avg/avg
after that you still need to limit dataset length
to 20 min
low igloo
Hiya! I'm very new to AI and ond't really know the technicalities of how RVC works. I'm trying to make little voice covers of things with voice models and I'm not sure what parameter settings are best to reduce choppiness/inaccurate pitching. I wasn't sure if ripped vocals that are produced affects this (like when there is tuning on the song vocals) or if the voice model needs to be trained to utilize the pitches head in music if that makes sense? Like if a voice wasn't trained on an E flat, it can't easily predict an E flat?
odd shale
can I ask why 20? I thought the ideal length was 30~90
Nope, i don't recommend using more than 30 mins.
You can use more than that but under your own risk
And if your dataset is clean enough
sour mason
wdym by risk? like OT?
odd shale
wdym by risk? like OT?
Not overtraining exactly
Instead, simply using an excessively big dataset won't improve your model's quality
Not even if your dataset isn't clean enough
I suggest using from 20-30 mins
sour mason
got it
if I cut parts of the song that's not clean and have a sudden break in one of the lines, would that affect the model in a bad way
odd shale
if I cut parts of the song that's not clean and have a sudden break in one of th...
Sudden break?
dawn nebula
if I cut parts of the song that's not clean and have a sudden break in one of th...
nope
sour mason
ensemble mode : vocal/instrumental, avg/avg, **vr arc 5 hp karaoke uvr**, **mdx ...
am I supposed to choose one of these or are they working on top of each other?
cuz I noticed each model is producing vocal files
stuck ingot
why when I try to make a request to post my model to the server, it doesn't work
(the application does not respond)
sand iris
@tawdry ruin mon dieu !!!
stuck ingot
<@1037338320960761998> mon dieu !!!
?
tawdry ruin
<@1037338320960761998> mon dieu !!!
sand iris
C'est lui qui gère QCBot
stuck ingot
ahh merci beaucoup
tawdry ruin
le bot est en ligne, donc il y a sûrement eu une erreur quelque part
je vais voir ça
dawn nebula
am I supposed to choose one of these or are they working on top of each other?
theres a total output
tawdry ruin
@stuck ingot j'ai rien de mon côté, donc c'est peut-être un bug de Discord. tu pourrais réessayer et voir si ça marche?
dawn nebula
you dont need to go into the subfolder
sour mason
you dont need to go into the subfolder
ohh I was looking at the subfolder I found the total output and yeah it's much better thanks!
sour mason
yeah for the model but eventually I want to use this song as a base for the cover
stuck ingot
<@295556186504298496> j'ai rien de mon côté, donc c'est peut-être un bug de Disc...
je vais re essayer
sour mason
ensemble mode : vocal/instrumental, avg/avg, **vr arc 5 hp karaoke uvr**, **mdx ...
btw is this a replacement of BVE or am I supposed to do this on the original song
dawn nebula
btw is this a replacement of BVE or am I supposed to do this on the original son...
replacement of bve
stuck ingot
j'ai pas bien compris ?
tawdry ruin
j'ai pas bien compris ?
quel est le lien que t'essaie de rentrer?
tawdry ruin
utilise celui-là à la place https://huggingface.co/muaota/RomanHorse/resolve/main/Romanhorse.zip?download=true
rare gobletBOT
Ayo? @wraith kayak level 6 !!!
stuck ingot
merci beaucoup @tawdry ruin
tawdry ruin
👍
lethal geode
whenever i close out of the collab webpage i have to install everything again. how do i stop this
stuck ingot
@tawdry ruin ça marche merci !!
tawdry ruin
de rien!
tawdry ruin
est après faut attendre pour avoir le rôle créateur ?
yep, un QC va checker ton modèle et s'il est approuvé, le bot te donnera le Model Maker role pour pouvoir poster dans #1175430844685484042
stuck ingot
yep, un QC va checker ton modèle et s'il est approuvé, le bot te donnera le Mode...
D'accord, ces prends combien de temps ? (desolé pour mes questions, est merci encore)
tawdry ruin
D'accord, ces prends combien de temps ? (desolé pour mes questions, est merci en...
normalement pas plus d'un jour
stuck ingot
normalement pas plus d'un jour
okay, merci beaucoup !
half cove
normalement pas plus d'un jour
it has instrumentals
stuck ingot
normalement pas plus d'un jour
je vais recevoir un message quand ça sera bon ?
tawdry ruin
je vais recevoir un message quand ça sera bon ?
oui
stuck ingot
oui
d'accord, merci beaucoup encore, ces mon premier model donc je savais pas trop comment ça marcher
tawdry ruin
it has instrumentals <:boohooh:1176674698629750975>
damn uhh
I'm letting the other QCs handle this
half cove
damn uhh
~~I'm letting the other QCs handle this~~
i dont think i can reject so ill wait for diablo
tawdry ruin
i dont think i can reject so ill wait for diablo
sounds good, spam ping him
half cove
bet
half cove
you could've said something useful at least
e
half cove
😈
tawdry ruin
fluid matrix
e
spare shale
е
half cove
e
hard pecan
whats the best program for rvc? rvcgui? applio? any other?
proper shale
whats the best program for rvc? rvcgui? applio? any other?
test and see tbh
hard pecan
also how long will training be with a 3060ti and r5 3600
proper shale
also how long will training be with a 3060ti and r5 3600
this varies a lot
proper shale
but is it viable?
yes definitely
odd shale
also how long will training be with a 3060ti and r5 3600
The training duration may vary
Because it will depend on your training settings, and PC specs
hard pecan
i listed my specs. what would be a range i could expect?
odd shale
i listed my specs. what would be a range i could expect?
Which is your dataset length?
hard pecan
for the person i have in thought it could be as big as i want. certainly no shortage of data
but lets say like 5-10 mins
heavy dune
hi i am very new to this and trying to use paul mccartneys voice and the tutorial video i found says in the comments google collab no longer works. where can i find an updated tutorial if anyone knows? thanks
oh it seems it does work but i have to download the model as opposed to paste in a google drive link like it says in the video? sorry if this is noob as hell
proper shale
i listed my specs. what would be a range i could expect?
i mean... probably like, over 3 hours definitely, this is for most models in average hardware tbh
proper shale
hi i am very new to this and trying to use paul mccartneys voice and the tutori...
https://aihubdocs.github.io/en has new colabs n stuff
odd shale
oh it seems it does work but i have to download the model as opposed to paste in...
Use huggingface links
zinc anchor
You should start over if you want to change pitch extraction algorithm
severe sand
Trying to use UVR through the webinterface, this seems to be an ffmpeg issue, anyone knows how to solve it?
molten pecan
Trying to use UVR through the webinterface, this seems to be an ffmpeg issue, an...
@viscid moss any idea?
viscid moss
severe sand
The path leads me to believe that it hates spaces in the path?
cuts right off in front of " files"
Though that should probably be a much more frequent issue
split latch
can I open more than one RVC instance? to use the model interface because my pc can handle that but I do not know if that will do something to the main instance
how do I do that?
thanks for letting me know
rare gobletBOT
Ayo? @split latch level 5 !!!
split latch
I guess I just have to test and see and hoooooly smokes that is a lot
I think it's the log folder
orchid sphinx
hi
rare gobletBOT
Ayo? @limpid rain level 1 !!!
split latch
is my model starting to overtrain?
orchid sphinx
I'm getting an error but can anyone help?
split latch
done
so is it looking good?
I am heading for 500 epochs and currently at 423
half cove
also try different model checkpoints
split latch
I mean I am currntly blown at how good this thing is, this is my second voice model and 300 epochs sounds so good already
hexed cave
you can use UVR through the cloud though
this is usually this error I get
molten pecan
this is usually this error I get
@viscid moss any fixes?
low shard
could you send me a link of the colab you are using?
half ember
How to use the plugin whit à AMD GPU ?
dawn nebula
How to use the plugin whit à AMD GPU ?
which plugin
hexed cave
That's normal
alrighty
half ember
which plugin
Ok what plugin i need to use ?
low shard
Ok what plugin i need to use ?
of what? what are you trying to do
half ember
I try to use the voice changer but to work he take my CPU and not my GPU cose it's AMD so i get res:20000ms
And a CPU use 100%
I take GPU0 here but nothing to do i wont wort whit it
low shard
did u download the directml version
also for the voice changer help its better if you ask in #🔍│help-w-okada
split latch
so for example this is the point where I liked the model is trained the best how do I know what file it is
yeah there are currently a save every 30
half ember
did u download the directml version
i download this version
half ember
also for the voice changer help its better if you ask in <#1159290161683767298>
Yep sry
low shard
i download this version
should be the right one, ask it in help wokada just to not confuse other people on the channel to use tho
half ember
should be the right one, ask it in help wokada just to not confuse other people ...
yep it's done
half ember
should be the right one, ask it in help wokada just to not confuse other people ...
SO you think it's just a bug or i can download a plugin ?.
low shard
so for example this is the point where I liked the model is trained the best how...
the auto saved file that is the most closest to that step, so as i seen in your pic, prolly the 's24120' one even if its not perfect
low shard
a bve in the cloud version?
of what?
low shard
SO you think it's just a bug or i can download a plugin ?.
sorry but i dont really know much of w-okada, you can check up this guide tho https://rentry.co/VoiceChangerGuide
hexed cave
with how much resources it has to use
hexed cave
sorry but i dont really know much of w-okada, you can check up this guide tho ht...
there is a w-okada help channel #🔍│help-w-okada
split latch
good to know because getting close to 500 currently at 484 epochs this is how it looks and I am contemplating stopping it
low shard
there is a w-okada help channel <#1159290161683767298>
yea i already told him that
molten pecan
a bve in the cloud version?
Yea
hexed cave
idk how training works but I'd say good progress n good work 👍
split latch
I guess..., I mean I still have the saves so it would not matter even if it overtrained
hexed cave
yea i already told him that
split latch
dang
molten pecan
Hehe that's normal. Check the troubleshooting chapter of the docs
brittle wing
i need help. I just installed, got the program up and running but really not sure what to do next
molten pecan
low shard
that is very weird, its prolly overtraining
molten pecan
I kept on going n I still get it
For how long has the colab been running?
hexed cave
completed at 4:26
idk
if u mean this when u said how long the colab has been running for
@molten pecan
low shard
look more about this https://aihubdocs.github.io/en/rvc-resources/epochs-overtraining--tensorboard/, its prolly model collapse or overtraining
low shard
maybe you trained the cleaned audio more than it needed, thats why it overtrained
dawn nebula
have u checked the tensorboard
brittle wing
anyone able to help me with installation?
hexed cave
someone explain to me why Im getting an error for Gradio
the cloud ver. of uvr
I kept on refreshing n retrying. Got the same error every time
hexed cave
this is what I mean
it keeps happening
even after refreshing or repeating my actions. I followed thr troubleshoot for the cloud version of UVR
low shard
mm pretty weird
low shard
where you got that colab btw
hexed cave
from the doc
here
since the local version costs too much gpu resources
I decided to use the colab
low shard
ohh that one lol the image was a bit pixelated on pc
hexed cave
u see
low shard
try to disconnect and delete runtime and to restart it from 0
low shard
i mean delete the runtime, refresh the colab and re install it yk
let me try it too, usually i use the local one
hexed cave
yeah
I didn't use the local
bc once I ran it with bve with specific settings
my gpu was going thru hell
im finding errors more than I am finding progress
viscid moss
OMG the BVE Model
low shard
OMG the BVE Model <:nails:1159569314848972891>
i tried all vr arch models
they seem to not work
i also tried the de echo normal
not working either
viscid moss
What lemme test rq
low shard
theres a problem only with vr arch models ig
low shard
ye
viscid moss
Testing...
viscid moss
It's probably just the BVE because that one has a somewhat different architecture and parameters
half cove
bve fucked tho
half cove
It's probably just the BVE because that one has a somewhat different architectur...
let me try more
low shard
also the de echo normal one doesnt work
half cove
also the de echo normal one doesnt work
low shard
then idk if it was me at that point 😭
viscid moss
It seems to be just the BVE
half cove
yea
viscid moss
so everything else works?
yep, just BVE models is broken
hexed cave
ah I see
viscid moss
I'll try to fix it
half cove
the bve curse
viscid moss
meanwhile u can use BVE model on MVSEP or UVR5 locally
brittle wing
-colab
hexed cave
meanwhile u can use BVE model on MVSEP or UVR5 locally
aight
I just want my vocals to be removed of the reverb and background vocal
can anyone see if it's good enough?
odd shale
can anyone see if it's good enough?
Yeah, it's good
hexed cave
Yeah, it's good
poggies
odd shale
The smooth is on 0.999
Use 0.987
odd shale
What do I look out for?
If the graph goes up then it's overtraining
brittle wing
If the graph goes up then it's overtraining
I'm aware about this, but how much high of the graph should I be concerned of?
odd shale
I'm aware about this, but how much high of the graph should I be concerned of?
If the graph goes up and doesn't stop going up
brittle wing
If the graph goes up and doesn't stop going up
Like this??
brittle wing
It'll probably stabilize
Should I let it do it's thing for like an hour?
odd shale
Should I let it do it's thing for like an hour?
If you want.
brittle wing
If you want.
Alright thanks! 👍
molten pecan
Should I let it do it's thing for like an hour?
yeah np. you can always use a previous checkpoint o resume training from where you left
let it go for 24 hours and you'll be fine
brittle wing
Hey, i need a little help. Is there any site that you can make your AI voice model for covers and that like stuff, just asking
molten pecan
Hey, i need a little help. Is there any site that you can make your AI voice mod...
https://aihubdocs.github.io/en/ here's a usage guide for newbies =3
brittle wing
Hey brother
molten pecan
hai
brittle wing
let it go for 24 hours and you'll be fine
Oh okay awesome!!
rare gobletBOT
Ayo? @brittle wing level 6 !!!
brittle wing
Applio AI covers do not work in Google Collab. What should I do?
thorny cargo
is there like, an example of a properly cleaned dataset I can use as a reference for cleaning my own datasets
odd shale
Hey, i need a little help. Is there any site that you can make your AI voice mod...
Use Ilaria RVC
odd shale
You don't really need that xd
All you need is know how to use MVSEP, RX10 and Audacity
brittle wing
Applio AI covers do not work in Google Collab. What should I do?
@odd shale
odd shale
Also having a good ear.
odd shale
<@628603171735994379>
Use Ilaria RVC colab.
thorny cargo
You don't really need that xd
I mean i've spent hours trying to clean up this dataset, i've used some audacity UVR5 and adobe premiere for editing the clips
rare gobletBOT
Ayo? @thorny cargo level 4 !!!
brittle wing
@odd shale How
odd shale
<@628603171735994379> How
Why are you pinging me on every reply?
brittle wing
Why are you pinging me on every reply?
Srry bro, how????
thorny cargo
Srry bro, how????
starts with patience
odd shale
Srry bro, how????
Calm down.
brittle wing
Hayda
thorny cargo
Srry bro, how????
sen 🇹🇷 ?
brittle wing
Evet
thorny cargo
there is a #1159595680659157132
brittle wing
I know
rare gobletBOT
Ayo? @brittle wing level 1 !!!
odd shale
there is a <#1159595680659157132>
You must learn principally how to delete noise
thorny cargo
You must learn principally how to delete noise
on the website i saw to use audacity for denoising
however
it doesn't do a complete job of it
odd shale
on the website i saw to use audacity for denoising
From where did you get those plugins?
thorny cargo
Pre installed afaik
I also tried this
with denoise
nothing has been effective
odd shale
with denoise
Yeah, that UVR denoiser won't do much.
odd shale
nothing has been effective
In X-Minus there are some denoise models that do the job properly
But it's paid
odd shale
<:joe_what:1169303301163077672>
Monthly sub
thorny cargo
Yeah, that UVR denoiser won't do much.
so is audacity's denoiser potentially a decent option since UVR's isn't effective
odd shale
so is audacity's denoiser potentially a decent option since UVR's isn't effectiv...
Ig so
BVE model in UVR, BVE V2 in X Minus or HP6 in UVR
brittle wing
Use Ilaria RVC
I used by in inferences dosen't show my model name
rare gobletBOT
Ayo? @brittle wing level 1 !!!
odd shale
I used by in inferences dosen't show my model name
You must put your HF link of your model you wanna download and use in the "Download Voice Models" tab
brittle wing
I can't from google drive?
Because i putted and they sayed your model is download, you can go back to inferences, but my voice model dosen't show
odd shale
I can't from google drive?
No.
odd shale
Because i putted and they sayed your model is download, you can go back to infer...
"Refresh" button
odd shale
No copyrighted samples plz.
brittle wing
You must put your HF link of your model you wanna download and use in the "Downl...
How can i made HF link of my model
brittle wing
"Refresh" button
I clicked but dosen't show my model
odd shale
How can i made HF link of my model
Any huggingface link of a model, you can find them on the #1175430844685484042 channel
odd shale
I clicked but dosen't show my model
Yeah.. I don't know how to solve that
brittle wing
What do I do if my model just randomly finished? How do I resume?
odd shale
What do I do if my model just randomly finished? How do I resume?
How so?
hexed cave
wait Leo, since I'm using a cover ver. of Cupid should I use the original instrumentals of Cupid or the cover instrumentals?
odd shale
@modern oxide Tootle needs help
brittle wing
How so?
Maybe because I put it on 500epoches and then I guess it just stopped.
How do I continue
modern oxide
?
odd shale
wait Leo, since I'm using a cover ver. of Cupid should I use the original instru...
Cover instrumental
hexed cave
alrighty
brittle wing
My graph just stopped like this. So it did 500 epoches then automatically finished. How do I continue another 500 epoches towards it?
I use Collab!
severe sand
Any other idea for removing a harmonizer? the BVE Model didnt manage :/
rare gobletBOT
Ayo? @severe sand level 2 !!!
odd shale
My graph just stopped like this. So it did 500 epoches then automatically finish...
Yeah.. Your model overtrained.
Go for the .pth file of your model that gets near the lowest low on the graph
brittle wing
Go for the .pth file of your model that gets near the lowest low on the graph
So 600 epoches?
odd shale
So 600 epoches?
NO.
odd shale
How do I determine
I mean, hop your mouse over the graph line, it will tell you the step count.
Put your mouse there
severe sand
Can you get these graphs with the local baseline RVC?
odd shale
Can you get these graphs with the local baseline RVC?
Yes
brittle wing
Any huggingface link of a model, you can find them on the <#1175430844685484042>...
How can i make my ai voice model on that site
odd shale
How can i make my ai voice model on that site
What site?
odd shale
I've not found them yet then, mind to point out where?
type "local" on the graph search bar
d/total, g/total and g/mel are the ones you should care for
brittle wing
What site?
odd shale
huggingface.co
No you can't
You need to use RVC Disconnected colab or install local RVC to make models
brittle wing
Search the model checkpoint that matches or got near that step count
the PTH?
severe sand
type "local" on the graph search bar
Sry not sure which search bar you mean. I'm talking about the WebUI
brittle wing
Your pths are on the "weights" folder
Got loads of randoms as I've done this a few times
odd shale
Got loads of randoms as I've done this a few times
Ig this is the sweet spot of your model
odd shale
Sry not sure which search bar you mean. I'm talking about the WebUI
You confused me bc i thought you were talking about tensorboard
brittle wing
Thanks!
severe sand
Yea sorry, some peeps call it baseline RVC, not sure what else to call it
brittle wing
Now I understand most of the concept lol
severe sand
Are you guys running tensorboard locally or through google colab?
brittle wing
Are you guys running tensorboard locally or through google colab?
Google colab. When I tried to do it locally I was told my GPU is crap so I had to use colab instead
odd shale
RVC Disconnected Colab
severe sand
I dont think anyone here is training without a GUI
soft stratus
I want to sing text to speech
soft stratus
You can't.
Why
soft stratus
Or you did mean AI Cover?
Yes
soft stratus
Is text to speech in AI Text to Speech
odd shale
Is text to speech in AI Text to Speech
I'm getting confused
tacit grail
can somebody help me please? No matter what "vocal type to deverb" i use in Ultimate Vocal Remover it still only gets this result when im trying to get ONLY the main vocals (just a sample for showing the result)
i have it set to "vocals only" and using UVR-BVE
severe sand
Can I tell UVR to not add a "1_" in front of file names?
azure marshBOT
RVC Guides (How to Make AI Cover)
odd shale
does this sound good?
I can hear artifacting.
odd shale
shown in the screenshot?
No, the quality will depend on your dataset quality
hexed cave
ah
I mean
im using a model made by someone
so idk where to find a better model than in here
wait by dataset u mean the extracted vocal?
ohhh the ai model
atomic lynx
How do i use the ai models
hexed cave
No, the quality will depend on your dataset quality
so should I train my own Hu Tao model? Cause I got the model from here
rare gobletBOT
Ayo? @hexed cave level 10 !!!
hexed cave
im confused
brittle wing
-colab
hexed cave
thin stump
Is it the model by voxel? It sounds good
hexed cave
does this sound good?
the one made by voxel is this one
thin stump
oh it's been a while since i played genshin
hexed cave
the yae miko model is made by dolyfin
brittle wing
If I was to make 10x 15 second clips of voice, is more clips of voices better for good voice results?
I obviously know about the overtraining etc dw
thin stump
the yae miko model is made by dolyfin
Someone made a 58 minute dataset of Hu tao 
and it's gone. I was going to say it's a huge amount of work to compile and clean
It was talked about in voxel's thread
hexed cave
Someone made a 58 minute dataset of Hu tao <:skullfacedistorted:1159654720735035...
fr
someone made a clean parody with that model
timid yoke
For example if I was training my voice what would the best words / sentences to say?
hexed cave
so i was expecting something good
brittle wing
Someone made a 58 minute dataset of Hu tao <:skullfacedistorted:1159654720735035...
that's because the voice actors like made us take them down 😭
hexed cave
that's because the voice actors like made us take them down 😭
DAMN U BRIANNA KNICKERBOCKERRR
thin stump
For example if I was training my voice what would the best words / sentences to ...
it needs to reach low and high pitches
Oh yeah I forgot how VA's lurk here
hexed cave
what
brittle wing
For example if I was training my voice what would the best words / sentences to ...
words and stuff don't matter tbh as long as all syllables are said. just have a good range.
if you want like, screaming and grit in it, add like a minute of you screaming stuff
15 minutes of data is good tbh
you don't need a lot
hexed cave
I dont wanna train-
brittle wing
SO far I have created 5, 15 second clips of clear voice examples. If I was to make more, would the overall result be better if I get some good training out of it?
brittle wing
yeah i had to deal with... fucking. diluc's va
hexed cave
I'd rather just use models made by people
brittle wing
ugh i HATE that mfer
timid yoke
words and stuff don't matter tbh as long as all syllables are said. just have a ...
Ok thank you
brittle wing
doki doki literature club
hexed cave
ugh i HATE that mfer
is he that much of a mf
hexed cave
doki doki literature club
when did sayori get a va?
brittle wing
is he that much of a mf
yeah he's weird
brittle wing
when did sayori get a va?
ddlc plus
thin stump
Cmon, it's Cdawg. He's that guy
thin stump
Oh
brittle wing
SO far I have created 5, 15 second clips of clear voice examples. If I was to ma...
..... you need 5 minutes minimum
hexed cave
I'll do a bit of japanese
brittle wing
iirc
severe sand
I dont really get the step 2a (sclicing) before the training. Some of the files end up pretty messed up, wouldnt the nice and manually prepared clips be better?
nono I mean in the rvc model training
brittle wing
Looking super good. Just look how low it's going lol
I think it's cause I made 15 second small clips & 15 examples
severe sand
how can I restart training from a previous point or from 0?
safe harbor
link to a RVC cover maker?
brittle wing
dude if its 2 pth files what do i do for index
brittle wing
oh cool
hexed cave
what do they mean by this?
brittle wing
what do they mean by this?
eq it
brittle wing
like this kinda
hexed cave
like this kinda
and did of that
brittle wing
this sounds better than 97% of ai covers so that's good
spiral sparrow
guys when i try different voices in VC sounds all weird My Viga Card Rtx 2060 any tips?
brittle wing
<@456226577798135808>
yesss
hexed cave
alrighty
tough crane
is this thing in the red circle normal when training?
glad zealot
Ye I see that a few times, it's fine
glad zealot
Doesn't read the dataset properly on the first step I presume
But it does still continue after it so that's fine
proud elbow
seems like it didn't log epoch/step 0 and the g value fluctuates as supposed, so it's normal
brittle wing
I already do that but how to convert/compress folders from Google Drive to zip?
ivory shore
does rvc allow for you to input your own vocals and convert them to different voices? also is RVC-GUI safe to download?
steel field
brittle wing
How to compress folders from Google Drive to zip because I can't download the pth file from Google Drive after I trained/created it from RVC Disconnected because it said my storage is already full so can anyone who know this pls help me? 😭
west agate
when using uvr5 for DeReverb and DeNoise what is a good window size and aggression setting to use?
strange schooner
hi https://ibb.co/Nrj0fLX which of those options i should select? looking into making a model myself
sour mason
how long does it usually take to train a voice model with ~25min of dataset?
cold plover
any help, im making a model locally and this is happening where the process data button is
tacit grail
How do i make the reverb sound like the character's reverb not the original
cold plover
any help, im making a model locally and this is happening where the process data...
here is the full error
cold plover
How do i make the reverb sound like the character's reverb not the original
after you get the ai stuff done just put it in fl stuido and put the presets on
sand iris
here is the full error
space in filename/folder name
cold plover
if you dont have fl stuido bandlab is a great to
cold plover
space in filename/folder name
fr thats it
sand iris
yeah 
brittle wing
hey does anyone know a good ilaria rvc alternative to make more realistic ai covers
tacit grail
after you get the ai stuff done just put it in fl stuido and put the presets on
presets?
rare gobletBOT
Ayo? @cold plover level 2 !!!
cold plover
if you look on youtube you will find the presets
brittle wing
are there any other rvcs that support huggingface
cold plover
space in filename/folder name
thank you btw
cold plover
are there any other rvcs that support huggingface
nuh uh
spring orchid
-colab
vague haven
hi
https://huggingface.co/apple-juice5237/elena_victorovna does anyone know how do i get a download link to this model? i made it, uploaded it on huggingface and now am having the stupidest problem ever
brittle wing
I can't download the pth file from Google Drive because the storage is already full so what's the alternative for this? (this pth file is from RVC Disconnected after I running the cell for saving pth files to Google Drive)
civic meteor
I can't download the pth file from Google Drive because the storage is already f...
Wrong pth file
It's not the D_ file
It should be the one with the name of the model and then . pth
brittle wing
It's not the D_ file
Oh okay but someone said I should download the pth files from logs
brittle wing
Not that one tho
That pth file is from logs tho
brittle wing
Or just logs
From logs
brittle wing
It's the one that says model-name.pth
I renamed the model but can't still download due to full storage, I still should download pth files from logs just to compress to zip along with the index to save it to Huggingface
royal linden
RUN_EASYGUI isnt working for me any help?
tacit grail
if you look on youtube you will find the presets
where is it? I cant find it
copper flax
does the ai thing still work the same as it did before
rare gobletBOT
Ayo? @copper flax level 1 !!!
dawn nebula
does the ai thing still work the same as it did before
yeah
copper flax
how do i do it i forgor @dawn nebula
copper flax
https://aihubdocs.github.io/en
How do i get music i have uvr i just need to find out how to install music @dawn nebula also srry if im annoying you
dawn nebula
How do i get music i have uvr i just need to find out how to install music <@690...
wdym by install music
copper flax
I forgot Honestly i think i got it thanks for the help
dawn nebula
all g
copper flax
all g
Acually I dont know how to import the model im pasting the link for the model on the thing but i cant get it to work how do i use the model do i paste link or something else @dawn nebula
copper flax
alr
main lichen
Ive been trying to download the ai vc but ive downloaded torch and python latest version but it still wouldnt work, I also have cuda. It keeps on showing me an error everytime i start it and it just shuts down instantly
sand iris
Ive been trying to download the ai vc but ive downloaded torch and python latest...
which "AI VC" ?
main lichen
Voice changer,
sand iris
Oh, you might want to take a look at this :
Also, pytorch and python are not required
brittle wing
I already do that but unfortunately the model is still doesn't work after compressed it to zip file and save it on Huggingface as a link I mean when I trying to infer that the model I created is still doesn't work😭💔 but I think we should download the index and pth files from logs folder from Colab right? or maybe because I wrote the title of the pth file,index file and the zip folder differently?
rare gobletBOT
Ayo? @brittle wing level 6 !!!
main lichen
Back, my vc is working but i cant get markipliers model
i downloaded, unzipped, and put it into the models folder
and restarted the app but it wouldnt show up
thin stump
Woah, this model is only a minute long too. It should work @brittle wing . Ignore the language
As long as there's no noise
brittle wing
@thin stump
I use this program to separate it
I want to put Korean songs on ai. Then I should teach them with Korean tutorials, right?
thin stump
You want to use foxjoy dereverb > uvr deecho > then the denoise
to clean up those audio files
brittle wing
You want to use foxjoy dereverb > uvr deecho > then the denoise
foxjoy dereverb > uvr decho > what is this?
rare gobletBOT
Ayo? @brittle wing level 2 !!!
brittle wing
I use Colab
thin stump
foxjoy dereverb > uvr decho > what is this?
I was talking about UVR. Your audio samples that you posted here aren't clean enough to use for colab yet
mdx23c-instvoc HQ->uvr deecho-dereverb->uvr bve 4b sn 44100-1 > denoise
That's the usual order that you should use
brittle wing
I was talking about UVR. Your audio samples that you posted here aren't clean en...
So you mean to change the UVR settings because the voice is not clean enough to make AI models?
thin stump
So you mean to change the UVR settings because the voice is not clean enough to ...
It needs to be raw vocals only. No noise or sound effects
Let me see if I got an example
brittle wing
What settings should I change in UVR to eliminate noise or sound effects? Or is there a program other than UVR?
I just saw something, why did you erase it?
thin stump
It's not very helpful for explaining. That's just an example of the results of someone's clean dataset. You can find plenty in #1175430844685484042
brittle wing
There's no voice that I want
thin stump
What settings should I change in UVR to eliminate noise or sound effects? Or is ...
mdx23c-instvoc HQ->uvr deecho-dereverb->uvr bve 4b sn 44100-1 > denoise is what I use
If you don't have those, you have download the models within UVR
brittle wing
There is no item called mdx23c-instvoc HQ
brittle wing
Where can I download it? I'm sorry that the translation is not smooth
Yes, what should I do here next?
thin stump
In Mdx, get MDX 23 downloaded
brittle wing
Is this it?
rare gobletBOT
Ayo? @thin stump level 12 !!!
thin stump
You want MDX23 from this list
thin stump
You want MDX23 from this list
From VR-Arch, you want VR-Arch 5_HP-Karaoke, VR-arc DeReverb, VR-arc DeEcho-DeReverb
brittle wing
Should I do it like this?
brittle wing
Do not download Inst 1 since you will not use it. You need the MDX-23 model
But there's no MDX-23 over there. What should I download...?
gritty tundra
How to fix this please help?
brittle wing
How to fix this please help?
Step 1, 2, 3. Start all of them
If the runtime breaks in the middle, an error can occur.
Keep the runtime running.
gritty tundra
=_=
thin stump
I can't help either sorry, I never used colab yet
gritty tundra
brittle wing
gritty tundra
Where do you think I will find 8 9 11
gritty tundra
It's a normal phrase
No, it's called a smiley, meet it, it's a smiley, it's an Englishman who doesn't know what a smiley is
thin stump
But there's no MDX-23 over there. What should I download...?
How is the translation on this website by the way? https://aihubdocs.github.io/en/vocal-isolation--datasets/vocal-isolation/#section-56
brittle wing
No, it's called a smiley, meet it, it's a smiley, it's an Englishman who doesn't...
The translator didn't understand what you were saying. The translator told me to smile
rare gobletBOT
Ayo? @brittle wing level 3 !!!
thin stump
because i'm about to log off
brittle wing
I can see the translation well, but there are only a few things I don't understand.
gritty tundra
The translator didn't understand what you were saying. The translator told me to...
No, it's not an emoticon, it's like an emoji, your brain is not very good
gritty tundra
Can you translate it into German if you are in Korean or let’s better translate it into Russian
brittle wing
@gritty tundra Always keep a connected check mark or disconnect. If it still doesn't work out, try every step of the way again. I'm sorry that all I can do is help
gritty tundra
I don't understand Korean
brittle wing
Are you German?
der Übersetzer hat nun die Übersetzung ins Deutsche übersetzt Fühlen Sie sich wohl, wenn Sie jetzt sprechen?
thin stump
I can see the translation well, but there are only a few things I don't understa...
You should be fine though. Think of what I said as a shopping list. I will post all the models you need again
gritty tundra
Are you German?
Nien it's me Russian-speaking Ukrainian CYKA BLEAT Чорт тебе побрав
thin stump
This page explains what they do. https://docs.aihub.wtf/vocal-cloning-guide/vocal-and-music-separation/uvr
thin stump
You can have someone translate it for you if you do not understand @brittle wing
brittle wing
<@876786741149827083>
Again, if it's disconnected or the steps go wrong, something can get twisted. Please keep that in mind
brittle wing
You can have someone translate it for you if you do not understand <@45622657779...
I request translation by translator
VR-ARCH:
VR-Arch 5_HP-Karaoke
VR-Arch 6_HP-Karaoke
VR-arc DeEcho-DeReverb
6_HP-Karaoke-UVR
MDX:
MDX23C-InstVoc HQ
MDX-net Reverb HQ
Can I download this?
gritty tundra
Again, if it's disconnected or the steps go wrong, something can get twisted. Pl...
https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS#scrollTo=g1X5oANker6l I'm doing an RVS cover on this site
brittle wing
What should I do here?
thin stump
Maybe I can share my screen?
thin stump
ok
Could you share those in private? You need to delete those two audios before you nuke the server
in my private messages
thin stump
Yes
brittle wing
@gritty tundra
Disconnections or steps are not smooth as follows
If the connection is disconnected, disconnect it again and proceed with the connection again, and when it is checked, proceed with the step
brittle wing
If the connection is disconnected, disconnect it again and proceed with the conn...
@gritty tundra
gritty tundra
yes
brittle wing
I tried to infer the models I trained/created from RVC v2 Disconnected and I downloaded it from Google Drive but unfortunately it's still doesn't work 😭💔 but I think after you saved your pth file and index file to Google Drive it doesn't work I mean the index and pth file from logs folder aren't work anymore
update: but we should the model public not private I mean after I public it on Huggingface it's already works but unfortunately it's got error when I infer it to Gradio or Cloudflare 😑
latent kettle
-colab
gritty tundra
If the connection is disconnected, disconnect it again and proceed with the conn...
Well, everything is fine with the Internet
brittle wing
Well, everything is fine with the Internet
It's just that. It was cut off using a free version..
gritty tundra
Not fair, but how to make RVS cover now
low shard
what colab is that
low shard
Not fair, but how to make RVS cover now
you mean rvc right
brittle wing
what colab is that
RVC V2 Easy GUI
rare gobletBOT
Ayo? @brittle wing level 4 !!!
gritty tundra
you mean rvc right
Yes yes yes
low shard
RVC V2 Easy GUI
could you send me the link?
low shard
Yes yes yes
to do that, you can check https://aihubdocs.github.io/en there are many ways online and locally
brittle wing
could you send me the link?
Are you talking about the RVC V2 Easy GUI?
low shard
Are you talking about the RVC V2 Easy GUI?
send me the link of the colab you are using
need to check if its the updated one
brittle wing
send me the link of the colab you are using
need to check if its the updated one.
gritty tundra
could you send me the link?
brittle wing
What is the solution to this?
low shard
What is the solution to this?
its an outdated colab
gritty tundra
could you send me the link?
Yey :З
low shard
https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS#scroll...
are you tryna inference or make a model?
brittle wing
@low shard Colab, please send me the latest link
gritty tundra
are you tryna inference or make a model?
I want to make a model and I don’t know what to do next?
low shard
@brittle wing to do inference you can check the ilaria rvc colab in the cloud inference part in https://aihubdocs.github.io/en
its updated
brittle wing
I don't know the language there. Please send me the colab link