#✨│ai-help
1 messages · Page 121 of 1
Yea, we can put the patch link on the docs.
I know already
Just use MVSEP then.
Leave that...
Make sure to select the Roformer model that got 11.24 of SDR
I want to get a computer to ensemble them
Is my model overtrained
?
Go and look for the .pth file that matches 50k+ steps.
Its gone
look on your "weights" folder
If it isn't there, you'll must restart training.
I had the 330 weights deleted the newest one and the program still says 500 epochs training finished
Show me your weights folder.
So you deleted it after someone else in a previous conversation told you thats the OT point or what?
....
Just stop
?
What?
You're asking me unnecessary questions
If you don't reply to him, we won't be able to help you further. (Also, please don't be rude towards others, that includes staffers)
😂
?
Is there a temporary folder where deleted files go to on colab?
I don't think so.
Then you have to retrain it. For the future keep in mind to not delete pth files that you believe are the OT point. In general, keep all the steps that are before an OT, anything else to save space if needed, you can delete
hey i'm new to this server, I need to know what I need to use to actually use the models?
For realtime purpose or for making AI covers?
ai covers
Whats your GPU name
So basically save latest checkpoint should be unticked
TICKED
You already had it enabled
why i have an ultra high ms
Whats your GPU
Nvidia 2060 super
does it say "GPU: Nvidia RTX 2060 super" on the right side, or does it say something like GPU (dml) with some buttons
gpu with (dm,l and some buttons
Thats the wrong version then
i use a chromebook so I don't know how to check that 😅 sorry I look like an idot here lol
which is the right version
i use 18a
No, you downloaded the AMD GPU version of the voice changer thats why your res is high since it cant detect your GPU
for any future questions about the voice changer ask in #🔍│help-w-okada
Open task manager, go to performances, check what it says under GPU
It doesn't look like ovetraining until I press the blue button below
It is not about if the graph looks wide or spread out, its about the lowest point in the graph
But on the right doesn't look like overtraining
It doesnt matter
It is not about if the graph looks wide or spread out, its about the lowest point in the graph
What is it
Yes, it is overtraining.
How with a 1 hour and 18 minutes dataset
It doesn't matter, every voice can overtrain at any point.
That's a problem
Very misleading
Nope, it's not misleading.
I think you're just not understanding properly what's overtraining.
But someone told me the more data the more epochs
That's wrong. You shouldn't guide yourself by things that random people tell you.
Intel UHD Graphics 600 is that what it is? 😅
Ayo? @unreal marsh level 1 !!! 
Is it gonna still sound decent even at 300 then
I don't know, you would have to test every .pth file of your model.
Its sadly not strong enough for the voice changer.
If your internet is good you can try the online colab version
Heres everything you need
Uh okay
I'll restart training then
You have to get rid of the fact that epoch number does not matter, its all about the steps. 55k steps is very high, it might be ready. 55k steps depending on the length of the dataset is a different epoch number everytime
I trained a 15 min dataset and I reached 40k steps at 1400 epochs which was the OT
So dont pay attention to epoch number
Uh that's something
So basically I reached a good number of steps
But overdone it
But wasn't steps an SVC thing
I guess it is a good number of steps yes
But you need to overdo it to make sure it is actually the OT, which you did - you went to 70k or something and it kept rising, so NOW you know that 55k steps was the ideal point
But the pth files were deleted
Dont know what that is
55k steps
I dont understand
Steps mattered in SVC
Steps are still a thing on RVC.
Which came before RVC
Yes but
I did a model with a similar dataset trained on crepe and it still sounded decent at 700 epochs.
The dataset was even smaller
That's the matter
steps number vary with dataset length as i said, so the number of epochs on the smaller dataset at 700 might be 300 on your current one
just dont pay attention to epochs at all only steps
But the other one sounded decent at 700 with a smaller dataset
And that might have been the same step number for all I know
I didn't even look at tendorboard back then
It still turned out fine
Ok
It's just weird
its not, throw the concept of epochs out your brain
Correct
So the steps are more important as you say
-guide
Full AI Voice Model Training Guide (Local)
Link: YouTube
Credits: Christopher Villanueva
Model training with Mainline RVC
Link: Rentry
credits: Raven (ravencutie21)
AICoverGen Colab Guide
Link: Google Docs
Credits: Eddy (Spanish Helper)
Create a model with RVC disconnected (colab)
Link: Google Docs
Credits: Angetyde
How To Make an AI Cover With Ilaria RVC
Link: Rentry
Credits: 👽 Julia (ailen2091)
No, epochs and steps are important equally
I just as a general question to figure out what is the most common way to do the back vocals do you A convert that vocals with rvc or B leave them as it is. I just want to know the most common one among them.
Ok
rmvpe_cpu is missing
I dont think it exists
oh ok
in the last versions it does exist lol
Ok then no clue
What colab is that
Ayo? @brittle wing level 11 !!! 
You're using for inference
how to solve this problem ?
??what are you trying to do
Training or inference
its wokada
It's not a colab?
move it to somewhere else like Downloads folder or Desktop
already unzipped
you can still move the folder
some ai voices i download dont work in rvc, i think its cos pf the file type? the ones that work have an index file and a pth files whereas the new ones i download have pth and a ckpt file and they dont work
rmvpe_gpu
Yes, but he asked for CPU.
Yea, there's no such thing as rmvpe cpu
Thats what I thought 😅
Is there any python library that can use these voices? I want to integrate it in my project?
Hey the Ilaria-RVC is not working for me, because of this error message
Ayo? @subtle holly level 2 !!! 
Ilaria closed down. Check the docs guide for any alternatives
Last update: Mar 10, 2024
pinned message for next time
can someone uvr for me? my pc is too bad to handle it
What are some alternatives for the Ilaria-RVC?
I can't find 'FIX' folder
RVC Disconnected
ilaria rvc dont work no more?
Is it on colab?
Sadly not
Yes open the guide i linked you, go to rvc -> cloud
what do i inference
ilaria dont work
how do i can find the song output even i can't find 'FIX' folder?
ilaria has been fixed
all colabs?
yes
ty
how does the ilaria rvc beta 2 work, like where do i upload the huggingface link
is 1650 super good enough or should i use my ryzen 9 3900x
the beta is still being worked on and put private again for a serious training issue
i suggest you using rvc disconnected for training
llaria
Ayo? @naive veldt level 1 !!! 

is my 1650 super good enough or should i use my ryzen 9 3900x
guys i can't find the Fix folder how can i do?
CPUs generally are not suitable so yea gpu is decent enough
what settings yall suggest ?
⠀
Settings for Nvidia GPUs 
F0 Det.: rmvpe (suggested for all series)
RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra
Advanced Settings
Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low
⠀
@brittle wing
thank you
Ayo? @brittle wing level 2 !!! 
if its lagging you have to raise the chunk
but you have a 2070 it should run 128 good
i know iam just saying that the embed is wrong
make sure to reduce ingame settings
i aint using it in game
Its not really wrong, maybe something else is messing up
can can u explain what does the extra option do ?
tried on freshly installed windows
CPU usage. If its too high it will lat
lag*
keep it to 16384
50%
and it affects quality of voice
and i have 1 more question its not about voice changer but with ai
what settings is good for gtx 1060??
GTX 10-series: 128-192 chunk | +8192 extra
should i use crepe?
best settings for Radeon RX 5500 XT ?
rmvpe
1sec iam trying to find the video
whats the best setting for Radeon RX 5500 XT GPU
⠀
Settings for AMD GPUs 
Don't forget that your models needs to be converted in ONNX!
F0 Det.: rmvpe_onnx (suggested for all series)
7xxx XT cards: 112-128 chunk | +16384 extra
6xxx XT cards: 128-192 chunk | +16384 extra
5xxx XT cards: 192-256 chunk | +8192 extra
RX 580: 192-256 chunk | +8192 extra
RX 570: 192-256 chunk | +8192 extra
RX 560: 256-384 chunk | +8192 extra
Advanced Settings
Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low
⠀
how can these guys make like these photos with ai ?
I am not educated on stable diffusion and all of that yet. Check the pinned messages in #🔍│help-ai-art and else ask in that channel
ok thanks
why voice changer not working
Ayo? @honest ruin level 1 !!! 
😭
Specific
how i remove that it picks up random noise
i picks up the slightest air movment in my room xd
enable sup2 and move s.threshold further to the right
I have one AI voice model (vocals of a singer) that was created (with some help) a few weeks ago. I have successfully made AI covers of more than a dozen songs mostly with no issues. I get the vocal model to do a cover of a song from a different album where the vocals are very different. For some reason my model still tries to copy the type of vocals from the song Im trying to get it to cover. How do you get it to do it's own style without fully mimicking the OG vocals?
what can i do about that when i talk and it plays back with the voice changer that it cuts randomly out
Is it only inside the app or on discord?
to test in the app set monitor to your headphones to hear yourself
nah bro im not living alone xd my monitors loud af
?
try
open task manager hit details
right click audiodg.exe and set Priority to HIGH
right click it again and press set affinity and only have an EVEN number CPU core setlike 2 AKA everything unchecked except CPU 2
what does affinity mean in german 💀
affinität
but theres nothing called like that
with row is it
for you
?
zugehörigkeit
ah bist du deutsch?
where do i change those advanced settings ?
on the bottom theres a small button
advanced settings works for every nvidia series ?
wokada
input: mikrofon
output: virtual cable
discord:
input: virtual cable
output: headphones
What's better RVMPE or mangio-crepe?
Shad is a chad for keeping the help channel alive.
rmvpe
i frogot what mangio crepe is good for but its good for a specific areas
Ah okay send me samples how both pitch extraction algorithms sound
But I don't like rvmpe
For training a model
@pastel oak was würdest du empfehlen?
30 minuten datenset oder?
all good, lass dir zeit
Ja ich glaub die default settings reichen so schon aus
hab überlegt ob man batch size erweitern soll aber 1:30h da muss man es noch nicht glaube ich
Nur eine sache
save_frequency meinen die soll man auf 15 machen wenn man newbie ist
total epochen würde ich eine sehr große nummer benutzen z.b. 2000 weil wir nicht auf die epochen achten werden sondern auf das tensorboard das sich auch nebenbei öffnen wird
Lies dir den bereich zu tensorboard auch nochmal durch
@brittle wing
I need to correct a small mistake I made when you asked about ticking save only latest ckp
I got it mixed up with another one.
You can leave that enabled. Just make sure that save small final model is activated, that is the thing that saves every model at X epoch/steps
Ich benutze persönlich 10 für die genauigkeit aber ist dir überlassen. werden dann halt mehr files
im grunde macht es wenig unterschied tbh
ok, dann ändere ich also nur die epochen auf 2k 🙂
I want it that way to have the weights in my drive
If you want the proper details
If you disable it, it will create a G and D file for every X epoch that you set. This will also mean that every X epochs its saving files that are probably larger than 1 GB in total. Since colab runs on your drive, it will overfill very fast, but you would have every file to resume training ; or for the case of losing pth files - you can quickly run the closest to the OT, have it create the final model and turn it off again
Yeah
If you really want that, make sure to constantly delete the first few files in realtime to avoid accidentally overfilling
@pastel oak
Wenn du das training startest stell sicher dass tensorboard sich auch mit öffnt. falls nicht musst du glaub ne localhost adresse eingeben die da angezeigt wird in nem neuen browser tab
öffnet sich, wenn ich jetzt lange afk bin. Wäre das schlimm?
von RVC Disconnected afk sein ist schlecht
von tensorboard nicht
heißt ich sollte alle 10 minuten mal die maus dort moven? und ist es richtig oder falsch das bei tensorboard noch nichts passiert
Weiß nicht obs alle 10 minuten sind, wenns da iwo stand dann passts
tensorboard ist normal ja
perfekt, danke dir für die große Hilfe, ein Tipp: Eventuell Deutsche Kategorie zu #1175430844685484042 hinzufügen 🙂 Würde sicherlich noch großer Community ansprechen.
RVC Guides (How to Make AI Cover)
Translation by country
is it possible to rip the dataset from a pth file
dug through my files and found an old oc model of mine and i can't find the dataset anywhere
yea no, don't think so
If its still the same rvc go into the logs folder then to 0_gt_wavs
its not pretty but it should be in there chopped up right?
ya the logs folder is the only chance i see if its the same as local
else no chance afaik
my gpu just rebooted itself holy shit
Ayo? @hoary eagle level 2 !!! 
EOFError: Compressed file ended before the end-of-stream marker was reached
on google colab
i don't think it liked me watching youtube while having an audio rendering
download failed for some reason, be it your internet or wherever downloading from iirc
when making an ai voice, aproximatly how long would 250 epochs take
i have a 307ti
if that helps
Depends on how long the dataset is, the batch size you choose etc. but mainly dataset
thats a relatively short dataset
maybe more
compared to what others use xd
to give an example, on my 3060 ti on a 14min dataset and 8 batch size one epoch took around 15-20 seconds
brother what
for 27 minutes dataset it takes 30-40 seconds
Idk maybe you selected a smaller batch size
8 batch size
how long does one epoch take what does the cmd say
im new to this
Im going to introduce you to a whole new world
Last update: Mar 8, 2024
should i just cancel this
so i just follow this ?
could u luink me that pls
guessing i just dl this
yes
Ayo? @wooden pasture level 2 !!! 
whats up
thats for inference
aka. you want to make a ai cover for example
skip that entire section to training
mangio
which step is that
and the steps before with processing the dataset worked?
im so smart
its inside a zip file
"ariesdata1.zip"
it can not read files that are inside a zip file, so you have to extract it and put it somewhere else
try pre process again tho im not sure if you have to restart anything
Think you can just select the path of dataset again and click on process data
fuck me this looks like its still gonna take ages
I dont suppose theres a model of Colab RVC with a text to speech option?
whats Ilaria RVC Mainline?
Ilaria RVC is.. RVC as colab hosted by Ilaria
Mainline is RVC locally
You can look into here
https://docs.aihub.wtf/tts/tts-tools/
is it like
shad can i dm u rq
Also the Applio Colab has TTS
I guess
I dont know what youre talking about im sorry lol
its nothing about tunes or not its just a colab space that ilaria created
it serves the same purpose as the others in creating AI covers
⠀
HuggingFace Spaces 🤗
⠀
Ilaria RVC
EasyGUI port with some improvements, by Ilaria.
RVC-HFv2
Applio port, by r3gm.
AICoverGen
AICoverGen port, by r3gm.
Advanced RVC Inference
Extended version of the GUI with advanced settings, r3gm.
⠀
⠀
Settings for Nvidia GPUs 
F0 Det.: rmvpe (suggested for all series)
RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra
Advanced Settings
Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low
⠀
⠀
Download for Nvidia GPUs 
Version 18a cuda
Download for AMD GPUs 
Version 18a directml
Download for Intel GPUs 
Version 18a directml
Download for Mac 
Version 17b Mac
⠀
⠀
Settings for Nvidia GPUs 
F0 Det.: rmvpe (suggested for all series)
RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra
Advanced Settings
Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low
⠀
There’s got to be a easier way…
mic sensitivity slider
rvmpe is for normal, nvidia gpus
rvmpe_onnx is for the amd ones
is there rvc gui link
if ur talking about rvc-gui specifically, no because it's outdated
https://docs.aihub.wtf/ has everything
Last update: Mar 10, 2024
-colab
- Applio, by IA Hispano Google Colab
- AICoverGen-WebUI, modded by Hina Google Colab
- AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
- AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
- Ilaria RVC, by thestingerx Google Colab
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Applio, by IA Hispano Huggingface Spaces
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HFv2, by r3gm Huggingface Spaces
- AICoverGen, by r3gm Huggingface Spaces
- Advanced RVC Inference, by r3gm Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
Ayo? @sinful nexus level 4 !!! 
ooh but someones said rvc-gui better but i dont think so someones wants to use rvc-gui
nah it isn't
it might be easier, but it isn't better
is there a way to run an RVC gui on one computer and use it on another? like to run the UI on your local network and use it on a another PC?
i know
yeah i think so, just open the localhost link?
Ayo? @proper shale level 95 !!! 
i think localhost only works on the machine it's running on
i think ive seen a way to forward it to your home network by setting "share=true" in launch()
i'm training a model rn so I'm gonna probs try it after it's done training
oh yeah true
i believe that's in a webui python file... 
im not entirely sure, never messed w that type of stuff
Heyo, I'm trying to do step two in the Easy GUI for RVC colab page but it's not downloading the model. I've used a link in the google doc
I'm trying to use junkratv2, (line 3344 in the drive)
Still waiting for some help on this if someone can
outdated colab, check https://docs.aihub.wtf/
Last update: Mar 10, 2024
You can try increasing index to help with the accent... but it'll still sound similar to what you give it
Ig it's about giving the AI what you want to hear
So I'll have to continue training it? As for sounding similar, I'm trying to get the vocal model I created to do it's OWN style of vocals for a song that has a completely different type of vocals. I just want to clarify that.
Real quick, I clicked on one of the model links in the drive and I downloaded a zip folder, is there a way to upload it manually instead of directly from the site?
It's the same vocalist but his vocals changed between the 3rd album (song from the album im trying to cover) and the 7th album (the ai model i created)
oh 
you'd probably need a diff model then..
What do you mean?
im making reddit stories
normally it might do that unless the voices are similar 
ngl training more might be a nice idea, but idk, i wouldn't bet on it
this isn't possible with RVC
uhm actually 🤓
It will always sound at least somewhat similar to what you give it
nvm ur right
The only way to get the same song in the voice AND style would be to have a cover of the song in that style
It's not possible with RVC? What is it possible with? I know it can be done because there is YouTubers who are doing the same thing Im doing and they are album to have the vocals from a different album do the ai cover as intended
nice grammar i got there
anyways
It's all about what you give it, training and inferencing wise... so yeah, maybe it didn't alter a lot because the voices are similar enough
No they aren't similar at all. Some covers Ive seen on YouTube use vocals from the 2nd album (COMPLETELY different) to cover a song from the 3rd album and this vocalist's style changes night and day. No exaggeration!
And yes it's the same vocalist
Maybe I should be specific and you can go see for yourself. The vocalist is Corey Taylor from Slipknot. His vocals are brutal (good way if you listen to metal) and he ruined his vocal chords between that album and the 3rd where he had to completely change his style. The AI YouTubers successfully had a 2nd album vocals cover a 3rd album song and it still had the 2nd album style despite it being different. That's why I need help because I know it's possible.

it really depends, there's a lot of variables... it can depend on the pitch too
seems like a complex issue
Hmm pitch
would changing the pitch help with that? or is it more complicated than that like you said?
u can try changing the pitch yeah
short question, i cant train my voice anymore cause i ran out of points or smt... Can i download my model or where is it now?
Ayo? @sly tulip level 3 !!! 
what are you using, and do you have colab pro 
Can I load my model into Applio (RVC I use) and change the pitch, or do I have to start all over again with a new model?
rvc v2 disconnected and no
you can change the pitch yea
if you didn't export the model before it disconnected... you lost it

By just loading the model?
i still could, iam still at the page but where?
yeah, and using the transpose (or anything related to pitch and that has a number on it) option
iam here, but cant train anymore
see, here's the problem
your session got disconnected, you lost ur gpu time, and you cant reconnect
@proper shale Okay, thanks.
yw, let me know how it goes ;)
so... i cant try it anymore right? its RIP
yeah... you lost that model 
you'd have to wait 24 hours for the gpu limit to come back and you'd have to start over...
do i have to start from 0 then?
yeah... sadly
but next time, you know
stop the training step after like an hour or two (make sure its not saving the d/g files) and then export model to keep the stuff safe
how do i use the ones with govits
you can't, if you're talking about using those models with w-okada
Im trying to convert my audio and it keeps giving me an error (and yes im using the updated colab)
that's an outdated colab, check https://docs.aihub.wtf
Last update: Mar 10, 2024
has newer colabs and guides
check the logs in the colab cell's output
Which google colab notebook are you using? The link
RVC Guides (How to Make AI Cover)
Translation by country
easy gui v2, pretty sure
The latest?
Fck welp that's why XD
so what are they for?
they're for gpt-sovits
what is that?
tts, can voice clone only with reference audio
u can load models for more accuracy, that one is a model
Oh ok, thanks for they help! sorry for my ignorance
it's okay to not know :)
i also am a bit confused bout gpt-sovits, dw
Why is every non-Google Collab RVC taking forever?
if its on huggingface, its using cpu
so yeah
hey
toggle voice on and off in w-okada /
voice-changer
i want to use the fake voice on discord for DND game
but if i turn the voice changer off no one can hear me
how do i toggle the voice between my own and the realtime voice?
i tired voicemeter bananna but the sound gets cruchy and i cant figure out nosie suppression
wrong channel buddy
hey mj are u there
i trained 495 epoch data
now just put the audio path there and click on convert?
have u read https://docs.aihub.wtf/
Last update: Mar 10, 2024
hey, using applio seems be giving me strange results where it wasn't acting that way for me a few months ago. First the inference processing is taking less than ten seconds, where it would previously be about a minute, and then the result is extremely faint/quiet with the volume randomly going high and low
idk if i'm doing something wrong, the only change i make to the default settings is check "split audio"
like this, this exact same model was fine before but now it's garbage results even when i completely reload the code
Anyone know what the fcpe pitch algorithm is?
Why is my voice making like robotic noises
lol
does anyone know what voice duckus uses
@proper shale hey I’m sorry to ping but could u help me with colab?
it seems like all the audio i generate sound like they have throat cancer, like not human at all
not used it much but kind of bootleg rmvpe which is faster and less resource intensive, also fine for voice changer as well (only availabe for gpu-cuda one)
- Applio, by IA Hispano Google Colab
- AICoverGen-WebUI, modded by Hina Google Colab
- AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
- AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
- Ilaria RVC, by thestingerx Google Colab
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Applio, by IA Hispano Huggingface Spaces
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HFv2, by r3gm Huggingface Spaces
- AICoverGen, by r3gm Huggingface Spaces
- Advanced RVC Inference, by r3gm Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
i have a ? everyone its because i want to join tiktok live as a guest with my voice changer but i want connect my pc mic to my iphone how can i do that because i cant use voice changer on phone
is there anything that can help my model being rapsy
I have a question, when making an Ai voice model, do ya’ll recommend to implement the character screaming along with its dialogue or nah?
Do i need a index file for ai covers, or is just the pth just fine?
RVC Guides (How to Make AI Cover)
Translation by country
Does anyone know how to make an ai model?
oh my gosh this is so much
Ayo? @inland badger level 1 !!! 
Hello alll
quick quetion, only Beatrice works, all
RVC models do not pick up on input
do you know a way I can configure it to have RVC models work?
Hey
Ayo? @sturdy laurel level 1 !!! 
You must be very good at music 🎵
huggingface isnt working
how much epochs is enough for 1 minute dataset
how do i get the voice changer to work in game
Guys, wouldn't it be great if we could add some Reverb to W-Okada? 🧐
what to do if the program is buggy in cs go
i need help about manglio rvc fork can someone help me.
how to fix echo guys?
Wait until it overtrains
enable "echo" box
I mean on my whole mic
like my mic is ass
it puts out game input or youtube videos
You can install steelseries sonar and watch a yt video on optimizing it
ok thx
Ayo? @tidal junco level 1 !!! 
can i get help from someone? i managed to use the ai model to change the voice of a wav file but i can't figure out how to download it
nevermind!!
how to fix random noice? when i am talking it repeat it selfs or it just makes random noice (i am using Voice Changer Client Demo)
enable sup2 and echo
yeah i did but it still does it sometimes
Ayo? @ashen thunder level 1 !!! 
move s.threshold to the right
dont move your mic close to your headphones
ok thanks it works now
RVC Guides (How to Make AI Cover)
Translation by country
mess with the pitch options
what's the issue?
Helper needing some help 🙏
Whenever I prepare a dataset I would just look up the sample rate on Ilarias Audio Analyzer, copy the sample rate written there and call it a day when training. Now I am using Adobe Audition CS6 to check for the spectogram kHz since the sample rate number is wrong in 99% cases 💀
If my spectogram peaks at 18kHz (*2 = 36khz), should I always go for 32khz since upscaling it to 40 would just leave gaps in frequencies?
I always trained on 40khz and got decent/good results. Looking back at my older datasets they also all peaked at 18 which more so made me question if my sources of getting audio is really that bad it never even gets up to 22khz lol
Should just use the choice that you feel safe
But based on my research, the audio is just downscaled to 16k anyway
@proper shale you mj.. if i want to train in the old style on rvc disconnected do i just need to download "original" pretrained.. i didnt train in a while
hi guys, the Ilaria RVC doesn't work, can i get some help
Just select the original
i'm not really sure, changing it +/-12 still sounds robotic and also way worse, and i haven't changed it to any other value bc i assume that would make it off key
Okay so I used the rvc discontinued colab link. I trained the model and followed the guide however, when I saved the model it only saved it as a folder apparently.
I've just tested on Colab, it works
i have an m2 mac
how do i get an RVC running
preferabbly for vocals
i have Whisky so if i need to port i can but so far no luck
Signed 16 bit PCM in extraction right?
i really lose my touch
yes
-colab
- Applio, by IA Hispano Google Colab
- AICoverGen-WebUI, modded by Hina Google Colab
- AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
- AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
- Ilaria RVC, by thestingerx Google Colab
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Applio, by IA Hispano Huggingface Spaces
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HFv2, by r3gm Huggingface Spaces
- AICoverGen, by r3gm Huggingface Spaces
- Advanced RVC Inference, by r3gm Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
anyone?
training voice models / ai covers / voice changer
what do you want to do?
ai covers
my friend had an RVC software ages ago and hes on windows
that was before GPTK got any good so i couldnt tag in
you need the mac release of rvc, its probably on github under the retrieval based voice conversion part. im on phone dont have the link atm
damn
if you could find it would be v appreciated cheers
where is the tutorial i can fix this? i have amd.
i camt find a binary release so you have to run it manually
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/en/README.en.md
installation steps
AMD is not an NVIDIA GPU...
oooh really? what a suprise
It says os is not defined
For me on Setup 2
Not sure what that means
I'm windows 10 rtx 4090 i9 13th gen
Sorry not tryna bug you but do you know my issue?
Ayo? @brittle scroll level 1 !!! 
I can send you a dollar if you need for the inconvenience
im also new here but my bro (@wispy burrow) teaching me what amd is 
Lol
I can't use the web version because of this error
What RVC are you using?
Basically that one on the website
Google drive
Yes after setup 1 worked the rest just don't saying is not defined
I still don’t know after all this time: which algorithm is generally best for singing models?
where do you put the model and index files?
The model just goes into the weights folder and the index files goes in logs folder I think?
OMG I FIGURED IT OUT
is there any sub folder?
im gonna try with cpu
if its working with cpu
how long does it normally take for someone to respond to your help form? Just asking
It puts a folder in there automatically after run the processes
yea but i dont put inside sub folder right?
just inside logs folder?
The index file should go in there automatically once you’ve finished training and you hit the ‘train feature index’ button
no i have a ready model and just want to learn rvc
so i have index and pth file now
You mean, you’ve finished training and you want to inference your new model?
Create a new folder with the Modelname inside the logs folder and place the INDEX there..
something like that yes. i bought one and wanna test on rvc
Ok so drop your index file in ‘logs’ folder. It doesn’t need to go into a sub folder
And then drop your .pth model in ‘weights’ folder
And when you refresh your inference data it should appear
yep it did work
thank you
does aihub haves something like explains all the settings what they does?
No problem
If you find out let me know. I still don’t know which algorithm is best
you are not satisfied from your exports?
Ayo? @trail trout level 3 !!! 
Sure, but I don’t know which algorithm is the ultimate for singing
so the settings on talking vs singing changes?
thank you
on youtube everybody is saying that creepe is best
Ah, possibly. Maybe I’ll try it
RMVPE is considered the best overall.
Harvest is outdated.
Thanks! I thought so. I just wasn’t 100 percent sure if that was the best for singing too
is this for singing or also for podcast so?
Works for speaking and singing.
thanks im gonna prepare 30 second vocal and try do some settings
@formal plover
i found this one which says crepe is the best option
lets see if i also get some errors
Thanks!
Ayo? @formal plover level 3 !!! 
oh this one, i did installed locally
-colab
- Applio, by IA Hispano Google Colab
- AICoverGen-WebUI, modded by Hina Google Colab
- AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
- AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
- Ilaria RVC, by thestingerx Google Colab
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Applio, by IA Hispano Huggingface Spaces
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HFv2, by r3gm Huggingface Spaces
- AICoverGen, by r3gm Huggingface Spaces
- Advanced RVC Inference, by r3gm Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
Ayo? @brittle scroll level 2 !!! 
This is All I get
What collab is this???
RVC BOH
-colab
- Applio, by IA Hispano Google Colab
- AICoverGen-WebUI, modded by Hina Google Colab
- AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
- AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
- Ilaria RVC, by thestingerx Google Colab
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Applio, by IA Hispano Huggingface Spaces
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HFv2, by r3gm Huggingface Spaces
- AICoverGen, by r3gm Huggingface Spaces
- Advanced RVC Inference, by r3gm Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
-rvc
Full AI Voice Model Training Guide (Local)
Link: YouTube
Credits: Christopher Villanueva
Model training with Mainline RVC
Link: Rentry
credits: Raven (ravencutie21)
AICoverGen Colab Guide
Link: Google Docs
Credits: Eddy (Spanish Helper)
Create a model with RVC disconnected (colab)
Link: Google Docs
Credits: Angetyde
How To Make an AI Cover With Ilaria RVC
Link: Rentry
Credits: 👽 Julia (ailen2091)
SVC is old
Rvx sounds more human
Few ninutes
Also in cover gen ui click on runtime > run all and a public link in blue 🔵 URL will appear
Okay me doing now
Taco libre
Thank you*
Dumb auto correct
Lol
Try
you mean rvc?
can u tell me wich url are you using cuz the Weights arn't workig for me
Thank you it's working so far
How much you need?
Have you found the solution to this problem?
it is installed on C: but this is what i get...
I try other colab
Ayo? @versed tusk level 1 !!! 
I use this one w-okada's Voice Changer | Colab
was workiong with me normally but not anymore
Ayo? @olive bay level 1 !!! 
this works for me
but i have colab pro
Just install the voices manually sometimes no models installed
Is it real time voice changer?
RVC Guides (How to Make AI Cover)
Translation by country
The W Okada yes
the link you sent me
The one that you talks and change your voice? Or the one you put an audio to change the voice?
The one that I talk and change my voice
ok
HI
Can’t I upload harvest models to the voice-models channel? If I remember correctly, I understand that you cannot upload anything other than rmvpe and mangio-crepe. Is this correct?
Help pls with that problem
JSONDecodeError Traceback (most recent call last)
<ipython-input-15-1d7b23896ae4> in <cell line: 31>()
31 if os.path.exists(config_path):
32 # File exists, proceed with creation of creds and client
---> 33 creds = Credentials.from_service_account_file(config_path, scopes=scope)
34 client = gspread.authorize(creds)
35 else:
5 frames
/usr/lib/python3.10/json/decoder.py in raw_decode(self, s, idx)
353 obj, end = self.scan_once(s, idx)
354 except StopIteration as err:
--> 355 raise JSONDecodeError("Expecting value", s, err.value) from None
356 return obj, end
JSONDecodeError: Expecting value: line 1 column 1 (char 0)
pretty sure thats an outdated colab
go on here and choose what youre looking for
Last update: Mar 8, 2024
try this
Last update: Mar 8, 2024
or this
Why do I keep getting errors?
so while I was generating my song, my rvc started smoking weed and showed me this message at the very end:
"During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/content/AICoverGen-NO-UI-en/src/main.py", line 346, in <module>
cover_path = song_cover_pipeline(args.song_input, rvc_dirname, args.pitch_change, args.keep_files,
File "/content/AICoverGen-NO-UI-en/src/main.py", line 316, in song_cover_pipeline
raise_exception(str(e), is_webui)
File "/content/AICoverGen-NO-UI-en/src/main.py", line 85, in raise_exception
raise Exception(error_msg)
Exception: index -1 is out of bounds for axis 0 with size 0
1"
any ideas?
Do you mean, the AICoverGen colab.
Yeah that
Ayo? @obtuse cape level 1 !!! 
i dont have this option on mangio
to your experience, which rvc tool is best for voice changing?
and which one of these also wanna know. it says on the site crepe is best but it doesnt tells anything about the others
for realtime voice changing ?
what youre doing has nothing to do with realtime voice changer if you mean that. what youre doing is used for for example AI covers or training voice models
yeye
check the contents inside that folder... hm
U can still change to other numbers but you need to change the pitch of the instrumentals too
Help! i dont know how to upload a voice model to huggingface 😦
Ayo? @rustic trellis level 1 !!! 
@proper shale can you help?
just find the section that talks abt that
Last update: Mar 10, 2024
RVC Guides (How to Make AI Cover)
Translation by country
Yeah all the stuff is in there, but on the old colab when u saved a model it saved it as a zip that u could then copy the link of that zip and paste it into the other colab that lets u use the model. When I used the export cell all it did was make a folder. There’s no model zip that I can get the link and paste it into the colab and use the model.
Looks like this
Frank1 is the dataset zip
The folder contains things like the index, .pth, etc
But there’s no model zip
Ayo? @feral saffron level 4 !!! 
In the old colab there would always be a finished folder and inside that folder would be model zips
download the index & pth, zip it, then upload to huggingface
Ayo? @proud elbow level 32 !!! 
How do you upload it to huggingface?
why i dont have this option even installed?
have u looked in the download center? i can show you where it is if youd like
i did downloaded from there but i cannot find it
it must appear here right?
you have vr architecture selected for your process method, choose mdx net instead
i did it already i know
you need to have this
dont take that one serious, just asking if this is the place it should appear after download. because it is downloaded not not here listed.
ooh
it is not something i can combine?
nah you have to do them one at a time
Ayo? @low tangle level 2 !!! 
after removing instruments, it starts little cracle in sound. why does this one happens?
bad audio input?
do you mean there's leftover noise, or it added audio artifacts?
no i mean crackle sound
like czzrzrtzt
.d
thats interesting, it could be either i guess. what model did you use?
i tryed many models like 2 models or 5 and same
Ayo? @trail trout level 4 !!! 
sometimes it can create artifacts when you process it many times
is the audio youre processing noisy or low quality?
decrackle and declick in izotope RX if u have it
not noisy but low quality like mp3 file
thanks for this info, it will help im sure
i also have a question. i was told that rvc doesn't learn anything new after 40 minutes. is that true? i can't find anyone else saying that
im new on rvc im not sure about this one
why do i keep on hearing myself cut out everytime i speak
What I meant that it's enough audio for training. If you do go for an hour, you'll have to set your batch size to 12 or 16 so it can read through all the files thoroughly
oh that was you lol
i would really like to try using all the data if possible so i'll look into batch size further, thank u
Yeah, i'm making a long dataset myself (49 minutes probably) so I might use 16
any idea what i should use for 90 minutes or is it not that simple?
16 then. I'm using colab and you're using local right? So really depends on gpu
go 16 esp if u have 12 gb vram
sweet thanks yall
ya my gpu isnt great (gtx 1650) but it seems to be getting the job done lmfao i just cannot get the colab to work anymore
what colab issue?
well for starters i cant even find the official notebook anymore so i was using rvc disconnected. i'll try it again later and tell you the errors
1650 seems only usable for inference
i mean it did train the model. the model sucks but i thought that there was just something else wrong with it
Can anyone help me? I've been trying to solve something on my own for a long time but now I've given up and decided to come here and ask about it because honestly I don't have any other alternative. Im going insane
whats ur question
I don't know if it's something dumb or not, but I want to understand why no GUI has the Record from Microphone option anymore.
I'm not talking about a Realtime voice changer option, im talking about an option to record your audio file using your mic directly into the GUI and use that audio file. All the GUIs had this in the past, but suddenly none of them do anymore, and I can't find any that have this function nowadays
is there a way to train an ai voice without google colab (i used to use kits.ai but they got greedy over time lol)
for free ofc
Ilaria RVC GUI still has it
-colab
- Applio, by IA Hispano Google Colab
- AICoverGen-WebUI, modded by Hina Google Colab
- AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
- AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
- Ilaria RVC, by thestingerx Google Colab
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Applio, by IA Hispano Huggingface Spaces
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HFv2, by r3gm Huggingface Spaces
- AICoverGen, by r3gm Huggingface Spaces
- Advanced RVC Inference, by r3gm Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
I didn't know about this Gui and it's really good, thank you! But it was my fault for not specifying it in my first message, i kinda need to know if there's any LOCAL GUI that still has that record from mic function. I don't work well with Google Collabs
?
The zip folder needs to have no special characters or spaces
so remove the underline if you can with the file you uploaded on google drive
Ayo? @dusk oracle level 3 !!! 
Remove it from the google drive and reupload. Not trying to be mean but read carefully https://docs.aihub.wtf/rvc/cloud/rvc-disconnected/#rvc-disconnected so you don't get errors
no its fine
thanks
oh im stupid the actual audio files had numbers and underscores
my bad
i thought this was easy 💔
wait chat im retarded
i forgot to run the set training variables
🐶💔
sorry to bug u guys
ah ok, that's it?
Can somebody help me please? Somebody told me to turn the .pth and index to a zip and upload it to hugging face and I think I did it right hopefully, but whenever I try to put the link into the llaria colab it keeps giving me an error
need download for RVC 2
Can anyone tell me whats wrong with this spectrogram
what happen to Ilaria? it gone
it's the new local installation
#📰│dev-updates message
no, that the one that gone, i did go to it github by the link
the old HF space was too slow, use the colab one instead
or the local one is enough for inference even on average laptop cpu
i'm looking for the local one, i think i did saw it zip there but not yet download it before it gone
well i will just wait for next update about that i guess
oof
hello, i'm trying to download required models, should I put " python src/download_models.py " in CMD?
@proven hill it was there before 
(or probably the next stable release)
ModuleNotFoundError
Traceback (most recent call last)
<ipython-input-1-0f3af68c63f> in <cell line: 4>()
2 #@markdown Link the URL path to the model (Mega, Drive, etc.) and start the code
4 from mega import Mega
5 import os
6 import
Shutil
ModuleNotFoundError: No module named
'mega'
Can you guys help me ?
how do i use the rvc gui for real-time? i have an AMD gpu
it either crashes when i choose anything other than windows directsound, or it just does nothing and passes my voice through
outdated colab
i'm talking about go-realtime-gui-dml, to be specific
nevermind, i think i figured it out
Can someone provide me the dated link ??? For AI colab
Ayo? @nimble widget level 1 !!! 
RVC Guides (How to Make AI Cover)
Translation by country
how do I use a voice pack
Is anyone able to make an ai model for me
You can find rvc google colabs in our docs https://docs.aihub.wtf
Last update: Mar 10, 2024
what colab u using?
@low shard Hi im sorry for pinging you but can you help me?
what problem do you got?
if i dissapear its bc im in class lol
Ok so when I exported my model on the RVC disconnected colab it saved it as a folder instead of a zip model. The folder has the index snd .pth in it, but I can’t copy the folder link and paste it into the llaria colab because it says it’s an invalid link.
How do I get a link I can use
Oh okay it’s all good
wait so do you got saved the folder on your device or not ?
I’m on iPad so I use google drive and it saved it on Google drive
Frank1.zip is the dataset
ohh on drive, you just need to download the modelname.pth and the added index from google drive and zip it yourself
I zipped the index and .pth and I put it on hugging face or at least I think I did
And I downloaded it off hugging face and put it on google drive and copied the link and it still didn’t work
Did I not grab all the right stuff?
There’s a lot of .pth’s in the folder
The G and D .pth I wasn’t sure if I was supposed to zip those or not
you need to download only those 2 files i said dw
nono dwa
you need to leave the zipped name.pth and added.index on hugging face, you dont need to put it back on google drive
Yeah but even when I used the hugging face link it still gave me an error
These r the ones I used
And there’s another index I used
send the hugging face link
Okay one second
You need to put only 1 pth
No I don’t know anything about the tensor board. I just followed the ai hub guide.
yea the ai hub guide has the tensorboard
well in this case
put only the modelname.pth and the added index
The added
Use only the added one
Okay thank you
I genuinely appreciate your help
So I repupload that on hugging face?
btw the tensorboard is a tool to know what epoch is good for your modle and like if u putted too many or too less, next time u should use that
yea put the newly zipped name.pth and added.index and get the new downlaod link
yea the tensorboard helps u get better quality
Bro thank you so much for helping me. I’ve been stressing all day trying to look for help. Do you mind if I add and message you incase it doesn’t work? I know ur in class and I don’t wanna keep bothering u if ur busy.
it may be better if u ask there so if im busy others can reply yk, also yw
Honestly bro I’ve been asking all day and I’ve barely gotten a response from anybody. You and NotMJ are pretty much the only ones who’ve helped me.
Is that overtrained
Dammi una risposta ti prego
Visto che parli italiano
Sorry but see what I mean?
This is with the added.index and the name.pth
that looks like overtraining, check the step of the lowest point and get the pth with the most similar step
send downlaod link again
lemme see in the files
I had it saved at 380 somewhere
AttributeError: module 'huggingface_hub.constants' has no attribute 'HF_HUB_CACHE'
can someone explain to me why i cant access public irl with my zip linkj
This is 390 overtrained