if it works, it works
#✨│ai-help
simple ore
simple ore
gradio is unlikely the cause of your problems
lost lodge
gradio is unlikely the cause of your problems
oh
then what caused this error to pop up...?
simple ore
what error?
uncut folio
.
lost lodge
wait ima run it again and show it to u
rare gobletBOT
Ayo? @lost lodge level 7 !!! 
lost lodge
when i run the last cell i get this message
simple ore
none of that are errors
simple ore
Ensure that GPU support is enabled in your Kaggle notebook and that a GPU accelerator is attached to your notebook session. You can verify this by checking the "Accelerator" setting in the "Settings" panel of your Kaggle notebook.
lost lodge
lost lodge
it took ages to finish running the cell on the message i got
[waited 3 hours and it still didnt finish running the cell yet]
lost lodge
Ensure that GPU support is enabled in your Kaggle notebook and that a GPU accele...
Its taking so long than usual to succesfully run the final cell
onyx wigeon
how do i make my own model
knotty moth
(blurred it cuz idk if its my ip or smh)
just open all those links
knotty moth
visit site
knotty moth
I couldve did that before 😭
that's why you should have read the guide carefully
https://rentry.co/RVC-Mainline-Kaggle
lost lodge
that's why you should have read the guide carefully
https://rentry.co/RVC-Mainli...
yeah
brittle wing
Is RMVPE+ better than RMVPE?
simple ore
no, it just has a f0 min/max limit
rare gobletBOT
Ayo? @turbid temple level 1 !!!
turbid temple
@knotty moth how do i make datasets
gray dagger
can you train a model with an amd gpu
dusky lion
This may be a dumb question, but i notice on this list that half of the models (ones that are 500mb and 1gb) do not work in my RVC program, is there another program for these and are they higher quality? https://huggingface.co/QuickWick/Music-AI-Voices/tree/main
simple ore
This may be a dumb question, but i notice on this list that half of the models (...
which one exactly?
uncut folio
hey im trying to finetune text to text AI and now i wanted to fine tune text to speech or speech to speech AI and anybody knows how to do it becouse everytime i tried it failed and in the internet there are no good tutorials
I got a GeForce RTX 3060 in my pc
simple ore
in that folder there are D/G pretrains
dusky lion
Here is one for example https://huggingface.co/QuickWick/Music-AI-Voices/tree/main/Kali Uchis 30k
Any of them that dont say (RVC) in the title, these are usually much larger zip files than the RVC ones
gloomy matrix
Anybody know if the voice changer works for macbook? I want to know that its possible before I waste my time.
I am aware it is available on Mac, but unsure if it would work on a macbook.
simple ore
Here is one for example https://huggingface.co/QuickWick/Music-AI-Voices/tree/ma...
those dont seem RVC models
config files are different
probably some other singing models
dusky lion
yeah would you happen to know what they are for and would they be higher quality than the RVC ones?
simple ore
gpt sovits probably
knotty moth
Here is one for example https://huggingface.co/QuickWick/Music-AI-Voices/tree/ma...
over 1 year ago lol
simple ore
yeah, probably that
dusky lion
over 1 year ago lol
Most of the RVC models on that link are also 1yr ago lol so i didnt know, what is the best place to find the highest possible quality models?
azure marshBOT
Most of the RVC models on that link are also 1yr ago lol so i didnt know, what i...
A model is the result of training on a dataset. You can learn more about it in the Applio Docs
gloomy matrix
Anybody know if the voice changer works for macbook? I want to know that its possible before I waste my time! apologies if thats a dumb question.
simple ore
technically macbook with MPS can run pytorch... in reality not so good
uncut folio
hey im trying to finetune text to text AI and now i wanted to fine tune text to speech or speech to speech AI and anybody knows how to do it becouse everytime i tried it failed and in the internet there are no good tutorials
And I hope somebody could help me
I got a GeForce RTX 3060 in my pc
gloomy matrix
I figured it'd be pretty awful, thanks.
knotty moth
Most of the RVC models on that link are also 1yr ago lol so i didnt know, what i...
that's why #1175430844685484042 and weights.gg are better place to find
simple ore
you can test weights.gg with their samples
brave garnetBOT
Made with love by Ilaria 💖
unreal hill
guys how do i install it
i cant figure it out
like i legit do NOT know what to download bruh 
rare gobletBOT
Ayo? @unreal hill level 1 !!!
rare gobletBOT
Ayo? @midnight relic level 1 !!!
midnight relic
Is that bad
brittle wing
-colab
brittle wing
-guides
low shard
Is it possible on iPhone
possible to do what?
low shard
<@681186927151546397> how would one install rvc
btw for helpers watching this, i already redorected him to the guide in general
cerulean sage
i was doing sum training but it was taking like a minute per 1 epoch whats up w that
@low shard nicky wicky
low shard
i was doing sum training but it was taking like a minute per 1 epoch whats up w ...
whats ur pc gpu
cerulean sage
whats ur pc gpu
im doing colab
low shard
@hoary nimbus im pinging u in the right channel, whats ur pc gpu
low shard
i was doing sum training but it was taking like a minute per 1 epoch whats up w ...
ig should be fine
Time kinda depends by gpu, dataset, batch size
hoary nimbus
<@957864410750275584> im pinging u in the right channel, whats ur pc gpu
I am using macbook
low shard
I am using macbook
you can’t train locally (on ur pc) on a macbook
its not powerful + not supported for training, only inference (use models)
low shard
I am using macbook
As you dont got a good PC, its better you use cloud (remote good pc) for training an RVC Voice Model:
- Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
- Applio (ui)
- Mainline (UI)
- RVCDISCONNECTED (no ui)
- Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
- Mainline (UI)
- Applio by Vidal (UI)
- Applio by Shirou (UI, no guide as of right now)
- Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
lilac carbon
hi i need help making an ai voice
hallow dragon
Thankyou ! i’ll try this
low shard
hi i need help making an ai voice
What's ur PC GPU
lilac carbon
What's ur PC GPU
one sec let me check
low shard
one sec let me check
You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU
lilac carbon
AMD Radeon RX 7600
lilac carbon
U could try with applio zluda
ok thanks
low shard
your welcome
low shard
make sure you dont skip any steps 🙂
nah skip all the steps 🔥 
whole vault
How do I use the voice cover thing for a song link on my phone
low shard
How do I use the voice cover thing for a song link on my phone
so, you want to make an ai cover?
misty owl
-colab
frozen zinc
guys how can i fix some machine sound
dreamy nymph
Which program do you use for auto segmtentaion?
low shard
@urban flint i pinged u in the right channel, whats ur pc gpu?
odd shale
guys how can i fix some machine sound
!howtoask
patent trellis
odd shale
Which program do you use for auto segmtentaion?
Do you mean dataset segmentation?
You can use Audacity's audio labeling function.
dreamy nymph
oh tanks!
gilded frigate
What's better for ai covers? Mango applio or like a realtime rvc voice changer thing.
I tried mango and applio but covers on yt sound alot less ai like and the voice overal sounds better.
gritty cargo
illaria falls?
low shard
illaria falls?
what?
gritty cargo
im triying to use it but dont work
low shard
What's better for ai covers? Mango applio or like a realtime rvc voice changer t...
all rvc versions have the same quality
what matters is the model quality and how much u cleaned the vocals
low shard
im triying to use it but dont work
"Illaria falls" doesn't exist wdym
gritty cargo
illatia doesnt work
gilded frigate
all rvc versions have the same quality
what matters is the model quality and how...
Whats the best stuff to use to clean vocals?
rare gobletBOT
Ayo? @gilded frigate level 1 !!!
gilded frigate
Reverb remover helps a whole lot probably
gilded frigate
Is there something i could do in audacity to clean it up more? Maybe make it louder or smth
gilded frigate
UVR https://docs.ai-hub.wtf/rvc/resources/vocal-isolation/
Oh shii
I didnt know that was a thing
gilded frigate
UVR https://docs.ai-hub.wtf/rvc/resources/vocal-isolation/
Do you also perhaps know what the best osaka (ayumu kasuga) model is?
I got a 900+ epoch one from huggingface but maybe theres a better one somewhere idk
low shard
Do you also perhaps know what the best osaka (ayumu kasuga) model is?
no, there are over 20k rvc models
gilded frigate
Damn
low shard
I got a 900+ epoch one from huggingface but maybe theres a better one somewhere ...
more epochs doesn't mean more quality
the only way to see is try
low shard
Really?
yes https://docs.ai-hub.wtf/rvc/resources/epochs-tensorboard/ people who train models use the tensorboard
gilded frigate
Yeh i read that it can have problems with new data bc its too dependent on the stuff it trained on
Interesting stuff
uncut folio
Hey idk why but my Appolo webinterface is not creating an index i checked it and its not there but its says in the webinterface that it has been succesfully created and i get this error then and im pretty new so idk if its okeys or bad:
Starting preprocess with 7 processes...
100%|████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:07<00:00, 3.82s/it]
Preprocess completed in 7.64 seconds on 00:03:45 seconds of audio.
Traceback (most recent call last):
File "D:\ApplioV3.2.6\rvc\train\train.py", line 84, in <module>
with open(config_save_path, "r") as f:
FileNotFoundError: [Errno 2] No such file or directory: 'D:\ApplioV3.2.6\logs\my-project\config.json'
An error occurred extracting the index: [WinError 3] Das System kann den angegebenen Pfad nicht finden: 'D:\ApplioV3.2.6\logs\my-project\v2_extracted'
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.
An error occurred extracting the index: [WinError 3] Das System kann den angegebenen Pfad nicht finden: 'D:\ApplioV3.2.6\logs\my-project\v2_extracted'
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.
An error occurred extracting the index: [WinError 3] Das System kann den angegebenen Pfad nicht finden: 'D:\ApplioV3.2.6\logs\my-project\v2_extracted'
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.
An error occurred extracting the index: [WinError 3] Das System kann den angegebenen Pfad nicht finden: 'D:\ApplioV3.2.6\logs\my-project\v2_extracted'
If you are running this code in a virtual environment, make sure you have enough GPU available to generate the Index file.
mystic marsh
-colab
delicate kettle
Can anyone tell me which of these are better
Jammable or musicfy lol for ai covers
odd shale
Jammable or musicfy lol for ai covers
Any rvc colab.
Because you don't have to pay for anything.
Jammable is just the same as RVC but paywalled.
odd shale
and does anybody know how to add voice effects
You can add them to your output made on RVC using any DAW of your preference.
mystic marsh
do anybody got a good link for ai covers i dont like these ones
latent kettle
do anybody got a good link for ai covers i dont like these ones
That's all we have
late charm
In Colab I have 3 hours of training available and I have a 15-minute dataset, can I manage to train the model every 100 epochs?
latent kettle
In Colab I have 3 hours of training available and I have a 15-minute dataset, ca...
Easily you can train 100 epochs in three hours
late charm
Easily you can train 100 epochs in three hours
thanks for your help.
rare gobletBOT
Ayo? @late charm level 3 !!!
latent kettle
thanks for your help.
Yw : )
latent kettle
thanks for your help.
If you want to train for a long time use kaggle
late charm
If you want to train for a long time use kaggle
Móvile.
latent kettle
Bit harder than colab but have 30 hours of GPU runtime. Without disconnecting
@late charm
late charm
I use a mobile phone, I tried to follow some guidelines and I was more lost than the boy's mother.
latent kettle
I use a mobile phone, I tried to follow some guidelines and I was more lost than...
It is possible to use kaggle on mobile
Anyways...
late charm
Yes, but I get confused and in the end everything goes wrong.
latent kettle
Then use colab
brittle wing
I use a mobile phone, I tried to follow some guidelines and I was more lost than...
You're here?
late charm
Then use colab
Yeah
brittle wing
Hey
Use RVC d
And
Do you want to send the files dataset and stuff so I can help you and the model gets trqined quicker
latent kettle
Do you want to send the files dataset and stuff so I can help you and the model ...
Do you have a GPU
late charm
Do you want to send the files dataset and stuff so I can help you and the model ...
It is confidential
latent kettle
Mhm I offered help
Which GPU do you have
latent kettle
For your local system
latent kettle
The person doesn't want help it's okay
But I want 😆
latent kettle
To train a deepfacelab model
latent kettle
I don't like to train models
Why did you offered him to train a model
brittle wing
I only know about voice models
Well because it's a voice model
And It's a model I want
latent kettle
I only know about voice models
I'll tell you how to do. It is used to make professional deepfakes
brittle wing
Sorry I'm not really interested
latent kettle
Well because it's a voice model
So you both want a same voice model.
Not a problem
brittle wing
So you both want a same voice model.
Nope it's a voice model I need
And I wanna help the person train it faster
brittle wing
BTW do you have tesla t4 on your computer. Like locally?
No in Colabs I dunno own a computer
latent kettle
No in Colabs I dunno own a computer
Oh.. then you can't help me. It is banned by Google colab. I thought you have it in your local pc
brittle wing
Oh.. then you can't help me. It is banned by Google colab. I thought you have i...
Yeah
brittle wing
Np
sonic mantle
-colab
sonic mantle
Use <#1202754985255764060> to execute the commands
brittle wing
-overtraining
azure marshBOT
-overtraining
You can detect if a model is overtraining if the TensorBoard graph starts to rise and never comes back down. An overtrained model will sound robotic, muffled, and won't be able to articulate words well.
Check these resources to learn more about this topic
- Epochs & TensorBoard from
AI HUB Docs - TensorBoard from 🍏 Applio Docs
brittle wing
-help
brittle wing
-audio
gloomy gate
When I open the mic in the Counter Strike 2 game, why does the sound delay and lag? is there any way to solve this problem? I use a GTX 1650 SUPER
marsh schooner
im tryna make a model and i did backround music removal but the music has singing in it so its peaking up the singing voice is there a way that i can automatically remove that singing sense its a lower volume compared to the voice model im trying to make
bc as u can see those spikes right there is the voice im actually trying to get and its louder than the music in the back
rotund cairn
why is my voice so laggy
mild sluice
Does anyone help me create ai cove?
azure marshBOT
Does anyone help me create ai cove?
mild sluice
@native crystal He doesn't answer me I need someone to help me on his behalf create ai cove
tiny dragon
Uhmm where do I post like the wips of a cover or whatever?
elder belfry
how to make rvc own voice model
low shard
how to make rvc own voice model
What's ur PC GPU
low shard
Uhmm where do I post like the wips of a cover or whatever?
Like post ur ai cover? Post it in yt then share the link in #1159290752195633273
elder belfry
What's ur PC GPU
rtx 6090 heheheh nly fun bro only tell me how i make like complete video link or whatecver to make own voice model i will make i will arange any gpu
rare gobletBOT
Ayo? @elder belfry level 2 !!!
low shard
rtx 6090 heheheh nly fun bro only tell me how i make like complete video link ...
... You seriously need to tell me ur gpu
and no, there is no single yt video for it, all on yt are OUTDATED
You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU
And training takes alot of computing, it cant work on any gpus
elder belfry
... You seriously need to tell me ur gpu
bro now i havce rc480 4gb but my czns have gaming shop so i will grab any gpu my i have rx480 4 gb
low shard
bro now i havce rc480 4gb but my czns have gaming shop so i will grab any gpu my...
that gpu sucks
and has low vram
you can't train locally (on ur pc) with that
u have to use cloud (remote good pc)
As you dont got a good PC, its better you use cloud (remote good pc) for training an RVC Voice Model:
- Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
- Applio (ui)
- Mainline (UI)
- RVCDISCONNECTED (no ui)
- Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
- Mainline (UI)
- Applio by Vidal (UI)
- Applio by Shirou (UI, no guide as of right now)
- Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
Kaggle suggested
elder belfry
u have to use cloud (remote good pc)
how i do i amn completeluy beginer even i dobnt know hich website or what hugging face use foir ytraining model
i am completely beginer anyone help me completelyonly 1 time
azure marshBOT
i am completely beginer anyone help me completelyonly 1 time
low shard
how i do i amn completeluy beginer even i dobnt know hich website or what huggin...
huggingface is the biggest ai platform, but you cant train on that, i gave you 3 cloud computing services
its better you click on Kaggle Mainline
its a guide to do it on kaggle, which gives alot of gpu time
elder belfry
huggingface is the biggest ai platform, but you cant train on that, i gave you 3...
can you msg me in personal if you do its help me alot plz brothr
low shard
can you msg me in personal if you do its help me alot plz brothr
i rather prefer we chat there, its easier for both of us and other helpers can also help too in case
first of all its better you read the guide
elder belfry
i click on kagles mainlin ui
its direct me on website what i do next
bro its such a too much hard can anytbody make free model of my own voice
@low shard how much you charged for my project
low shard
<@911742715019001897> how much you charged for my project
charged?
I didn't charge you anything
If you meant 'how much do you take for a paid commission of a model', i don't do those things
you can either #1159289738314919936 or #1191429836321849435 if u want someone to do it for u
low shard
its direct me on website what i do next
you need to read the guide, it tells u what to do
elder belfry
you can either <#1159289738314919936> or <#1191429836321849435> if u want someon...
oh ok thanks
brave garnetBOT
⠀
Google Colabs 
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
⠀
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
⠀
Made with love by Ilaria 💖
uncut folio
Hey does anybody know how I can create my own llm to text like chat gpt with cuda?
brave garnetBOT
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
snow hazel
Is there a website that's good with robotic voices?
brave garnetBOT
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
distant turtle
-colab
umbral grotto
-help
brave garnetBOT
⠀
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
⠀
Made with love by Ilaria 💖
⠀
Google Colabs
⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.
AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.
RVC Disconnected
To train new voice models, by Kit Lemonfoot.
EasyGUI
The OG interface, by Rejects.
⠀
Made with love by Ilaria 💖
⠀
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
⠀
Made with love by Ilaria 💖
⠀
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
⠀
Made with love by Ilaria 💖
pliant axle
@low shard mines the same as yours but uhd
low shard
<@911742715019001897> mines the same as yours but uhd
that wasn't mine btw, random from google
pliant axle
oh
low shard
both ones are like that?
pliant axle
both ones are like that?
NVIDIA GeForce RTX 3050 6GB GPU
low shard
bc that's integrated graphics which is bad
low shard
NVIDIA GeForce RTX 3050 6GB GPU
kinda better, what are you looking for?
pliant axle
sorry im on a work laptop
pliant axle
kinda better, what are you looking for?
i want like to change my voice in the studio to a specific artist
low shard
training (making models), inference (using models) on pre-record audiso, inference on realtime for calls
low shard
i want like to change my voice in the studio to a specific artist
so, inference on pre-recorded audios right? like u record ur voice in the studio then change it into the ai one
pliant axle
so, inference on pre-recorded audios right? like u record ur voice in the studio...
yea
low shard
yea
Your pc is good enough to use RVC locally (runs on ur pc), you can choose between those for doing it locally if u really want to:
- Applio: A fork of RVC with some extra features like Applio TTS, same quality tho
- Mainline: The original RVC
but honestly its not that good so it will be slow, id suggest cloud (remote good pc) like Ilaria RVC Zero which will also be easier than installing it locally, the only bad thing is its not unlimited as local so you willl have a zerogpu quota for your account
brave garnetBOT
Ilaria RVC: CLICK HERE 🤗
Guide on how to use it: CLICK HERE 📝
Don't forget to thank Ilaria if you find it useful! 💖
pliant axle
Your pc is good enough to use RVC locally (runs on ur pc), you can choose betwee...
thank you
knotty moth
NVIDIA GeForce RTX 3050 6GB GPU
that's good enough and still faster than cpu mode
btw you can even port the repo of ilaria RVC space locally (not to be confused with the old "ilaria mainline")
simple ore
3050 is colab speed
low shard
that's good enough and still faster than cpu mode
btw you can even port the repo...
Ilaria RVC Zero is optimized for ZeroGPU, Ilaria Mainline isn't the old version, its the only local ilaria rvc version
simple ore
kaggle will be faster
low shard
3050 is colab speed
that huggingface space runs on ZeroGPU
which is an A100, alot faster than google colab and kaggle
knotty moth
Ilaria RVC Zero is optimized for ZeroGPU, Ilaria Mainline isn't the old version,...
I thought Ilaria Mainline may have been abandoned, and btw I'm not sure but I noticed that somehow ilaria's rmvpe+ is less likely to produce voice cracks than the og mainline's rmvpe
pliant axle
which is an A100, alot faster than google colab and kaggle
how long this usually take
low shard
I thought Ilaria Mainline may have been abandoned, and btw I'm not sure but I no...
Ilaria Mainline may have been abandoned
Ilaria RVC Zero doesn't get any updates since 4 months either lol
but I noticed that somehow ilaria's rmvpe+ is less likely to produce voice cracks than the og mainline's rmvpe
That's super weird, are u comparing also the mainline rmvpe or mainline rmvpe+? iirc mainline has rmvpe+ too
pastel oak
That depends on your own upload speed
low shard
depends all on your internet, shouldn't take long
pliant axle
oh nvm it worked i thought it was frozen
rare gobletBOT
Ayo? @pliant axle level 2 !!!
low shard
knotty moth
> Ilaria Mainline may have been abandoned
Ilaria RVC Zero doesn't get any update...
the og one only has vanilla rmvpe, and iirc rmvpe+ was introduced in applio 2.x
low shard
the og one only has vanilla rmvpe, and iirc rmvpe+ was introduced in applio 2.x
yea ur right
hidden dew
Where do I drop my rvc models in applio local?
knotty moth
Where do I drop my rvc models in applio local?
logs folder, both the index & pth file
hidden dew
Thx
obsidian palm
hello, i have a massive delay between my real life input and eventual output into discord via vbcable, does anyone know how i could fix that?
pastel oak
hello, i have a massive delay between my real life input and eventual output int...
Whats your GPU, whats your settings, which version of wokada do you have (top right somewhere it says it)
delicate kettle
I’m just trying to figure out how people make covers but are able to block out the background vocals
jaunty shale
so I did some research and I am trying to make my models sound much crispier. but this made me confused.
so I used my voice for test. I put these settings and exported it.
regularly I leave sample rate as it is. It always sounds normal, but this time I changed it to the same sample rate
the low pitch voice sounds good, but when it gets higher, it gets distorted
any ideas why?
rare gobletBOT
Ayo? @jaunty shale level 9 !!!
cloud lance
does llaria support realtime?
azure marshBOT
odd shale
does llaria support realtime?
And no, Ilaria Mainline/RVC doesn't support realtime
brittle wing
so this is
rare gobletBOT
Ayo? @brittle wing level 2 !!!
shell dagger
how do i use #🔍│find-models
simple ore
so I did some research and I am trying to make my models sound much crispier. bu...
it is quite the opposite
you pick the sample rate of the model, the result is that anything you infer using this model will be in the same sample rate
whichever sample rate of the data set, it will be resampled to the sample rate of the model for learning and to 16khz for f0 extraction
resampling does affect quality, so keep that in mind
simple ore
the low pitch voice sounds good, but when it gets higher, it gets distorted
the distortion comes from the model never learning how to reproduce high pitched voices
odd shale
the distortion comes from the model never learning how to reproduce high pitched...
Yep, that's basic RVC logic
simple ore
funny thing, if you use a mix for male and female voices in the same dataset, the model sometimes switches to a different voice to sing, etc
odd shale
funny thing, if you use a mix for male and female voices in the same dataset, th...
That's true too.
shadow geyser
Anyone having issues where a voice app gets stuck or seems like it isnt loading?
rare gobletBOT
Ayo? @shadow geyser level 2 !!!
brittle wing
hey friends has anyone tried hooking up their voice model to some kinda shitty noise filter
I found a program called voicemeeter but it's kicking my ass rn
and cantabile another program
brittle wing
right I'm tryna use it with cantabile which lets you use plugins and stuff
it feels like it could work but idk I can't figure it out
calm vector
sorry for the dumb question but i downloaded a voice model and put the .pth file in the assets>weights folder and it works great. I was wondering if I should do anything with the index file? Thanks very much for any help
simple ore
sorry for the dumb question but i downloaded a voice model and put the .pth file...
if using rvc it goes into logs, I think?
jaunty shale
the distortion comes from the model never learning how to reproduce high pitched...
Strange.. I did it before without changing sample rate and it went fine
I use rvc disconnected and applio for conversion
Cuz this was first time this happened to me
jaunty shale
Cuz this was first time this happened to me
I tried 32k sample rate, and obviously it got worse-
I'm confused
cloud finch
my applio is in my E drive but when i use it my C drive loses space, can i make it so my E drive loses space instead?
simple ore
my applio is in my E drive but when i use it my C drive loses space, can i make ...
run WizTree, see what actually consumes your space
cloud finch
alright
simple ore
- pip cache, 2) huggingface cache 3) gradio temp folder
cloud finch
also
simple ore
that's likely culprits
cloud finch
Im tryna train a model, does the dataset have to be online, or can it be a local directory?
cloud finch
that's likely culprits
alr ill def check those out
simple ore
dataset creator is only needed when you have no access to a local drive
cloud finch
oh
cloud finch
When i did this i got an error
simple ore
no space
cloud finch
simple ore
simple ore
well, actually using a space is fine... so perhaps no files in such folder or no supported files?
anyway, it would really help to check the terminal window to see what's happening
simple ore
yeah, that wll do
karmic fjord
anyone who downloaded them as html files?
low shard
anyone who downloaded them as html files?
no need for that
read the message i linked u
there are new temp docs to use
upbeat portal
-colab
reef geyser
-rt
azure marshBOT
-rt
This interaction has expired, use the command
/guidesif you wish to see it again.
Command timed out. Use /help to for a new interaction
brittle wing
One message removed from a suspended account.
azure marshBOT
One message removed from a suspended account.
This interaction has expired, use the command
/guidesif you wish to see it again.
brittle wing
One message removed from a suspended account.
azure marshBOT
One message removed from a suspended account.
This interaction has expired, use the command
/guidesif you wish to see it again.
rare gobletBOT
Ayo? @brittle wing level 1 !!!
brittle wing
One message removed from a suspended account.
One message removed from a suspended account.
One message removed from a suspended account.
red kayak
One message removed from a suspended account.
its asking for an audio file and u are giving it a pth file...
brittle wing
its asking for an audio file and u are giving it a pth file...
One message removed from a suspended account.
red kayak
One message removed from a suspended account.
yeah any audio file u want to inference with
brittle wing
One message removed from a suspended account.
One message removed from a suspended account.
red kayak
the pth file should go into ur assets/weights/
and your index should go into logs/(create a folder inside the logs folder)/
brittle wing
the pth file should go into ur assets/weights/
One message removed from a suspended account.
red kayak
yeah thats where the pth goes
brittle wing
One message removed from a suspended account.
One message removed from a suspended account.
One message removed from a suspended account.
rare gobletBOT
Ayo? @brittle wing level 2 !!!
red kayak
One message removed from a suspended account.
yeah but dont name the folder in the logs folder, "logs" , it may confuse you. instead name it something like your model's name e.g MOZE
brittle wing
One message removed from a suspended account.
One message removed from a suspended account.
One message removed from a suspended account.
One message removed from a suspended account.
red kayak
One message removed from a suspended account.
yeah thats how u infer
brittle wing
One message removed from a suspended account.
One message removed from a suspended account.
distant turtle
-colab
fathom geyser
Hi, I have been dealing with this problem for days and still couldn't fix it. I restarted my computer, changed the folder location, I really did everything I could but none of those things helped me. Can someone please tell me how I can fix this error?
I'm using Applio by the way
simple ore
I'm using Applio by the way
did you select the model at the top?
fathom geyser
yes
simple ore
show it
fathom geyser
Well... I was testing it again and now it works. But the download failed
fathom geyser
Can someone help me with this please?
brazen helm
can someone help me setup rvc for mac if its possible to? im not sure how to do it
brave garnetBOT
⠀
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
⠀
Made with love by Ilaria 💖
fiery raptor
what does this error mean? (its in the UVR5 NO UI thing)
simple ore
I think it does not like the file name
fiery raptor
idk how to change it
low shard
can someone help me setup rvc for mac if its possible to? im not sure how to do ...
u can only inference for mac, not train
if u really want to do it locally u can try https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/en/README.en.md
or else use cloud (remote good pc)
late charm
Hey, I need help. I don't understand why I train a model to 40k and if it reaches 16k I have to train it again to 32k. How does this benefit the model?
simple ore
Hey, I need help. I don't understand why I train a model to 40k and if it reache...
" model to 40k" what?
late charm
Yeah.
late charm
i'm trying to make a sense of this question
It does not reach the frequency of 20k but 16k.
simple ore
if you have an audio signal that reaches 20KHz, then you need 40k sample rate model to train
if you have an audio signal that reaches only 16KHz, then you only need 32k sample rate model to train
there's no 'again'
here I have mp3 that is 48k, but the content is below 16k
fathom geyser
I've executed everything and it still fails. Why is this happening to me? 
stable hollow
Hey guys. What would you say is the best rvc real time voice changer currently? I'm using deiteris' Fork for W-Okada rn
Asking cause it's been a couple months lmao
simple ore
that's the one, maybe some minor changes since then
stable hollow
Good to know. Thanks :)
rare gobletBOT
Ayo? @stable hollow level 1 !!!
low shard
I've executed everything and it still fails. Why is this happening to me? <:skul...
Did you get a no-feature-todo in the feature extraction?
Bc it seems that there is a problem with your dataset
Are u sure it's in .wav
copper ether
How do I retrain starting from the same epoch I left off at
simple ore
you just start the training again for the same model, keep in mind that RVC has a bug that overwrites the latest weights/.pth, so you may want to keep a copy of that
you dont need to run preprocess/extract features step
copper ether
Thanks
mild sluice
Hi, I have been dealing with this problem for days and still couldn't fix it. I ...
How to
distant turtle
-colab
austere compass
hello, i am having issues viewing the grad_norm_d graph from the tensorboard. the data points just show NaN and nothing appears. Any known way to fix this?
quasi dagger
hello, i am having issues viewing the grad_norm_d graph from the tensorboard. th...
Did you click on the refresh button?
austere compass
Yes, nothing changes
civic thicket
can someone help me?
pastel oak
can someone help me?
Just ask instead of asking for help
low shard
can someone help me?
btw helped in #🧬│ai-chat message
simple ore
hello, i am having issues viewing the grad_norm_d graph from the tensorboard. th...
All is NaN? your model got f'ked
austere compass
All is NaN? your model got f'ked
oh nyo
simple ore
I've seen it a few times
simple ore
depends on whether you saved all weights or just latest
with just latest it is dead
you may recover a model you saved every epoch, if you did
austere compass
fun
i didnt
i guess i trained in a FP32 env THEN switching to a FP16 env killed it
rare gobletBOT
Ayo? @austere compass level 1 !!!
ebon topaz
idk why but when i use Gura the ping goes op to above 10Kms ping, but when i switch its down to 1 or 3ms
rare gobletBOT
Ayo? @ebon topaz level 1 !!!
simple ore
can anyone help me with this
you're trying to use iGPU
flint grail
oh
rare gobletBOT
Ayo? @flint grail level 2 !!!
simple ore
change HIP_VISIBLE_DEVICES in the bat file
craggy solar
guy
i'm gonna make my first 3 models of the slaughter me funkin opponents
but i have a feeling this will be hard
pulsar forum
craggy solar
i gonna make the fnf slaughter me street muppet models
knotty moth
hello, i am having issues viewing the grad_norm_d graph from the tensorboard. th...
stop using mangio that screws GTX gpus
oak edge
hello
My name is Arkeshan, shortly Ark, I'm trying to know about what is tensor thingy
in Mangio RVC, I'm training a data set 7.30 mins, is 5000 epochs enough?
low shard
My name is Arkeshan, shortly Ark, I'm trying to know about what is tensor thingy
it’s basically a tool to check how the training is going, u could see more about it in https://docs.ai-hub.wtf/rvc/resources/epochs-tensorboard/
low shard
in Mangio RVC, I'm training a data set 7.30 mins, is 5000 epochs enough?
5k epochs are way too much, id go with 500/1k, there isn’t a rigut amount of epoch but that’s enough time to see how it’s going with the tensorboard
what’s ur pc gpu btw?
oak edge
rtx 4060
rare gobletBOT
Ayo? @oak edge level 1 !!!
oak edge
its laptop btw
oak edge
Ayo? <@439461469688692740> level **1** !!! <:lfg:1159355870119993496>
Man what the
oak edge
How much vram (dedicated gpu memory)?
uhm 8
low shard
also laptop gpus are weaker than desktop ones
oak edge
I already trained 5000 epochs for a 1.30 minutes data set twice
oak edge
also laptop gpus are weaker than desktop ones
RTX 4060 laptop and desktop are same die
pseudo kite
never seen anyone do that
That's what i said
low shard
RTX 4060 laptop and desktop are same die
die??
oak edge
pretty much 5-10% deviation from desktop gpu die
low shard
oh thought u were telling me to die
oak edge
Sorry man I'm kinda new so could've overtrained a tiny teeny bit
low shard
That's what i said
yea usually with the tensorboard people find the right epoch under 1k epochs
oak edge
I'm trying to master the art of seperating vocals from songs, swapping it with ai , and make all the songs sang by my fav singer
low shard
Sorry man I'm kinda new so could've overtrained a tiny teeny bit
btw more epochs don’t mean more quality
oak edge
btw more epochs don’t mean more quality
one thing i noticed the voice quality doesn't change
but the regular electronic distortions
and some parts where volumes fluctuates becomes less
pseudo kite
Sorry man I'm kinda new so could've overtrained a tiny teeny bit
You've gone four digits 5 times 🙏
low shard
training too much could overtrain which could make worse quality or have no impact so no reason in training more
oak edge
training too much could overtrain which could make worse quality or have no impa...
i saw a video where a guy recommending to train 20000 so i thought i start with 5000 (Guy had 1650 super btw)
low shard
i saw a video where a guy recommending to train 20000 so i thought i start with ...
never follow yt tuts
they are like a year old
oak edge
training too much could overtrain which could make worse quality or have no impa...
Can I send 2 seperate Comparisons one with 250 Epoch then with 820 epoch
low shard
and wtf who would train with a gtx 😭
oak edge
made fresh today itself
oak edge
and wtf who would train with a gtx 😭
Idk man he just trained around 5 epoch for demonstration purposes in vid
low shard
Can I send 2 seperate Comparisons one with 250 Epoch then with 820 epoch
can’t listen to them rn but sure
(i’m at school)
oak edge
Man still your school allows, 1st world privilages I guess..
oak edge
oak edge
nah its digital school
even better
low shard
Well usually no one finds improvement (in rvc context) after that
oak edge
so trying all weird stuff out..
oak edge
Well usually no one finds improvement (in rvc context) after that
Man but the stuff I found was clear especially stuff i noticed is when the data training time improving the pronounciation becomes better
rare gobletBOT
Ayo? @oak edge level 2 !!!
oak edge
and sometimes the random voice cuts is lesser and the best of it all higher pitch songs are clear
low shard
when u train an rvc model, you train using a pretrainex model, basically a “model” trained by the creators (even if there are community ones) with like multiple voices in it, it’s done to help people train their models
without that, it would be from scratch, meaning you’d have to train with tons of hours of data and way more time
oak edge
I have a data set of same song one with 5000 epoch 1.30 minute training set, then now 7.30 minute 820 epoch, the 820 one has better clarity, very less distortions than 5000 with 1.30 minute set
oak edge
when u train an rvc model, you train using a pretrainex model, basically a “mode...
I don't even understand fraction of how this happens😭
Btw training took me to 800 epoch with 7.30 minute set only took about 6 hours so not a big time
low shard
I don't even understand fraction of how this happens😭
fortunately there are pretrains, without them ud really have to train alot more
training pretrains takes alot of time and data
(and power )
oak edge
fortunately there are pretrains, without them ud really have to train alot more
sooo is all getting sorted out for me?
by creators of mangio rvc?
low shard
sooo is all getting sorted out for me?
im just saying that you usually wouldn’t need to train with over 1k, or need to train with more than 1 hour of data (if u ever do) as u might have no improvements
low shard
by creators of mangio rvc?
btw, mangio rvc is a fork (modified version) of rvc, its outdated since a year
oak edge
Man then which gui you recommend i'm not good with CLI at all
low shard
i mean, its still usuable but usually people rather go with more updated things like mainline (the original rvc) or applio (an updated rvc fork)
they all have a UI
oak edge
they all have a UI
where can I download it man... I don't know much about installation either...
low shard
i mean, quality wouldn’t change, rvc did no improvements in a year unfortunately and has been left to rot from the original creators itself
Its more suggested to use them in case mangio does weird errors
low shard
where can I download it man... I don't know much about installation either...
ofc u can use mangio too, i was just telling u tho
oak edge
do I need to put 23 hours of training again??
low shard
i mean, quality wouldn’t change, rvc did no improvements in a year unfortunately...
the main developer rvc boss kinda left it ye, he works now on gpt so vits which is a tts program (while rvc is a sts program)
low shard
do I need to put 23 hours of training again??
23 hours of training?
is that how much it took u for doing the 5k epochs?
dang 23 hours are alot
oak edge
is that how much it took u for doing the 5k epochs?
no 9000 epochs
simple ore
5k epochs on 7 minutes is 4800 too much
oak edge
- 6 hours of this 820 epoch 7.5 minute set
oak edge
5k epochs on 7 minutes is 4800 too much
no
low shard
you have ALOT of patience
oak edge
5000 + 4000 of 1.30 minute different sets
simple ore
trying to squeeze a bathtub of lemon juice from one lemon
oak edge
5000 SPB, 4000 my own voice
oak edge
never train again over 1k
aight then I'll fix at 1000, because I seem to have a lot of time these days up until february
well so which one you recommend me to download
low shard
well so which one you recommend me to download
if u want a simple ui, u could try out applio
oak edge
I hope my laptop is not gonna die out before I go to uni due to all training.. one time when I put crepe extraction it just started to BSOD🥲
low shard
oak edge
if u want a simple ui, u could try out applio
is it better than mangio
low shard
don’t train alot
oak edge
don’t train alot
is crepe extraction better than rmvpe? for training>
but crepe seems to be ram hungry and makes my laptop BSOD almost all time
my lap is legion pro 5i, 4060,13700HX
low shard
is it better than mangio
well it still gets updates and the ui could be easier than it, personally yea
low shard
but crepe seems to be ram hungry and makes my laptop BSOD almost all time
it’s better to use rmvpe, it’s less sensitive to noises
oak edge
my lap is legion pro 5i, 4060,13700HX
oh yes ram is 16 gb
low shard
my lap is legion pro 5i, 4060,13700HX
damn u lucky
oak edge
it’s better to use rmvpe, it’s less sensitive to noises
okk... So man what are the benefits i can see by switching
low shard
i got integrated graphicsh with an i3 of 10 or 11 gen
oak edge
damn u lucky
nah it was a gift for my hard work
low shard
okk... So man what are the benefits i can see by switching
Btw u could see more about the f0 methods aka pitch extraction in https://docs.ai-hub.wtf/rvc/resources/inference-settings/#also-known-as-f0-they-re-the-algorithms-for-converting-the-vocals
copied the wrong thing oops, edited it
technically crepe has slightly better quality but it’s slower and more prone to noises
which is why most people use rmvpe which is still good quality, less proje to noises and faster
i’d suggest to use rmvpe
please ask in #🔍│help-w-okada
i’m not a wokada helper, but if ur following yt tuts, don’t
true owl
ty very much
low shard
ty very much
-rt
azure marshBOT
-rt
This interaction has expired, use the command
/guidesif you wish to see it again.
low shard
its better u follow the 1st written guide
its the wokada fork, which has better performance too
but yt tuts are mostly outdated
oak edge
i’d suggest to use rmvpe
yes mine already has noises
true owl
thanks for the help
oak edge
uhm man I was wondering should i jump to appolio
low shard
yw
oak edge
so what kind of benefits i'm looking at, less tearing artifacts
rare gobletBOT
Ayo? @oak edge level 3 !!!
oak edge
like electronic stuff between voice, will those reduce
do I have to train all over from scratch in applio>
low shard
uhm man I was wondering should i jump to appolio
technically mangio is less suggested as its old, id suggest applio
low shard
do I have to train all over from scratch in applio>
If u meant as “are there pretrains in applio”, yes there are pretrains in every rvc
oak edge
no no, I mean training the specific voice
low shard
the models are usable in both mangio andapplio
oak edge
in my case SPB
low shard
no no, I mean training the specific voice
didn’t u train that model already?
oak edge
I want to train his voice to swap to songs
oak edge
didn’t u train that model already?
yes in mangio
low shard
yea, u can use that model in applio too
oak edge
but will that work with applio i d k how to transfer this file to that
low shard
the models can be used in all rvc versions
oak edge
ohh... thankgod, can I continue to train from where i left(820 epoch) in applio?
low shard
They are like interchangeables
oak edge
I mean if the set is same will there be anydifference in outcome at all
like less artifact output, more clear voice
low shard
ohh... thankgod, can I continue to train from where i left(820 epoch) in applio?
i don’t really do local, but technically u should be able too
low shard
I mean if the set is same will there be anydifference in outcome at all
well nope with the same set
oak edge
i don’t really do local, but technically u should be able too
Man how old r u
low shard
if ur having noise issues, it’s better u retrain with rmvpe
oak edge
if ur having noise issues, it’s better u retrain with rmvpe
Oh my gosh again...
wait can you tell if a model has noise issues by listening?
low shard
started about ai like july 2023
oak edge
started about ai like july 2023
Dayum... I d k what I was even doing when I was 14, probably playing some games in my android tablet T_T
low shard
Oh my gosh again...
well yea
low shard
Dayum... I d k what I was even doing when I was 14, probably playing some games ...
oh yea i play games too, used to mod and hack mvs in 2022
(hack in modifying UE4 files to get free cosmetic, i dont ruin other experiences)
oak edge
oh yea i play games too, used to mod and hack mvs in 2022
Man is a prodigy
this is my current training set...
low shard
Man is a prodigy
did some basic mods, unfortunately idk about 3d models
oak edge
did some basic mods, unfortunately idk about 3d models
man that's still very very advanced at that stage, maybe I'm surprised since I'm from 3rd world T_T
Can I ask doubt about what the sliders do in RVC
low shard
man that's still very very advanced at that stage, maybe I'm surprised since I'm...
where are u from btw? I’m from italy
oak edge
where are u from btw? I’m from italy
Sri Lanka
low shard
Can I ask doubt about what the sliders do in RVC
about the sliders?
oak edge
yes what these even do
btw why shouldn't I place the appilo in D drive My C is really cramped atm
simple ore
you can place it anywhere
just dont place it into a folder with onedrive/spaces/funny local characters
low shard
yes what these even do
https://docs.ai-hub.wtf/rvc/resources/inference-settings/ these might answer you more easily than doing a wall of text lol
oak edge
thanks
oak edge
just dont place it into a folder with onedrive/spaces/funny local characters
oh thank you man
simple ore
this makes no damn sense... unless Mangio does something stupid
the model takes audio at 16k and produces the sample ratio it was trained with
oak edge
any idea what should I download
simple ore
v3.2.6
oak edge
the model takes audio at 16k and produces the sample ratio it was trained with
idk what those even mean so i'm just sliding them here and there
low shard
this makes no damn sense... unless Mangio does something stupid
Average mangio moment
oak edge
I just sometime put all sliders to max and click convert
simple ore
it marks all non-safetensors as sus
oak edge
Man 4.21 gb T_T
I'm gonna take my time train 1000 epochs in Mangio then in Appilo, then compare them with various songs T_T gonna take a week..
rare gobletBOT
Ayo? @oak edge level 4 !!!
simple ore
once again, 1000 epochs on <10 min file is crazy
but you can test the results every 10 epochs if you want
oak edge
but you can test the results every 10 epochs if you want
i mean are there any downsides other than more time ??
simple ore
it will be overtrained af
oak edge
it will be overtrained af
and....???
simple ore
robotvoices
oak edge
robotvoices
Won't the electronic slider thingy won't do anything to reduce that effect?
simple ore
no
oak edge
simple ore
if you want quality, get more audio
oak edge
Oh, then I'll go and check out 100 200 300 400 500 600 700 then 1000 = 3800 epochs of same file but different, Like SPB_100, SPB_200, SPB_300 and what about checking them individually to find the point where voice turns robotic?
oak edge
if you want quality, get more audio
I searched yt for a long time to find this 7.30 minutes clip, the singer is old and he passed away, only some pure clips of his singing is there in yt
I'm not comfortable about feeding voice ripped from UVR to train as it has many artifacts...
low shard
I'm not comfortable about feeding voice ripped from UVR to train as it has many ...
What uvr model did u use? Bc most people here use uvr
low shard
yea the with bs roformer iirc
oak edge
yea the with bs roformer iirc
idk names man T_T, But the ripped is still bad for the songs
I can hear those small echoes apart from pure voice some times violin music too
low shard
low shard
I can hear those small echoes apart from pure voice some times violin music too
echo can be removed using a de reverb or de reverb (used on the extracted vocals)
oak edge
echo can be removed using a de reverb or de reverb (used on the extracted vocals...
but that makes the song hard to voice swap, like voice cuts
and fluctuating volume of singer
low shard
but that makes the song hard to voice swap, like voice cuts
mm weird
azure magnet
how do i get the model to run from the .json file?
im tryna get the juice wrld one to work
low shard
how do i get the model to run from the .json file?
im tryna get the juice wrld ...
the only files u need to run are .pth and added.index for rvc
no json
azure magnet
oh, sorry
what do i use to run the .pth?
low shard
oh, sorry
what do i use to run the .pth?
are you following a yt tut btw?
azure magnet
are you following a yt tut btw?
no, i couldn't find a good one. i js got the model off the google sheet
rare gobletBOT
Ayo? @azure magnet level 1 !!!
low shard
no, i couldn't find a good one. i js got the model off the google sheet
its good bc those are old, and the google sheet is old too
don’t use those
You can search rvc ai voice models at:
- #1175430844685484042
- Send " @gusty kestrel search (name of the model)", without the ()
- Do /find with @earnest musk
- https://weights.gg/ (login required)
- https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice models)
- https://applio.org/models
- https://voice-models.com/
- https://thevoicemodels.com/ (for Turkish Models, login required with discord and level 2 on their server)
if there isnt one, you can:
- #1159289738314919936
- #1191429836321849435
- make it yourself with our docs guides https://docs.ai-hub.wtf/essentials/how-to-make-voice-models/
What’s ur pc gpu and what are u using btw?
low shard
i have a 1650
its not the best, good only for inference (using models)
what are u using btw
like applio, mainline
low shard
bc u can run the program either locally (on ur pc) or using cloud (remote good pc) computing services like google colab
azure magnet
ohhhhh
azure magnet
what are u using rn to make ai covers ?
nothing, this is my first time trying. i mean fl studio if thats what u mean, mvsep to seperate the stems of the song im using
low shard
nothing, this is my first time trying. i mean fl studio if thats what u mean, mv...
ohh lol
the program to use those speech to speech model is naked rvc
one sec lemme tell u
This is for locally
brave garnetBOT
Ilaria RVC: CLICK HERE 🤗
Guide on how to use it: CLICK HERE 📝
Don't forget to thank Ilaria if you find it useful! 💖
low shard
ilaria rvc zero is on cloud, its a zerogpu (A100, way faster than ur gpu) huggingface space
so ilaria rvc zero will be faster than ur gpu and easier, but u got limits so can’t use it 24/7, u got a quota per account, while local will be kinda harder to install and slower but unlimited
so its kinda ur choice
low shard
your welcome
austere compass
what difference does the index file have on the final inferred audio
low shard
what difference does the index file have on the final inferred audio
the index file contains the accent of the voice
austere compass
i see, thanks
low shard
yw
brittle wing
@low shard Can the uvr website release more drum models?
rare gobletBOT
Ayo? @brittle wing level 6 !!!
low shard
the x-minus website
i dunno about drum models
oak edge
hey man I tested 100 epoch vs 1000 epoch
for 7.30 minute audio..
also can I ask about UVR vocal + instrument seperation here?
odd shale
also can I ask about UVR vocal + instrument seperation here?
Your best bet for isolating vocals is BS Roformer.
If you ask.
odd shale
hey man I tested 100 epoch vs 1000 epoch
Also, more epochs doesn't mean more quality.
oak edge
I currently tried BS Reformer beta 4
also the epoch, the accent was heavy with 1000 epoch than with 100 and the major difference I noticed was the loudness
the spikes shown in audacity for the same song everything same settings except for the models one with 100 and other with 1000, the 1000 shown more loudness and bold pronounciations
with accent much closer to the model singer's voice
while 100 epoch was pretty low on loudness and some words slightly mispronounced and the accent wasn't much convincing as 1000 epoch
also are the ensemble with 1296+1297+23C INSTVocHQ better than MELBAND_reformer_big_beta_4?
odd shale
while 100 epoch was pretty low on loudness and some words slightly mispronounced...
Probably on 100 epochs it was overtrained.
odd shale
also are the ensemble with 1296+1297+23C INSTVocHQ better than MELBAND_reformer_...
You can just use BS Roformer (11.31 SDR version) on mvsep.
oak edge
I use the local version
odd shale
I use the local version<:skullsob:1159372531992645662>
You can try mvsep too.
oak edge
Probably on 100 epochs it was overtrained.
then is the loudness and accent getting much closer to the AI singer a bad thing? because lower I go on epoch it gets worse T_T
odd shale
then is the loudness and accent getting much closer to the AI singer a bad thing...
Ah i mispelled.
I mean, undertrained.
oak edge
You can try mvsep too.
MVSEP has these queue too
oak edge
I mean, undertrained.
than can I go above 1000? like 10000?
when can we actually see the limit?
odd shale
MVSEP has these queue too
Tho there are some models that aren't available on UVR.
odd shale
than can I go above 1000? like 10000?
You got a 7 min dataset, that would be overkill.
I'm not sure if you overtrained the model when reaching 1000 epochs.
Have you checked tensorboard?
oak edge
Have you checked tensorboard?
Still no idea how to check it
is that some software to be downloaded?
oak edge
Tho there are some models that aren't available on UVR.
so what's best on MVSEP for songs? I just need clean vocals
odd shale
so what's best on MVSEP for songs? I just need clean vocals
BS Roformer (11.31 SDR version)
oak edge
okk
aight so the tensor board bat is doing it's cmd thing now
Well it's not opening the localhost address for me
rare gobletBOT
Ayo? @oak edge level 5 !!!
supple hatch
where i take a link ? pls help
pastel oak
where i take a link ? pls help
Of what, theres many things you can do
oak edge
it shows a step number of 24.36k at the lowest
wait its around 3.96k... but this graph is weirder than the one in tutorial T_T
distant turtle
-colab
lean steeple
does anyone here have okada working with a very low delay / sample size? dm me
gilded frigate
how do i extract background vocals with uvr5? The "main vocal" extraction thing on the best models list doesnt do anything. In fact the vocal file is distorted and really low quality while the actual okay sounding bit still including background vocal file is named instrumentals
odd shale
how do i extract background vocals with uvr5? The "main vocal" extraction thing ...
You can try using Mel Karaoke or UVR BVE V2 on either MVSEP or UVR Online
gilded frigate
alr
rare gobletBOT
Ayo? @gilded frigate level 2 !!!
gilded frigate
You can try using Mel Karaoke or UVR BVE V2 on either MVSEP or UVR Online
are those exclusive to the websites or smth?
odd shale
are those exclusive to the websites or smth?
Yep, these are exclusive from these websites.
gilded frigate
im doing everything locally so i dont have to deal with those queue's and stuff and now you just get cock blocked by a paywall again smh
what a pain
brave garnetBOT
⠀
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
⠀
Made with love by Ilaria 💖
terse turret
so i was doing a resume of a training and this appeared, does anyone know how to fix it?
pastel oak
You are trying to remove background noise from a normal song? Can you even hear "noise" when the beat is playing
If its vocals, BS Roformer removes background noise and instrumentals. You can use it on mvsep or uvr
Make an account which is for free to skip like 95% of the queues and get to yours asap
fathom geyser
Hi. How do I upload my model to Huggingface? I'm using RVC Disconnected and I don't know how to use it. Sorry for the dumb question, but I'm new to AI Covers. How does this work?
upper scarab
Hello im new in this what can i use to test a model that i train in local ? i was trying to use \Mangio-RVC-v23.7.0 but im having an error when i open de ui
leaden latch
I notice my RVC models not being able to pronounce certain adlibs like grah and brrt is it impossible for rvc to do these sounds or do I need more adlibs in my dataset should I remove the adlibs from my dataset if rvc cant do them?
brittle wing
whats the rvc that u can use in ur browser link?
low shard
Hello im new in this what can i use to test a model that i train in local ? i wa...
mangio is kinda outdated, what's ur pc gpu?
low shard
Hi. How do I upload my model to Huggingface? I'm using RVC Disconnected and I do...
you can just upload it urself https://docs.ai-hub.wtf/essentials/voice-models/#uploading-to-hugging-face
low shard
whats the rvc that u can use in ur browser link?
every interface is in the browser lol
upper scarab
mangio is kinda outdated, what's ur pc gpu?
i was using an amd r5 430 but i think its no compatible cuz too old
low shard
i was using an amd r5 430 but i think its no compatible cuz too old
i searched online and it seems to have 2gb vram.. Idek how even you are on your browser rn
upper scarab
☠️
rare gobletBOT
Ayo? @upper scarab level 1 !!!
low shard
☠️
You can't do it locally (on ur pc)
u need to use cloud (remote good pc) computing services
what are u lookinn for
upper scarab
what are u lookinn for
for now im trying to do an AI cover
low shard
for now im trying to do an AI cover
use ilaria rvc zero, zerogpu (A100) huggingface space
brave garnetBOT
Ilaria RVC: CLICK HERE 🤗
Guide on how to use it: CLICK HERE 📝
Don't forget to thank Ilaria if you find it useful! 💖
low shard
fastest free cloud way
flint wadi
any tips to make the vocal sound less like autotune guys? Im trying to make a song cover by using Ilaria RVC (yes i have already separated the vocal and the instrumental)
Index influence -> How much accent is applied: 0.9
Respiration median filtering: 3
Envelope ratio: 0.25
Consonant breath protection: 0.5
rare gobletBOT
Ayo? @flint wadi level 1 !!!
subtle crater
I've had my settings set so the ai voice plays back to myself so I can hear if it's working but now the voice all of the sudden plays like 4 seconds after I talk. It used to be instant
simple ore
any tips to make the vocal sound less like autotune guys? Im trying to make a so...
turn pitch down in the original audio
subtle crater
Nevermind i fixed it
flint wadi
turn pitch down in the original audio
ty, it kinda reduces it a bit
simple ore
ty, it kinda reduces it a bit
sometimes it may help to remove reverb/echo, but mainly the robot voice comes from several vocalists joining in
flint wadi
ah i see
orchid glade
if I train a model is it okay if I just have one wav file or do I cut it into 5 second clips
simple ore
preprocess does the cutting
if youre using applio it also does some basic filtering/normalization
wicked nebula
why when I press start
rare gobletBOT
Ayo? @wicked nebula level 1 !!!
wicked nebula
I don't even sound like the model I'm using
if not, the model doesn't even works
I tried using other models
still not sounding like the model I'm using
orchid glade
what's the best pre trains for an english female voice
lusty dove
extra
orchid glade
I'm using applio atm but there's a storm if my pc goes out while it's still not finished do i restart from the beginning
ornate gulch
hi
orchid glade
I got this what do
low shard
I got this what do
When you finish your Free Daily GPU on Google Colab, you can:
- Use an alt google account
- Use kaggle which gives more gpu time but its harder
- Wait until tomorrow
- Pay for colab pro
just use kaggle
what’s ur pc gpu btw?
orchid glade
4070
should I just do it locally
how do I save my progress tho it timed out at epoch 150
is there anyway to save this
low shard
should I just do it locally
yea ofc
why do it cloud when u got a good enough pc to do it locally
orchid glade
most of my tries are pretty bad was wondering if it would be better on a different gpu
low shard
how do I save my progress tho it timed out at epoch 150
i think u can’t if the gou time finished, unless u see got the G & D files in the google drive, in that case u could use those as “pretrain” to continue the training
orchid glade
ah I guess I have to start over
low shard
most of my tries are pretty bad was wondering if it would be better on a differe...
that depends by ur dataset and if ur using the tensorboard
orchid glade
what does using a tensorboard does it tell you when to stop training or what epoch is the best
that's what I remembered at least maybe I'm wrong
low shard
what does using a tensorboard does it tell you when to stop training or what epo...
it tells u which epoch to use
like helps u to not under/over train
knotty moth
I got this what do
the reason why you should use your 4070 locally rather than that
orchid glade
do you reccomend I cut my audio into 5 second clips then if I train them locally
or does not matter
I'll just give it another go tomorrow I suppose
any advice how to make a good data set
knotty moth
do you reccomend I cut my audio into 5 second clips then if I train them locally
use this method rather than cutting inbetween several sentences which compromises quality with popping artifacts
https://rentry.co/RVC-dataset-RX11#noise-gating-and-audio-labeling
covert aurora
My models sound great for short durations but feel incredibly monotonous after 1 minute because it doesn’t fully clone over expression and emotion of source sound. Anyone know if this is a limitation of RVC or is this just that I need more expressive training data?
spiral cape
I am trying to train my own AI model, but the tool Im using got too old to do its work. Is there tool I can use to make my own AI voice model?
low shard
I am trying to train my own AI model, but the tool Im using got too old to do it...
I'm guessing you want a speech to speech ai model, so RVC
What's your PC GPU?
rare gobletBOT
Ayo? @spiral cape level 1 !!!
spiral cape
as I was making, I got
WARNING: Ignoring version 2.0.6 of omegaconf since it has invalid metadata:
over and over, and I cant find way to figure this out
low shard
#1159290752195633273 or else ur msg gets deleted
low shard
im using macbook right now
Oh yeah nvm, that's not good enough to train (make models) locally (on ur PC) you can only inference (use models)
low shard
as I was making, I got
WARNING: Ignoring version 2.0.6 of omegaconf since it ha...
Does it work after the warning?
low shard
Oh yeah nvm, that's not good enough to train (make models) locally (on ur PC) yo...
As you dont got a good PC, its better you use cloud (remote good pc) for training an RVC Voice Model:
- Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
- Applio (ui)
- Mainline (UI)
- RVCDISCONNECTED (no ui)
- Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
- Mainline (UI)
- Applio by Vidal (UI)
- Applio by Shirou (UI, no guide as of right now)
- Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
It should be just a warning and u should be all good, but I'd suggest u use kaggle mainline instead
spiral cape
Im using google colabs for AI, and does it mean that there will be no problem making the voice model?
Do i just have to run anyway?
my bad, seems like this code doesnt work anymore.
I see github sites exposed already inside.
white bough
Is anyone else having problems with the Applio Google Colab (no UI) and Pytorch? I can't seem to make it work...
low shard
Im using google colabs for AI, and does it mean that there will be no problem ma...
It would be better u use kaggle
Google colab is easier but gives only 4 hours daily of gpu
Around 4 hours, it's not even granted and u could get only an hour
While Kaggle is bit harder but gives u 30 hours weekly granted
low shard
Is anyone else having problems with the Applio Google Colab (no UI) and Pytorch?...
@nocturne mural seems like people getting errors in the applio no UI colab btw
Already 2nd report
simple ore
it likely installs torch 2.5
should be fixed with a new clone
Vidal did a PR last night
distant jasper
-rt
azure marshBOT
-rt
This interaction has expired, use the command
/guidesif you wish to see it again.
distant jasper
> This interaction has expired, use the command `/guides` if you wish to see it ...
.
rare gobletBOT
Ayo? @distant jasper level 2 !!!
sand ermine
gilded mortar
im just looking to understand some basic terminology. is this something i could use to make an AI that will mimic a character's voice based on voice clips?
if so, how do i get started? and if not, what would that be called generally?
latent kettle
im just looking to understand some basic terminology. is this something i could ...
It can convert any voice to your favorite character's voice
latent kettle
if so, how do i get started? and if not, what would that be called generally?
Your have to use RVC for it. If you have a decent GPU use it locally if not use cloud services such as kaggle and colab
gilded mortar
hmmm i see, thank you... i assume its compatible with text to speech in some way?
latent kettle
hmmm i see, thank you... i assume its compatible with text to speech in some way...
You can use APPLIO fork for TTS
gilded mortar
alright then! so to get started i would need to install an RVC program and find or create the voice model i want to use to go with it...?
rare gobletBOT
Ayo? @gilded mortar level 1 !!!
gilded mortar
im very curious about the actual process of creating a voice model and what that involves
im also curious what an epoch is
latent kettle
im also curious what an epoch is
Epoch is a training cycle
If you read a book for 1 time that means 1 epoch.. more times you read more information will you get from it. Similarly If Ai algorithms study your dataset for 1 times it's 1 epoch.
-rvc
gilded mortar
i see. im familiar with the principles of ai training but i guess not all the terminology
latent kettle
All you need @gilded mortar
latent kettle
i see. im familiar with the principles of ai training but i guess not all the te...
So here is a guide for you to do all this stuff. Read it.
gilded mortar
thank you very much!
brittle wing
i need help with rvc , i keep getting an error can someone help me?
latent kettle
thank you very much!
Yw ; )
latent kettle
i need help with rvc , i keep getting an error can someone help me?
Show me the error
brittle wing
do i copy paste what came up in cmd
latent kettle
do i copy paste what came up in cmd
Yes. Or level up yourself to send pictures
brittle wing
what shall i do
brittle wing
Yes. Or level up yourself to send pictures
Traceback (most recent call last):
File "C:\Users\Daanish\Desktop\RVC1006Nvidia\runtime\lib\site-packages\gradio\routes.py", line 321, in run_predict
output = await app.blocks.process_api(
File "C:\Users\Daanish\Desktop\RVC1006Nvidia\runtime\lib\site-packages\gradio\blocks.py", line 1007, in process_api
data = self.postprocess_data(fn_index, result["prediction"], state)
File "C:\Users\Daanish\Desktop\RVC1006Nvidia\runtime\lib\site-packages\gradio\blocks.py", line 953, in postprocess_data
prediction_value = block.postprocess(prediction_value)
File "C:\Users\Daanish\Desktop\RVC1006Nvidia\runtime\lib\site-packages\gradio\components.py", line 2076, in postprocess
processing_utils.audio_to_file(sample_rate, data, file.name)
File "C:\Users\Daanish\Desktop\RVC1006Nvidia\runtime\lib\site-packages\gradio\processing_utils.py", line 206, in audio_to_file
data = convert_to_16_bit_wav(data)
File "C:\Users\Daanish\Desktop\RVC1006Nvidia\runtime\lib\site-packages\gradio\processing_utils.py", line 219, in convert_to_16_bit_wav
if data.dtype in [np.float64, np.float32, np.float16]:
AttributeError: 'NoneType' object has no attribute 'dtype'
rare gobletBOT
Ayo? @brittle wing level 1 !!!
latent kettle
guys when i start the programm it has all the time whitescreen
Which program ?
brittle wing
w-okada
brittle wing
Which program ?
w-okada
brittle wing
Which version are you using? And which GPU do you have
windows 17b
azure marshBOT
This interaction has expired, use the command
/guidesif you wish to see it again.