#✨│ai-help
1 messages · Page 48 of 1
The percentage does not mean much. When you start the process, you get to decide after how many steps you record the data for plotting/printing. If you choose this number so that it happens exactly after each epoch (a full scan of all the data), then the percentage will be 0%, because it is printed at the beginning of each epoch. If not, then you are printing it in the middle of an epoch. And this percentage tells you where you are, when you print it. To be honest, it does not matter that much.
ah ok
Anyone else getting Keyboard Interrupt errors on Kaggle and then it just stops the whole run?
hi whats the best rvc software right now for making covers? I have a 3060
im using mangio rvc but i don't know if its the best one i can currently use
Yeah Mangio seems about right
Where did you make the model?
In Colab?
How to install Tensorboard
https://docs.aihub.wtf/guide-to-create-a-model/tensorboard-rvc
More info on Tensorboard and extra training tips (until UVR section)
https://rentry.org/RVC_making-models
no no i made it on my beefy gaming pc w/ applio, i figured it out and saved a few voices as zips
Is it in the weights or logs folder?
i just went through the training thing and set the epochs as the finished amount so it doesnt actually do anything and then i saved the voice
i never actually saved any of my voices because i just mess w them on there
@true vortex @proper shale let's settle this once and for all
Does batch size affect quality or nah
Hey, is this outdated? https://discord.com/channels/1159260121998827560/1182344731800387585
Yeah!! Replace them for these ones please:
https://rentry.org/ilarvc_inf_guide (eng guide)
https://rentry.org/ilarvc_inf_guide_es (esp guide)
Ok, thanks for letting me know!
Full AI Voice Model Training Guide (Local)
Link: YouTube
Credits: Christopher Villanueva
AICoverGen Colab Guide
Link: Google Docs
Credits: Eddy (Spanish Helper)
How to Make AI Covers using Ilaria RVC
Link: Documentation
Credits: Nick088, Kanav
Create a model with RVC disconnected (colab)
Link: Google Docs
Credits: Angetyde
How to make an AI cover using Hugging Face 🤗
Link: Rentry
Credits: 👽 Julia (ailen2091)
@quasi dagger you'll also have to change the bottom one with the same link too. Use the eng guide
If you can
__<
I actually copied and pasted from the guide channel xD
Ah
no.
batch size affects how much of data per iteration is going to be used for training
this only affects performance as higher batches take less time to process completely back and forth
Yeah that makes sense
i literally have no idea where you guys got an idea that batch size affects the quality
😭 ask @proper shale
@molten pecan
Yes?
Yes?
I need help with this website: https://rentry.org/ilarvc_inf_guide.
what is ilaria rvc, can anyone explain? idk if someone trolled me
can i upload an mp3 file and turn its voice into a text to speech with it?
or not?
its for using rvc on huggingface spaces
Ilaria RVC is a fork of RVC hosted in Hugging Face
Hugging Face is an open-source online platform for hosting interactive AI apps, models and datasets. It's free and of unlimited use
does it do what i just asked
What do I do?
Yes
but it doesnt
😰😰😰😰
what do you mean?
it shows the same voice no matter what i do
All the instructions are there 😭
did u upload a voice model 
no
theres a part that it says ''model'' and its empty
Wait you want to use the TTS? Because TTS doesn't convert audio files, only text
u need to put in a voice model to use it
doesnt it take samples from an mp3
i dont get it
so it doesnt do what i want, it uses the ''models'' that already exis
t
or something like that
and the place where we upload an mp3 file is for no reason
and theres lots of buttons for some reason
the mp3 is to convert audio files into the voice you want
the buttons are for fine tuning and other things
what do you mean
so like
someone speaking
and then it changes to voice
to the person you want
the tts does the same thing
thats how rvc works
what does rvc mean exactly
the tts is just using a tts model then converting it into rvc
Retrieval-based-Voice-Conversion
all of this is so complicated, i just want to use a text to speech thingy where i can upload an audio and make it talk, and make it have emotions
okay so thats not what i want i guess because i never heard that
what are differences between rvc and tts
tts creates audio from text
by emotions i mean like the ones on clipchamp
rvc converts audio into another voice
what are you trying?
RVC (Retrieval-based Voice Conversion)
Is an advanced AI tool to modify a voice to sound like another voice. This can be done either through manually inferencing audio files or real-time, as you speak.
Text-To-Speech (TTS)
A technology that converts text into human speech as an audio.
there are cartoon characters that i want to make them say what i want, in the way how i want, yelling, whispering, angry, etc
you could use a tts that says the things you want and then use rvc to make it the voice you want
Exactly
then that mod just told me about rvc for no reason
😭😭
how will the text to speech work without a voice
The TTS already have built-in models
are you guys coders, is that what youre called
But they can't make yelling and anger noises though
Text to speech without a what? 
i have seen it before though
It's TEXT to speech for a reason. Not audio to speech
you can
clipchamp has ai that has different emotions to chose
Bark moment
Really? how?
yep
Bark tts but its kinda low quality
what is bark
Ask wendy
Another different text to speech software
Wendy knows a shit ton about TTS. I'm sure she can help you
Ok tell me what you want
I will explain very very VERY slowly
Good luck...
Did you try searching a tutorial on YouTube?
i want to make characters talk using text to speech, and make them talk how i want, raising voices, whispering, sad, happy etc
not yet
...there are plenty of tutorials on YouTube
Ok do you want to make them what they say? Like example for this character in an angry tone say "I HATE THIS WORLD"
what do i search to find these tutorials, i dont wanna watch something irrelevant accidentally
yeah or make them sigh sometimes
Wait does ila rvc tts use bark
or make them say ''heeey whatsup'' instead of ''hey whatsup'' like, i wanna make them lengthen the consonants and vowels
No clue. @half cove does it?
i think it uses edge tts
it also has an option for googles and for elevenlabs api
@half cove can edge tts do this
i dont think so but i can check
how did you guys learn about these stuff at the first place
where cani put rvc models for a song
did you not ask stuff like me
how long did it took to learn
Its not that hard once you get the hang of it
really
it looks really complicated
because i want them to talk in emotions you know
and it already looks complicated without that
How can I fix the delay?
Yeah thats the complicated part
Ayo? @brittle wing level 1 !!! 
nt think so
dang
oh whoops
Yeah its currently not possible for tts to have longer consonants
how exactly did you guys start, how did you learn all these inside slang
where can i use the rvc modals anyone got the co;ab link?
You might have to use your own voice and record what you want the character to say
where can i use the rvc modals anyone got the colab link?
you talk in secret words that i dont even understand
Ayo? @brittle wing level 3 !!! 
but then it wont be that characters voice
where can i use the rvc modals anyone got the colab link?
You can input your audio to RVC to make your voice sounds like the character
Using your voice gets you more control than tts
answer someone lol
Ayo? @fickle dust level 1 !!! 
okay so what do i need, tts or rvc?
-colab
Suggestions for @fickle dust
- 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
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HF, by r3gf Huggingface Spaces
- AICoverGen, by r3gf Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
- Advanced RVC Inference, by neuclya Huggingface Spaces
1-2-3
the one modded by hina?
Pick ai cover gen
the ones by google colabs?
Yes
Rvc
thanks cause i wana make a personal song with the model
Do you have rvc installed?
so i will just change my voice? does it still count as ai?
Yeah just upload the audios
Yup
its all in spanish though
Do youhave rvc installed?
no
Pick the english one ofc 
-rvc Read the guides
Suggestion for @brittle wing
Full AI Voice Model Training Guide (Local)
Link: YouTube
Credits: Christopher Villanueva
AICoverGen Colab Guide
Link: Google Docs
Credits: Eddy (Spanish Helper)
How to Make AI Covers using Ilaria RVC
Link: Documentation
Credits: Nick088, Kanav
Create a model with RVC disconnected (colab)
Link: Google Docs
Credits: Angetyde
How to make an AI cover using Hugging Face 🤗
Link: Rentry
Credits: 👽 Julia (ailen2091)
dont i need to upload the speech of a character for the ai to know what is it
Oh you want to train?
okay so i will read all these stuff?
what is train
Training the ai to mimic the character sound
Nope
You know what step aside tts we dont need it
You are going to use your voice
okay
Model training
Is the process where a machine learning algorithm learns patterns from data. This involves feeding the model a set of data known as the training dataset, which contains examples of the inputs the model is expected to handle, along with the correct outputs (RVC already knows inputs for training because of pretrains).
The goal of model training is to allow the model to learn from this data so that it can make accurate predictions or decisions when it encounters new, similar data.
The quality and quantity of data used for training significantly impacts the model’s performance. The better the training data, the better the model’s ability to make accurate predictions.
In the context of RVC or AI covers, you make a voice model based on a certain person or sound, and when you input an audio, the goal will be the output sounding like the model.
A comprehensive training guide for AI voice models. Goes through dataset creation & vocal isolation, training setup, Tensorboard, and vocal inferencing
Further Reading & Downloads:
Vocal Isolation - mvsep.com/en/home
Mangio RVC Installation - youtu.be/ixB9oalT3cQ?si=wLMTnFOqABQIeLBj&t=79
Tensorboard Installation - us.aihubfrance.fr/guide-to-...
That tutorial should help
Watch the vid
And follow what it suggested
ok
@brittle wing you dont need to train if the voice model is already made by someone
check weights.gg
is it a file or a number
Check on the website.gg or #1175430844685484042
Search by typing the name
Never trust a person with a wojak pfp
so i shouldnt watch the video if its been made before?
Yeah
u need login and download the model
@finite galleon keep helping em
bro ditched us
yea
theres also a save button
saving on a collection
idk what that means ig its like youtube playlists
but its stuck
oh wait its not
i guess i need to create a name for the ''playlist''
wat
THIS PERSON IS LYING!
i cant create a list
THIS PERSON IS LYING!
well ig i will just download them all
im not
i know theyre annoying
Lmfao
is it normal that its a 100mb compressed file
a zip file
so how do i use it
i downloaded it
now upload it into a google drive and put it in ila rvc
whats the next step, i downloaded it
unless you wanna do it on your own pc
im gonna do it on pc
Ayo? @brittle wing level 4 !!! 
then u gotta install rvc
is that an app
i thought it was just short for 3 words
like
text to speech being called tts
so rvc is a program?
a spefic one program?
its a program
tts is a name for things that use text to speech
rvc is the name of the program
YES
IT AFFECTS TRAINING QUALITY
what does it mean again
so i gotta search rvc on google?
it will show up?
okay im doing it
-links
- Applio-RVC-Fork, by IA Hispano GitHub
- Mangio-RVC-Fork, by Mangio621 Huggingface
- RVC Studio, by SayanoAI Huggingface
- AICoverGen, by SociallyIneptWeeb GitHub
- Replay, by Replay Team Website
- Original RVC, by the RVC-Project team Huggingface
You can find more info on the #1159513888199540817 channel. If you can't find your answer, feel free to ask for help in #✨│ai-help. Credits to Faze Masta and Antasma for compiling these links.
@brittle wing
Not exactly model quality but the quality of training itself
yea i was so confused 😭
What is bro sayin
All I'm saying is, there's a best batch size for each model, and certain models can flatline way earlier due to it
-rvc
Full AI Voice Model Training Guide (Local)
Link: YouTube
Credits: Christopher Villanueva
AICoverGen Colab Guide
Link: Google Docs
Credits: Eddy (Spanish Helper)
How to Make AI Covers using Ilaria RVC
Link: Documentation
Credits: Nick088, Kanav
Create a model with RVC disconnected (colab)
Link: Google Docs
Credits: Angetyde
How to make an AI cover using Hugging Face 🤗
Link: Rentry
Credits: 👽 Julia (ailen2091)
@true vortex how?
_ _
i think we are getting mixed up with model quality and training quality
@proven hill could you please change the bottom link for the ila rvc guide? It's outdated.
https://rentry.org/ilarvc_inf_guide
it affects performance only

you can view torch's source code and see exactly what batch size is responsible for
RVC Guides (How to Make AI Cover)
Translation by country
i like to use the example of baking a cake, where you can set a high temperature and get the cake earlier, but it is worse than a cake that was baked at a lower temperature. that's because it was cooking for longer.
either way, id just recommend using different batch sizes in every model
its a performance setting, but it does affect quality
Hey just want a quick answer, which local RVC build/fork the best currently for voice changing
have you actually read this yourself? the difference is so small that it doesn't matter on rvc due to the model complexity
nobody would even be able to run rvc with more than 20 batches
which one should i click
either way, batch size still affects the training
difference in batch size 1 and batch size 20 is literally invisible
if u want real time stuff, W-Okada
if you want inferencing, maybe mangio fork
but telling people it affects the model's quality is overall just misleading
okay fair point
I'd just suggest people using different batch sizes
it does, but literally no audiophile would be able to hear the difference
It can make better models in the end
i mean applio has bark preinstalled so you could try that
i mean based on what people are saying, installing applio is a journey
what is a fork
a software based on a technology, adds features n stuff
can't think of a better explanation
😭
lol
that didnt explain anything
so its like what links are called?
or something
but does that have bark 😭
or theyre modes or something
they're pieces of software based on a certain technology with a new gui n stuff
the flavors to the vanilla
bro just use ila tts or colab or elevenlabs
you don't need built in bark
he wants emotion and shi
eleven labs could work ig
which fork sounds the best for voice changing
yea ElevenLabs ftw
all of em are the same
i mean does the best
just different gui and sometimes features

ig mangio gets the job done
damn
im sticking with my mangio it is
thats not free
eleven labs wants subscribtion and stuff
it is free but limited quota
you could use bark then just put the audio into rvc easy
nah its straight up not free
what?
no
then what
you use text to speech
then you put the audio into rvc
it changes it into the voice you want
oh
why tho
why do i need eleven labs at the first place
oh wait it has emotions
to make it say what you want
right?
it has emotions?
yea i think
should i select from its own voices
or should i add one
i think adding one was not free
right?
Error: Failed to load logs: Not Found. Logs are persisted for 30 days after the Space stops running.
why?>
ilaria rvc
they do this to save computational space, just restart the space
how do i restart
yeah
how do i restart space
meatball icon (aka 3 vertical circles) on the top right?
Error: Failed to load logs: Not Found. Logs are persisted for 30 days after the Space stops running.
still shows that
hmm
wait for it to finish building
yh
does it matter if the voice of the tts doesnt sound similar to the character? should i make it as close as possible so i can change it better?
poopmaster
id just use a voice thats kinda similar for example male if your using a male model
is there a setting for that
whatever that is
on elevenlabs
why is it taking so long
===== Application Startup at 2023-12-17 00:54:00 =====
===== Application Startup at 2023-12-17 00:54:00 =====
===== Application Startup at 2023-12-17 00:54:00 =====
WHyy
thats on rvc
ok
yeah
bruh im so dumb i was trying to test ila rvc but i was using a space where i was messing with the code 💀
it just takes a while
do labels change anything or theyre just like a description that shows the viewer what its about?
do i have to write things on there, will it change the voice? or it doesnt?
why is nothing happening when i click the start bat file, it runs it like it would and then nothing occurs
its a description i suppose
im downloading the one that says original
im assuming its the most basic one
is it
yea
so its easy to install and easy to use?
i suppose the installation is straightforward, but the gui might be confusing at first
greetings im trying to find a fix for an error that has been preventing me from training voices in rvc 2. Error number 38, i think it's pretty common
what do the logs say?
and uh don't think thats common
what does gui mean? buttons and stuff? is that what is gui?
A graphical user interface, or GUI ( GOO-ee), is a form of user interface that allows users to interact with electronic devices through graphical icons and visual indicators such as secondary notation. In many applications, GUIs are used instead of text-based UIs, which are based on typed command labels or text navigation. GUIs were introduced i...
so ive tried to run the cmd prompt to open and run the start_http.bat file properly and that also instantly closes. Is there some other way that im supposed to be installing the programs files onto my computer
hold on im trying it one more time
does everyone hold themselves from making a waltuh joke or its just me
hey it says i 100% completed the download yet the bar is not fully green
its slowly filling
oh yeah nevermind i downloaded it
Ayo? @brittle wing level 5 !!! 
this is the error that keeps on repeating
i just wanted to rebuid one of my cloned repositories
btw is this a common issue
i tried to rebuild a duplicated space that i had
alr thats fine
How do I fix this?
hi how do i get the voice thing running? im having problems setting it up
It still said the same thing
which one do i click to open rvc
Drag the bat file out?
Bat folder
Whats the rvc link
that’s what you click right?
bat file or bat folder, i cant see a folder named bat
THIS PERSON IS LYING!
does anyone know where to download VAC by Muzychenko the full version for free? I tried the trial and it worked a charm (unlike th VB cable which even with the recommended settings gives me issues), but now its saying "trial" over me when i speak
Thanks! Also i thought since it was real time voice changer it would go on this channel my baddd
Hello, people. I changed my computer recently and I tried to use RVC again for covers, and i get this message
"C:\RVC1006Nvidia>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897
'runtime\python.exe' is not recognized as an internal or external command,
operable program or batch file."
Does anyone know anything about this? Thanks in advance.
@molten pecan hey so, im trying to figure out where is the link for the rvc software and im overwhelmed
that's okay!
im sure im onto something i got the rvc conversation one
Ayo? @wide plinth level 1 !!! 
conversation? you mean conversion?
i cant ask yet xd ime xtracting
THIS PERSON IS LYING!
ye
i dont wanna use voice generator
i want to write the lyrics and cba performing rn
cba?
dunno who that is
that's good
let's talk about this first
when you say "write the lyrics" you mean typing them out and make the AI sing it?
yes correct
alright
is that not a good result?
there is a tool to type out text, but it is for normal speech, not singing
you know what I mean?
which other one?
to my knowledge no version of RVC have a tool to type out lyrics and make it <sing> it. you can only write text and it will convert it into <normal speech>
so do i need to download python to open the file?
probably yeah
so yeah if you want the AI to sing it, you'll need an audio file of someone singing the lyrics
no problem i got alot of samples from locals
alright then
yw. have fun
ykow Julia you should try applying as helper someday :3
yeah helpers tell me the same too >.<
but
can i get papago voice?
apparently I need to know how to make a model first
did you search on weights.gg?
yep
it's a website
you could try searching #1175430844685484042 or use /search on #🔍│find-models
You know whats so annoying? Supporting the latest python version (3.12) for rvc. I had to fork rvc, update pytorch to a nightly version, update scipy, update numpy, update numba to git+https, etc. Plus I had to update the entire github actions workflow to even support this. And ofc it still doesnt even work - poetry install fails on the CI 💀
Keep in mind the python 3.12 release was already 75 days ago, and yet most python packages are still way behind....
hmm
or even try asking for it :3
k
https://rentry.org/aihubstuff i did this for helpers and the community
I compiled the best guide for each tool + a link to the app
why use 3.12 when the older ones are still used
bookmark it ^^
why when iload start http bat nothing happens
someonehelp pls
id have to go build python 3.11 from source for that
with pyenv
nice
and that means two versions of python, one of which is not the latest anymore
im just annoyed with how out of date the python ecosystem tends to be
i mean.... you don't really need the newest python if all the other apps still use older ones 
newer versions aren't always good. 3.9 - 3.11 already have almost all of what RVC needs, or can u mention what the new 3.12 features could be useful for RVC?
RVC Guides (How to Make AI Cover)
Translation by country
Does that contain mobile?
so after downloading a rvc model how do I actaully use it like assuming im on a collab drive or something etc. also what are the best settings to get highest quality for covers? should you upload vocals unproccesed with some tune and compressor for cleaning up purposes etc?
hello
why i cant hear my sound when im talking
is there any settings to hear my sounds ?
Does anyone know what retrieval rate does
Careful there, 3.11 and 3.12 do not work
You're supposed to use 3.8/3.9/3.10
never tried it, 3.9 is already best working version.
3.11 and 3.12 just crash because python slightly altered one of the language's policies. fairseq (which is used for index files), does some stuff which is no longer allowed in those versions and just throws an error
any help ?
Traceback (most recent call last):
File "infer-web.py", line 27, in <module>
import gradio as gr
File "/opt/homebrew/lib/python3.8/site-packages/gradio/init.py", line 3, in <module>
import gradio.components as components
File "/opt/homebrew/lib/python3.8/site-packages/gradio/components.py", line 32, in <module>
from fastapi import UploadFile
File "/opt/homebrew/lib/python3.8/site-packages/fastapi/init.py", line 7, in <module>
from .applications import FastAPI as FastAPI
File "/opt/homebrew/lib/python3.8/site-packages/fastapi/applications.py", line 15, in <module>
from fastapi import routing
File "/opt/homebrew/lib/python3.8/site-packages/fastapi/routing.py", line 22, in <module>
from fastapi import params
File "/opt/homebrew/lib/python3.8/site-packages/fastapi/params.py", line 4, in <module>
from pydantic.fields import FieldInfo, Undefined
ImportError: cannot import name 'Undefined' from 'pydantic.fields' (/opt/homebrew/lib/python3.8/site-packages/pydantic/fields.py)
How'd you install it ?
I am trying to run it on Mac M1.. it worked a while ago but now this error appears when i "sh run.sh"
That's weird. did you ever move the folder around by any chance ?
It's looking for packages which I wouldn't expect run.sh to be using
maybe.. but i also re-downloaded the original folder from github and i get the same issue
Ayo? @brittle wing level 2 !!! 
also "pydantic" is installed already
Thing is, run.sh should be creating a virtual python environement
and not use system-wide python packages
any solution or command to properly run the webgui ?
gradio i mean
also what does this mean WARNING: The directory '/Users/kai/Library/Caches/pip' or its parent directory is not owned or is not writable by the current user. The cache has been disabled. Check the permissions and owner of that directory. If executing pip with sudo, you should use sudo's -H flag.
That's... interesting
why is it so static and does echo like thing in the background
you using realtime voice changer ?
and it never comes out as perfect as in videos
?
i use it yeah
I don't really get what's going on on your setup to be frank, it does a lot of stuff the way it's not supposed to do it
Mind moving over to #🔍│help-w-okada ?
it is stuttery with too much static
results vary on the models you use btw, and also your microphone's quality
but im using rvc modles
or it is the voice changer that is okada
Yeah, though there's a dedicated channel for realtime there
yeah, people call the voice changer "w-okada" because that's the maker's github name
oh
got it
trying to run Mangio RVC
its the same as on the github
not altered
Oh mangio, I haven't checked that in a while
also is pytorch neccesary to be installed
i got python only
Ayo? @wide drift level 1 !!! 
for voice-changer everything is bundled
There's always the original project
assuming python 3.8/3.9/3.10 is the default version on your system, it should work
just download the repository's code as zip, extract somewhere, and launch run.sh
Note that this is important, else it breaks
oh there's an english readme
i dont really get it
do i have to install pytorch?
Nope !
Only unzip & have fun
https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/en/README.en.md
My bad on that one x) Should've linked the english readme
ah well i got a nvidia gtx 1650
which version should i dl
like gu or direct ml
cuz i downloaded direct ml and it is very stuttery and static
get the one that doesn't mention "directml"
or refer to https://rentry.co/VoiceChangerGuide
This guide is written by: Raven
Please give credit if you use elsewhere, thank you!
Other Links
Antasma's Local Error Fixes
Antasma's Colab guide
Sushi's useful Links - You need to follow this if you are on an AMD or INTEL ARC graphics card
Frequently Asked Questions
Jump To The Important Section...
XD
thanks.. installation now. But the original doesn't have rmvpe+ right?
It does nowadays
ok cool
Mangio used to be relevant, but now pretty much everything it had is available elsewhere
i wish applio would work for mac...
downloaded the latest version of the base rvc fork and i have no idea where the hell my models are being saved
i realised i have to redownload it cuz i downloaded directml 
there was a weights folder with all of them in older releases but now i have no idea where tf they are
I think it's under /logs/<modelname>
Could be wrong
where do i put my model in the RVC original? there is only a logs folder.. now weights folder
all thats in here is discriminator and generator pth files tho, and its generating a lot of them for some reason
every 60 or so steps it seems
/assets/weights/
just checked
sorry about the earlier misinfo
ahh ok
/assets/weights/
thx i found it
installation takes ages
Yeah unfortunately
hmm
everytime i try to train a voice model it always stops training by 150 epochs
keep getting this error
space in filename most likely doesn't help
"new voice training version v3" is gonna cause issues
oh i forgot about that
ight now its stopping at epoch 4
still installing
gonna try clearing storage
What's it stuck at ?
It's downloading a lot of stuff, and even building locally
But unless your network's having trouble keeping up, it should be done by now
Ayo? @brittle wing level 3 !!! 
should be ok then
its eating space :/
how much?
👍
yayy
does this run with M1 chip or is it CPU?
20 seconds audio takes like 10 seconds to inference
i think it uses the iGPU yeah
also its only RMVPE.. no RMVPE+
you aren't really missing out on much tbh, rvmpe+ just allows you to limit the frequencies that will be inferenced, not a big big thing
how do i find out if its run with GPU ?
i mean, 20 sec audio taking 10 secs is already a sign its running with gpu, cpu would take 40-50 secs
- pytorch (the thing that RVC uses for a lot of stuff) has support for GPU acceleration now
whenever i set my output device and press start i hear a static voice from the model ( model voice saying ahhhhhhhhhh) it get louder when i increase out meter
and their voice isnt clear and feels muffled or quite silenced
that's your background noise
ok
weird
how do i apply a noise suppresion
it might be the pc cpu noise
if you reeeally aren't sure though theres a "task manager" on Mac, called "Activity Monitor", i think it tells you how much GPU certain stuff is using
sup2
that is a program?
Ayo? @wide drift level 2 !!! 
it's in the GUI, sup2, check it
Guys
I have 6 minutes and 40 seconds dataset
How many epochs should I do?
400 or 500?
great thank you
but at times when im talking he suddenly keeps repeating what i say non stop
Activate Echo and increase a lil bit of S Thresh
i maxed it
Thanks
thresh*
i maxed thresh out but he doesnt catch up to what im saying
like i say a word he says the second half of it
is that the chunck ms?
i even set it low so he speaks faster
maybe decrease it a bit then
he muffels some of my words
like he says them but they arent clear as if talks underwater
increasing chunck fixes it kinda
but the delay is un believable
112/128 should be fine
yeah i tried it it is almost perfect but there is small very small stutter that has like a robotic sound
it exposes that its ai
i will work on it
using others settings turned out a bad idea
since it varies from mic to another
man this model was training 30s/epoch yesterday and today when resuming it's training 1m/epoch and all of the settings are identical
im trying to train a model in google docs and when i click training this error keeps repeating
anyone knows what's wrong?
these are my settings
RVC Guides (How to Make AI Cover)
Translation by country
-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
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HF, by r3gf Huggingface Spaces
- AICoverGen, by r3gf Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
- Advanced RVC Inference, by neuclya Huggingface Spaces
Ayo? @old oracle level 1 !!! 
Hi, can someone send me a guide for creating new voices locally using my pc hardware (I don't want to use Google Colab)
-rvc
Suggestion for @timber terrace
Full AI Voice Model Training Guide (Local)
Link: YouTube
Credits: Christopher Villanueva
AICoverGen Colab Guide
Link: Google Docs
Credits: Eddy (Spanish Helper)
How to Make AI Covers using Ilaria RVC
Link: Documentation
Credits: Nick088, Kanav
Create a model with RVC disconnected (colab)
Link: Google Docs
Credits: Angetyde
How to make an AI cover using Hugging Face 🤗
Link: Rentry
Credits: 👽 Julia (ailen2091)
Thank you, I will ask for help if I encounter something I cannot solve by myself
sure, just ask here!
Hey @proven hill could you please change the bottom link for the updated guide? https://rentry.org/ilarvc_inf_guide
not my bot
Who should I ask?
should be chthulu
@robust needle them?
yup
Gotcha ty
do i check the total loss for the g model or the d model??
-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
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HF, by r3gf Huggingface Spaces
- AICoverGen, by r3gf Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
- Advanced RVC Inference, by neuclya Huggingface Spaces
📈 Tensorboard info and training tips 💾
https://rentry.org/RVC_making-models
THIS WAS ALL TESTED ON THE Mangio-RVC-Fork WITH THE F-RVC-exp CONTINUATION FORK OF Mangio-RVC-Fork
(From what I know, each fork logs differently for TensorBoard)
Things you may need for this guide:
Audacity
iZotope RX Audio Editor
Adobe Audition
Spek
TL;DR is at the bottom of the page
Mangio logs...
Good luck!
Those are extra tips if you are interested
how tf do i make my chromatic on kits.ai sound good with the official bf chromatic scale?
hello everyone, can you help me with Voice Changer Client Demo? when I say something, vol does not change, and when I select passthru, everything is fine, but just when I use the program, my vol slider does not change, please help
sorry but we don't recommend kits.ai and most of us don't bring support to it
it's recommended to train using RVC either locally with a decent 8 GB+ GPU or in some colab fork, with much more settings and customization, and also good support here
i think you'll have to ask somewhere outside
have you heard pekora's eepy voice when she wakes up
guys
Ayo? @strong flax level 1 !!! 
please
Ayo? @brittle wing level 2 !!! 
help idk how to get voice modulation
wdym?
idk how
Ayo? @pulsar coral level 1 !!! 
u looking for a voice model? voice changer?
voice modulation is that one google drive link some random youtuber shared
you're looking for https://github.com/w-okada/voice-changer
guide is found at https://rentry.co/VoiceChangerGuide
This guide is written by: Raven
Please give credit if you use elsewhere, thank you!
Other Links
Antasma's Local Error Fixes
Antasma's Colab guide
Sushi's useful Links - You need to follow this if you are on an AMD or INTEL ARC graphics card
Frequently Asked Questions
Jump To The Important Section...
im doing shit today
Nah it's alright lol
-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
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HF, by r3gf Huggingface Spaces
- AICoverGen, by r3gf Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
- Advanced RVC Inference, by neuclya Huggingface Spaces
any efficient way to determine the amount of time it's going to train epochs based on the dataset length?
im getting a traceback error on step 2b
If Colab disconnects at any point before the training process (happened while I was extracting a dataset's features after pre-processing the data I think) do I have to restart the process all over again?
WHY THE HELL IS MY MANGIO RVC NOT WORKING WHEN I TRY TO CHOOSE MODEL OR DROP AUDIO FILE HELP
THE BUTTONS DONT WORK
Ayo? @simple raft level 1 !!! 
JUST HELP ME SAY SOMETHING
RVC Guides (How to Make AI Cover)
Translation by country
AH THANK YOU BOT I FINALLY MADE A AI COVER
guys why iam having echo in the AI voice app?
you mean w-okada realtime? then the monitor option should be set to your headphones, not the speakers that could be caught by the mic
i fixed it thanks
-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
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HF, by r3gf Huggingface Spaces
- AICoverGen, by r3gf Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
- Advanced RVC Inference, by neuclya Huggingface Spaces
what rvc collab should I use as a beginner in using ai voices to make ai covers?
ilaria rvc should be good
check the #1159513888199540817 in case you're confused
Thank you so much 
you're welcome :)
Ayo? @proper shale level 58 !!! 
lmk if you have any issues with it
-local
- Applio-RVC-Fork, by IA Hispano GitHub
- Mangio-RVC-Fork, by Mangio621 Huggingface
- RVC Studio, by SayanoAI Huggingface
- AICoverGen, by SociallyIneptWeeb GitHub
- Replay, by Replay Team Website
- Original RVC, by the RVC-Project team Huggingface
You can find more info on the #1159513888199540817 channel. If you can't find your answer, feel free to ask for help in #✨│ai-help. Credits to Faze Masta and Antasma for compiling these links.
I made the ilaria rvc guide !! https://rentry.org/ilarvc_inf_guide
-help
-online
Yo uhm i try to make rvc models with the google collabs, my dataset are 1h longs and i try to make 250 epochs after like 3hours when i reach 100 it stop itself do you know how to fix it ? ty !
完整包 Complete package
For Nvidia GPU users:
https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC1006Nvidia.7z
For AMD/Intel GPU users:
https://huggingface.co/lj1995/VoiceConversionWeb...
regular RVC is more up to date than mangio atm
the only con is the UI is a little worse, but, it's faster, so
my rvc isnt working it says Running with the runtime Python, Please wait.
Where is the voice changer download?
Sup Tech folks, i kinda want to do shit posts (i do blender animations) but i suck at voice acting, kinda thought about taking voice lines from old video games and using AI clones to voice in my dialoge, thing is i hav no clue how to do any of this
Any idea where to find someone who can do that for me or where i can learn said knowledge?
https://www.youtube.com/watch?v=Vpmiq8sTH0k Example
bruh no one helps here
Ayo? @gaunt narwhal level 1 !!! 
Some please help
/colabs
Ayo? @polar crest level 2 !!! 
-colabs
- 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
- RVC Disconnected, by Kit Lemonfoot Google Colab
- easyGUI, by rejects Google Colab
- Ilaria RVC, by thestingerx Hugginface Spaces
- RVC-HF, by r3gf Huggingface Spaces
- AICoverGen, by r3gf Huggingface Spaces
- RVC v2 Huggingface version, by Clebersla Huggingface Spaces
- Advanced RVC Inference, by neuclya Huggingface Spaces
Oh ye the applio colab should also not be working atm cause of that
Guess I'll take care of it
applio issues its gradio right? its already found a tricks?
I have a fix but yesnt
Still uses gradio like the covergen webui colab fix
whatts i are you recode with NoUi style?
oh i see
No.. I make web ui works

what
what are you trying to do ?
what told you you need to install pytorch ?
um i was following a youtube video to how get custom voices and like the video said to download python and pytorch to get other voices, I can send you the link so you can see
Ayo? @solid moon level 1 !!! 
Yeah, don't listen to that guy lol
oo
I've yet to see a youtube tutorial for this that isn't misleading
All you need to do is download the right file for your machine
What's your graphics card ?
um
(you can check under task manager)
how do i use voice models from weights.gg on RVC v2
thank you <3
hang on
i almost found it
dont go yet 😭
erm
is it supposed to be NVDIA GeForce RTX 2070 SUPER??
Yeah!
The appropriate file for it is
https://huggingface.co/wok000/vcclient000/blob/main/MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.17b.zip
-realtime
This interaction has expired, use the command
/guides realtimeif you wish to see it again.
did you execute as admin?
Is this still improving? If not is there a specific point I should take the weight from?
RVC Guides (How to Make AI Cover)
Translation by country
everytime i use the google collab one once ie get to the 2nd partr where u put the audio in it just says error
looks good, keep going
damn really? lol I was expecting someone to say "nah the mode collapses ruined it start from scratch"
im 2k epochs deep
idm collapses anymore, they're annoying but model keeps improving so
will keep it going then
something happened just before 35k, you could try using a checkpoint just before or keep training. Collapse happen, but checkpoints after can still sound fine. Although there is a drop off at some point in terms of how good a model can get which would probably get reached soon.
I did compare a 1k epoch inference and a 2k epoch inference and couldnt tell the difference lol so maybe that point is soon if not now
depends on how long the dataset is. I could never really go much past 300 with 40m dataset
but yeah have to consider steps and batch size as well
~6min
tried 17min and got shit result. not enough clean data so I let meh data through. pruned it to 6min and so far the result is great. miles better than before
it struggles on breathy high notes unfortunately. not sure what to do about that
its probably gonna be as good as its ever gonna be soon
maybe an earlier checkpoint struggles less?
Sometimes it learns bad things as much as the good things when you train a long time
I got 400 weights I can go through to check lol
oh pitching down the original slightly in audacity at the sections where the voice cracks works
Ayo? @serene solstice level 5 !!! 
run as admin?
or extract the folder as admin
this is probably not the solution to the problem
ye
someone please help me
like i need it to get it working
So I started to make a model and right as I start the 'Dependencies' I see this message. Clicked 'Connect without GPU' and got this error later. Don't know what happened.
make new acc
or just wait 24h
ok
What are prediction windows in UVR? I'm going to go with 50 overlap and 1024 Segment size
check a model
would it be still the same layout
Ayo? @gaunt narwhal level 2 !!! 
the other one was simpler😭
btw Mia you said that you do your index after training is done?
i always do mine before
okay good lol
im annoyed rn trying to look through my tensorboard
bc there are no points for so many epochs
like i wanna see the loss of 8k
but theres 7.992k
why not?
this is my mel
its good
but i dont wanna train it more
also with bad datasets it will always sound kinda robotic right?
no matter how much you train it
Mia do you ever make datasets from songs themselves instead of stems?
ah
makes sense
idk how much more I can clean a dataset than what I have
Just kinda karaoke it, dereverb, maybe dereverb twice
thats about it
i dont use uvr i use mvsep bc i dont have a good enough gpu
and I train on paperspace
wym on your own? like what if there is noise over the audio itself?
yes but wym manually clean
and wym rx10
Oh it used to be called aggression settings 🤔
but again what if there is noise while the speaker is speaking
audio editors can do that?
so you could manually create a vocal only stem with it for example? it would take forever, but is it possible?
Ayo? @fallow glacier level 3 !!! 
oml, okay, your own vocal only version of a song
can you do manual cleaning in audacity?
I assume the answer is no
are there videos on how to read spectrogram..?
how do you clean a spectrogram though? isnt it just the same as cleaning a waveform?
i do not 😦
then why is spectrogram so significant if its the same thing
it makes my time to make a cover increase, to the point of taking 40 minutes to make just one
You are a helper, do you know anything about this error?
okay. Example, whats wrong with this one, or what should I be looking for?
Yes, what does that mean?
I'm a little confused, so RVC uses finished models in /weights unlike sovits? also I noticed the file size of the model is much smaller
I understand, thanks for the explanation
Ayo? @modern surge level 4 !!! 
Do you know any helper/mod or member who knows about colab to help me?
What does this mean?
Please I need help
i think you're using the wrong sample rate
Is there a pattern to when RVC prints the loss_disc, loss_gen, loss_fm, loss_mel, and loss_kl line to console? It seems to do it randomly.
What do I do?
change the sample rate
How though
the real fix is to use a wav and not an mp3
not sure I dont run locally
untrue
audacity
How in audacity
paste the audio into audacity
make sure this is set to 44100
in the bottom left
and re export it
I don't see project rate
@fallow glacier
i tried updating RVC using these, and theres still no RMVPE option, am i doing something wrong?
it says in the bottom right that your rate is 44100 so id say just put your audio file in and export it
idk i dont use local but i dont really find rmvpe to be better but ive only used it like once
It worked. Where is it?
where is what
The audio
1_1_Pharrell Williams - Happy (Video)(Vocals)(No Other).wav->Success.
Index not used.
are you running locally..?
Yes
uhhhh idk show me your root folder
check audio-outputs
Ayo? @fallow glacier level 4 !!! 
are you inferencing?
it doesnt auto save afaik. well idk how it works on a command line only version. on the gui verions for me I have to right click the generated audio and save it.
^ yeah there should be a button
try specifying an output folder
it has an option for "Specify Output Folder"
type the location for the audio outputs folder
(shift+right click then copy as path)
and remove the quotes
is it not just outputting to the input folder?
i run on paperspace so this isnt the same, and I use applio RVC
maybe thats a contributor
but it also auto saves mine to the audio outputs folder
I just figured if there's no generated audio in his ui then it might be saving to the input folder by default
oh huh I wonder if it does for me too. I never checked
yeah it says "opt" in the output folder field
yeah i didnt know that either
i randomly figured it out
btw does it do the thing for you when it almost never works when saving the lowest points of a model
wdym