✨│ai-help

[dawn nebula]

yeah

[charred topaz]

I mean the first voice model i created i just created without tensorboard and worked really well

is it necesary?

[dawn nebula]

its not absolutely necessary, but it will be a very good reference on whether your model is overtrained or not

[charred topaz]

UVR extracted vocals with 44100 sample rate, can i just put it in audacity and change it to 48K then export?

because theres only 40k, 48k and 32k in the Train sample options

[half cove]

if its 44.1k then use 40k

[charred topaz]

ok thanks

[brittle wing]

can anyone help me with how to download rvc, ive already gotten python but whenever i open the "start.http.bat" file it just shows this: MMVCServerSIO.exe ^
-p 18888 ^
--https false ^
--content_vec_500 pretrain/checkpoint_best_legacy_500.pt ^
--content_vec_500_onnx pretrain/content_vec_500.onnx ^
--content_vec_500_onnx_on true ^
--hubert_base pretrain/hubert_base.pt ^
--hubert_base_jp pretrain/rinna_hubert_base_jp.pt ^
--hubert_soft pretrain/hubert/hubert-soft-0d54a1f4.pt ^
--nsf_hifigan pretrain/nsf_hifigan/model ^
--crepe_onnx_full pretrain/crepe_onnx_full.onnx ^
--crepe_onnx_tiny pretrain/crepe_onnx_tiny.onnx ^
--rmvpe pretrain/rmvpe.pt ^
--model_dir model_dir ^
--samples samples.json

[hot lagoon]

My puting the sed -i 's/pyworld==0.3.3/pyworld/g' /notebooks/Applio-RVC-Fork/assets/requirements/requirements.txt
but it gives me this

[sand iris]

which platform is that on ? Cloud ?

[hot lagoon]

Paperspace

[sand iris]

requirements is now directly under the Applio folder

not under /asset/... but right at Applio-RVC-Fork/requirements.txt

[hot lagoon]

[twin prawn]

Could someone help me with the google colab link I'm using? The one I'm using doesn't send me the gradio link https://www.youtube.com/watch?v=g1OCqPI7TXc

[young pendant]

hello guys, I have a little question.
Does the dataset have a limit of recorded voice or not?
I had readed the datasets have up to 45 minutes or something, is that true?

[young pendant]

Google colab will not work for voice cloning anymore, that was banned several months ago, you must use RVC offline, either for voice retraining or cloning new ones.

[half cove]

!pip install huggingface_hub==0.19.4

[half cove]

it does work just webui was banned but there are ways of bypassing

[half cove]

having over 45min is fine but i usually recommend around 20 min

just make sure it isnt too much cuz rvc struggles with very long datasets

[young pendant]

I am making Valvatorez's voice dataset, I already reached 8 minutes from episode 1, was so easy. 😉

¡¡¡SARDINES!!! 🐟

[half cove]

if your dataset is long just cut out the best bits

[young pendant]

thanks. 👍

[onyx ferry]

[karmic kite]

wait how do you even do ai covers now isn't the colab dead?

[half cove]

some colabs still work

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• GPT-SoVITS (demo), by litagin Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[half cove]

try ila rvc

[karmic kite]

woah I didn't even know there were more-

@pale nimbus check this out

[karmic kite]

Thank you

[young pendant]

I am feeling nervous, cuz I never trained a voice before, let's hope I don't overtrain it.

[half cove]

use tensorboard

[young pendant]

ok.

[twin prawn]

@young pendant is there a link for that?

[young pendant]

I think both RVC and Applio have Tensorboard, but I don't remember it exactly, I saw it on tutorials.

[half cove]

also click this to save every checkpoint as a model

[young pendant]

that sounds even better than I thought. 😄

[half cove]

applio has it setup already for mainline and mangio you gotta run a bat file

[karmic kite]

Yep it works

Thanks

[half cove]

[pale nimbus]

which is the best one to use?

[karmic kite]

ila works, just install first then start the second thing for the link

[rare goblet]

Ayo? @karmic kite level 1 !!!

[karmic kite]

...what kind of clyde rip off are you

[half cove]

it should also what sample rate of ov2 super and make sure its the same as the pretrain your using also send full error

the 32k one did you download my fix or the main one?

for the 32k use this https://huggingface.co/poiqazwsx/Ov2Super32kfix/tree/main

others should be fine tho

[daring widget]

Hello?

How do I extract high notes and ad libs?

Excuse me……?

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[blissful belfry]

Is there a way to fix it?

[lost dune]

Train model?

[blissful belfry]

Don't use a model to train the sound 🥹

It's easy GUI

[lost dune]

Do you want to use the model?

@blissful belfry

[rare goblet]

Ayo? @lost dune level 1 !!!

[blissful belfry]

Yes, I'm used to making music.

[lost dune]

You can try Hina Mod

[blissful belfry]

But suddenly it just happened like that. I thought it would just be a moment and then it would go away 🥹

[blissful belfry]

How 😲

[lost dune]

https://colab.research.google.com/github/hinabl/AICoverGen-Colab/blob/main/Hina_Mod_AICoverGen_colab.ipynb

It's much easier than easy gui

[blissful belfry]

Thanks, I'll try it out :3

[rare goblet]

Ayo? @blissful belfry level 2 !!!

[lost dune]

Okay

[blissful belfry]

Thank you very much, now I can use it, thank you very much 🥹✨

[hot trail]

How do you use

The models

[dawn nebula]

do u want to infer or use voice changers

[charred topaz]

im trying to create another model but now everytime i train it the epoch load really fast like 1 second after another, my first ever voice model was way slower, any reason?

[strange schooner]

Hi
What would a recommended batch size be for a 30 min dataset?

[narrow compass]

what is the correct link for this download

https://huggingface.co/itt0lp/taylorswift/blob/main/taylorswift.zip

[young pendant]

Idk, but I hope it wont happen to me as well, I just finished my first dataset (20 mins).

[charred topaz]

my first dataset was like 7 minutes long, training was like 1.5 hours, that first voice model works really well
Then every single new voice model i try to create it just train too fast, like a couple of minutes, and the result is really bad

[young pendant]

my first voice I've done right now have 20 minutes, and I have a very potent computer with a RTX 3070, so... I think I may complete it, as long as I do not overtrain it. I will search a youtube video for using tensorboard on RVC. 😉

[charred topaz]

i have a rtx 3080

[covert turret]

https://huggingface.co/itt0lp/taylorswift/resolve/main/taylorswift.zip

[young pendant]

now I saw that you have a problem with cloning voices after the first one. right?
let's hope doesn't happen to me.

[charred topaz]

yeah, first model neede 1.5 hours, after that it loads too fast and I think thats the reason the newer voice models are bad

[young pendant]

Do you use Local RVC @charred topaz ?

[ruby lava]

do anyone have the weeknd model?

[lucid crow]

Hello guys, I have installed vbc but I lost audio how do I get my audio back?

Can someone help me pls

[charred topaz]

yes

[young pendant]

then I don't know how to help, I am planning to use RVC loval as well, so if I get into the same problem and manage to solve it, I will contact with you.
-added as friend just in case I manage to solve the problem while cloning newer voices-

[charred topaz]

thanks like i said my first voice model was good, the ones after that very bad

[twin prawn]

darn. cuz for me, this started happening last night

[rare goblet]

Ayo? @twin prawn level 1 !!!

[lucid crow]

hello im using an amd ryzen what version i need to download on windows 10

[reef ruin]

does anyone know why my voice ai isnt wokring

[thin stump]

Have you even took a look at this? https://rentry.co/VoiceChangerGuide

assuming you need a real time voice changer

[brittle wing]

I can still train models on Easy GUI without GPU?

[lucid crow]

@thin stump can u dm me

[thin stump]

I never used Okada. I only know the guides to refer to here since i've been in this server long enough

[keen oxide]

i downloaded the zip and extracted, what do i do next? i forgot what the instructions said and i lost the link😭

[rare goblet]

Ayo? @keen oxide level 1 !!!

[fluid dust]

Does anyone know what this means?

[fluid dust]

Open it by goin to start http

[keen oxide]

?

[fluid dust]

Is it for to add voice model or for the whole client

[thin stump]

also you can search up your problems here 50% of the time

[lucid crow]

i dmed you. @thin stump

[thin stump]

#🔍│help-w-okada and use the snipping tool on your PC so it's more clear to read

[lucid crow]

@thin stump DMS.

its more easier in dms.

[thin stump]

I'm not receiving anything? and I cannot help you like I said lol

[lucid crow]

done

[fluid dust]

Not sure what that is buddy

[sour spire]

Can i use your easygui to make cover song?

[rare goblet]

Ayo? @sour spire level 1 !!!

[vocal narwhal]

I didn't get the public URL after step 3 ( I did step 1 and 2). Does anyone know what cause this?

[tacit grail]

It should not take me 10+ minutes to get past a percentage of rendering with GPU conversion on ultimate Vocal Remover

[keen cedar]

whats the best gui to use

[rare goblet]

Ayo? @keen cedar level 2 !!!

[forest sluice]

Always replace "blob" with "resolve"

[forest sluice]

On colab/Huggingface Spaces, Ilaria RVC

[forest sluice]

GPU?

[keen cedar]

tried that and its literally traking forever to convert

taking*'

[tacit grail]

Rtx 2070

It's using %100 of my gpu

[keen cedar]

pleaseeee

guysss

i need a one with aa GUi

the illari one does not have one

wait where is the illari colab

ilaria

[tacit grail]

Stilllll on 78%

Wtf?

[gloomy moth]

Anyone know a good non-local host for ai voice stuff, deadass just gonna fw it for a minute and kits got a lil hungry and removed free voice training

[tacit grail]

I don't understand why it's taking so long for it to render

[rare goblet]

Ayo? @tacit grail level 5 !!!

[hard wing]

i wabt to sound like jouffa de stremer

[tacit grail]

should i just reset? its been on 82% for a while

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[dawn nebula]

-colab

[azure marsh]

Suggestions for @gloomy moth

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• GPT-SoVITS (demo), by litagin Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[dawn nebula]

rvc disconnected

[dawn nebula]

#1175430844685484042 #🔍│find-models

[dawn nebula]

yeah reset probs

[wise horizon]

I need help with the Ilaria hugging space, ive been downloading the model and it's been saying "model downloading" but i can't select it and it doesn't show up

[rare goblet]

Ayo? @wise horizon level 1 !!!

[charred topaz]

hello, I have a question abour continue training a model, I was training a voice model 600 epochs, and saved at 300 (eas saving every 25), I test that model (saved at 300) and its fine, but I want to continue training the model to complete the missing 300, how do I do it with Mangio-RVC? I just go directly to "STEP 3", put the Model G and Model D paths and then just train the model with 300 epochs? (300 to complete the original 600 I wanted)

[quick trench]

Yes

[hexed shard]

what happening if train a voice without pretrained models? hypothetically, for 20k epochs

or, less stupid question- how do i make a costum pretrain D and G for a specific voice?

[sand iris]

Not sure it even lets you do that

The whole point of pretrains is that you won't start training a model from purely random data, but instead, start off a model which is supposed to correspond to the "common traits" of speech models

[finite galleon]

Pretty sure you can just remove the G and D files to train with no pretrains but not to sure

[sand iris]

Doesn't it fail to load them in such case ?

I think you gotta edit the code if you want to start from scratch

[finite galleon]

Just looked at the train.py code and it looks like if the pretrain section is none it wont try to load it

[sand iris]

oh interesting

[gentle mural]

is there a new version of the voice changer?

[sand iris]

Haven't had any new releases for over a month

[gentle mural]

can i have a link to the newest one please i have a NVIDIA gpu is there's one better for that

[sand iris]

https://huggingface.co/wok000/vcclient000/blob/main/MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.17b.zip

[gentle mural]

thank you

[rare goblet]

Ayo? @gentle mural level 1 !!!

[lilac glacier]

Hii

The audio is not uploading on applio

Send me a best collab RVC Link

I want to Run it Locally RVC on mac studio(M2 MAX) 2023

[gentle mural]

why aren't any of the voices working?

[lilac glacier]

Punjabi ch dsoo broo

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• GPT-SoVITS (demo), by litagin Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[hexed shard]

1. yup u can do it by simply unload the pretrained models
2. in Applio latest ver u can actually unchek "pretrained" in the train tab

my question is, if i train a non-english voice without pretrained models (which i assume are trained on english), is it even effective?

[proper shale]

I don't think it'll even sound good... but eh, try it

[finite galleon]

You need a lot lot lot of audio data and it must be clean

For it to be good

[elfin dock]

https://gyazo.com/c3ffa6de3dfa08b2a86731e3a30e0e30

got this error with ai cover gen

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[young pendant]

I finally managed to complete my first time at cloning/training a voice, just I need a sample audio to test it out and also uploading right here on AI HUB. 😉

[hexed shard]

how lot? in minutes? hours?

[rare goblet]

Ayo? @hexed shard level 2 !!!

[hexed shard]

and also, is there a guide for creating a specific language pretrained model?

[forest sluice]

It's too costly to make

[low shard]

making pretrains takes alot

[hexed shard]

if i have like a 4 hours samples of a specific language, whats the process of train it to a pretrained model?

[hexed shard]

what is the process brother?

[low shard]

are you trying to make an ai voice model?

or are you really trying to make pretrains

cus those are 2 different things

[tranquil cliff]

he said he wanted to create a pretrains

but I understood voice model

from how he explained

[low shard]

pretrains are some training done by experts to give the ai some ground voice to let other users sue it to make their training faster or better quakity based on the pre trained voice

instead

a voice model is just you cloning a voice

like making it to use it for ai covers later

[hexed shard]

my great guys i know how to train a voice model

I want to make a pretrained model out of 5 hours ready dataset i have in a specific language, so it would make the clones in this language more accurate

https://www.youtube.com/watch?v=qxLF-RegLWs I guess thats the method

[odd shale]

Nah, i don't think 5 hours of a language will be enough to make a pretrain

You'll need probably from 15 to 20 hours.

[hexed shard]

cool ok thanks man.

k we can collect 28 hrs. is the youtube link i provided here is an optional method for that?

[odd shale]

Idk if you can also train a pretrain on RVC.

[low shard]

that video doesn't even have anything to do with rvc

[odd shale]

But i don't recommend making a pretrain using that Tutorial.

[odd shale]

Bro thought that pretrain tutorial was for RVC 💀

[low shard]

Nope its opd and is nothing related to rvc

I dont even know where u got thst

Also why would you need to make pretrains?

Pretrains are a thing downloaded in rvc made by experts that are trainings already made to have some ground of voice of like english accent or male voice for example, so that when people use rvc with that pretrains, the better the pretrain is, the faster is the training that other users using that pretrain

[hexed shard]

does someone here can just tell me whats the method of creating a custom pretrained g and d for rvc? or even refer to a doc about it? cause i cant find anything related to it.

[low shard]

Its a thing done by experts

[hexed shard]

bro i swear i know whats pretrained and voice cloning is..

[fluid matrix]

but you never trained a model?

[hexed shard]

ok dont worry about that. do you have sources or docs about the method or just the knowledge its for experts?

[rare goblet]

Ayo? @hexed shard level 3 !!!

[hexed shard]

like 75 models

[fluid matrix]

using rvc?

applio probably

[hexed shard]

applio is rvc

[fluid matrix]

yeah i know

[hexed shard]

yeah i made through applio and before that rvc and before that svc

[fluid matrix]

why do you even need to make a new pretrain?

what language? what usecase?

[hexed shard]

i want to take over madrid through yiddish pretrained models

[odd shale]

What the hell are you planning?

[hexed shard]

do you have links to help now?

[fluid matrix]

madrid? isnt yiddish a jewish thing?

[odd shale]

Madrid, España

Why he wants to conquer Madrid?

[fluid matrix]

also i doubt making a new pretrain would benefit your models

[hexed shard]

jesus guys im not gonna take over madrid

[sand iris]

kkkkek

[fluid matrix]

😭

[odd shale]

XD

[fluid matrix]

yeah its a joke 💀

nvm

[hexed shard]

https://huggingface.co/ORVC/Ov2Super

[fluid matrix]

look if you need to ask how to make a pretrain then you probably shouldnt be making one

it wont improve your model anyways

[hexed shard]

thats an enhanced custom pretrained g d

[little island]

how would I go about using a model for text input to speech (rather than having a vocal input)
for general speaking, where it adds things like pitch automatically

[rare goblet]

Ayo? @little island level 1 !!!

[hexed shard]

my man forgive me, i highly doubt you know what you are talking about

[fluid matrix]

my brother in christ, i highly doubt YOU know what you are talking about 💀

stop being so pretentious

[hexed shard]

believe me, i know.

[odd shale]

Blaise is currently trying to fix that VoRAS thing, so until then, there's no way to make an better pretrain than all the ones usable.

[fluid matrix]

bro you are asking us to help you how can you know 💀

and dont expect any help if you talk like that lol

[hexed shard]

the default pretraines g d trained on english.

[odd shale]

Bruh, just go and read a guide about how to make models, NOT pretrains

[low shard]

you can use any tts like 11labs and then input the generated audio

[odd shale]

Stop it.

[half cove]

The default pretrain is multilingual

[little island]

I do use elevenlabs for tts, but I was wondering if I could run something like that locally with similar/better results

[low shard]

you could try searching up for edge tts or bark tts

[hexed shard]

i can teach you how to do it bruh

[odd shale]

Pretrains?

Nah, no.

[little island]

yeah I actually got bark tts I don't know why I haven't used it yet

💀

thanks tho

[hexed shard]

no. you told me to read a guide about creating a model

[sand iris]

Stop

[odd shale]

Yeah, i don't understand why you exactly wanna make pretrains

[fluid matrix]

bro stop arguing

you dont need to make a new pretrain, its useless

[odd shale]

Stop insisting.

[hexed shard]

you really wanna know?

[fluid matrix]

are you trolling

[odd shale]

Are you serious?

[molten pecan]

they must be. if they refuse to collaborate then you can ignore them

[fluid matrix]

there is nothing to collaborate he should just stop wasting time 😭

[hexed shard]

😦

[low shard]

yww

[fluid matrix]

and acting like he knows shit when he is asking US for help

[little island]

what am i witnessing rn

[half cove]

[hexed shard]

ok first of all, https://huggingface.co/ORVC/Ov2Super i highly recommend u guys to train ur next model with those enhanced pretrained models

[molten pecan]

the stuff we see everyday

no worries

[fluid matrix]

yeah he is trolling

[low shard]

the most sane discussion in ai hub

[odd shale]

If you really know how to do things, then stop typing on this help channel.

[molten pecan]

yea we already do. they are pretty good

[fluid matrix]

i mean we can argue about that

[odd shale]

Those pretrains were made by SimplCup.

These are meant just for short datasets that don't get past from 9 mins

[molten pecan]

dang I should make a pretrains guide too

[hexed shard]

yeah they are

[fluid matrix]

on how to make one?

[hexed shard]

thanks. a normal person

[odd shale]

Necesitarás investigar un montón entonces

[little island]

real question if i live in an area where electricity is expensive is it over

[molten pecan]

no el como hacerlos de cero, sino qué son y como usarlos nomas

[odd shale]

So you're saying we're weirdos for not agreeing with your idea?

[fluid matrix]

Si necesitas una guía para eso, entonces no deberías realizar una capacitación previa, porque es básicamente el mismo proceso pero mucho más difícil.

[hexed shard]

no. you just didnt helped at all

[molten pecan]

it's not how to make them from scratch 😭

just what they are and how to use them

[odd shale]

Julia sería innecesario una guia de como usar pretrains.

[fluid matrix]

how much data have you gathered already anyways? and from where?

[odd shale]

Tú al seleccionar un sample rate ya los estás usando

🤦‍♂️

[fluid matrix]

isnt that explained already? i guess you can add a couple of sentences lol

[molten pecan]

pero puedes elegir entre varios pretrains si mal no recuerdo

[odd shale]

Almenos en el colab de RVC Disconnected puedes seleccionarlos.

Pero en cualquier version de RVC local NO

Tienes que si o si descargarlos y poner la ruta a los archivos D y G en la GUI

[fluid matrix]

fine. how much data have you gathered already? and from where?

[hexed shard]

28 hours of dataset contains audio segments and transcriptions

[fluid matrix]

you already have that or plan to?

[hexed shard]

already have that on my computer rn

[fluid matrix]

and where did you get it from

[hexed shard]

a place called HF

[odd shale]

Huggingface?

[fluid matrix]

can you link it?

[hexed shard]

yeah HF is hugging face

[hexed shard]

huh?

[fluid matrix]

you said you got it from huggingface

im asking you to send a link to something from the dataset or the whole thing

[hexed shard]

dont you know there are datasets on HF?

[fluid matrix]

bro

im asking where you got yours

not some random one

https://huggingface.co/datasets/imvladikon/hebrew_speech_kan?row=96

maybe its this

maybe its something else

how am i supposed to know

[hexed shard]

not this

but why is that important?

[fluid matrix]

to see if its actually useable?

[hexed shard]

so search for the 28 hours one

[fluid matrix]

you cant just send the link?

[hexed shard]

but i really dont think you will know if its useable because u dont have a clue about the process of making pretraineds

[rare goblet]

Ayo? @hexed shard level 4 !!!

[fluid matrix]

then dont ask for help

good luck

no one will help you here

[hexed shard]

then help!!!!

if you know the process just provide help

[fluid matrix]

BRO HOW TF ARE WE SUPPOSED TO HELP IF YOU WONT COOPERATE

[half cove]

#✨│ai-help message

How tf

Will he help you

If all you say is "I know what I'm doing and you don't "

[hexed shard]

what?

[fluid matrix]

sorry man

[hexed shard]

i think there is a language barrier maybe or i dont express myself properly

[fluid matrix]

no you just talk shit and dont actually want help

[molten pecan]

this is going nowhere

[fluid matrix]

im not going to click through all the links in the world to miraculously find the dataset you used if you can just link it

[hexed shard]

lets say i have a proper dataset ready. what should i do? can you just say?

[fluid matrix]

no i cant

[sand iris]

Let's end it there

[fluid matrix]

yeah

[molten pecan]

https://aihubdocs.github.io/en/essentials/how-to-make-voice-models/ here's how to make voice models

[fluid matrix]

he knows already 😭

[sand iris]

stop

pls

[fluid matrix]

🙏

come back when you made your pretrain

dont even bother typing anything

[hexed shard]

chill man

[sand iris]

Find out how to make it initialize the training models (G & D) with random weights. There might be features in place for it, or not. Haven't checked in a while.
From what I remember the original RVC project requires you to load files, but it should be real easy to make it start from nothing. From there, just train a model on all the data, and hope for the best, and then save the G & D files resulting from it

[fluid matrix]

conversation ended, good luck

[charred topaz]

hello, how do I continue training? I was doing 600 epochs but saved at 300, voice works fine but I want to complete the missing 300?

[fluid matrix]

are you using a colab or local rvc? this might help:https://aihubdocs.github.io/en/ . But usually its just entering the exact same settings and clicking train again

[charred topaz]

loval rvc

[fluid matrix]

yeah then you need to reenter the same exact same settings

but dont preprocess again or anything like that

just press train

https://aihubdocs.github.io/en/rvc/local/mainline/

[charred topaz]

so, just skip directly to STEP 3?

[fluid matrix]

read the section of whatever you are using to be sure

but basically same settings and dont press anything else besides training

only thing you change is your epoch count

no reprocessing or anything

[charred topaz]

that page doesnt mention "Load pre-trained base model G path:" or "Load pre-trained base model D path:"

[fluid matrix]

because you dont need to do that

just said to not load anything again

also you sure you are using local rvc?

[charred topaz]

Yes local Mangio

[molten pecan]

not yet. I will in the future. for now the docs are mostly oriented to newbies

so keep an eye on dat =3

[charred topaz]

oh, so, I have to put that value?

[fluid matrix]

same exact settings. meaning if you used a custom pretrain, add that again too. everything has to be the same

[molten pecan]

iunno about pretrains so idk

[half cove]

It will load the correct ones for you if you choose the same settings unless you used a different pretrain than the base which I doubt

[charred topaz]

I tried with youtube tutorials and one of them said to continue training you had to put that G and D vales

[fluid matrix]

just increase the epoch amount and press train again after reentering all the same settings, simple as that. but do not reprocess or anything like that, otherwise you will screw up the model

[half cove]

Not the fucking yt videos

[fluid matrix]

dont use outdated and misinformed youtube tutorials

hope you didnt mess it up already

[charred topaz]

I mean I did used the page

just had to search the video cuz the page didnt mention specifically about G and D values

[fluid matrix]

please just do it like i just said, reprocessing it will mess up your model

might even delete the progress

[hexed shard]

man send a screenshot of the train tab

[molten pecan]

yea we don't recommend youtube tutorials, as usually they have outdated info

ensure the vid was uploaded fairly recently, no more than 2 months. that's my personal tip

[hexed shard]

you just need to keep the sample rate and the ver (1 or 2) from the original training process and the pretraineds will appear automatically

and press training. doing the preproccess again would not harm anything though, dont worry.

[fluid matrix]

(it will)

[molten pecan]

v1 arc >>>

[hexed shard]

nah it wont if he use the same settings. just place the same files again in the log folder

[fluid matrix]

reprocessing will redo the pitch analysis and mess everything up, if not worse

recently tested by charlotte too

[hexed shard]

if you dont remember the exact settings you trained with, simply open the "train.log" file in your model log folder to see the settings

[fluid matrix]

@charred topaz

first amazing thing you said

[charred topaz]

I created "robie1" voice model hours ago and works fine, but only did 300 epochs of the 600 total

so i skip directly to STEP 3?

[hexed shard]

thats the corrected settings from the original train?

[charred topaz]

I didnt put the G or D values yet

[hexed shard]

you dont need to put any of that anywhere

[fluid matrix]

what do you want to put there? it already has them? or did you use a custom pretrain

[charred topaz]

I created "robie1" voice model from 0, stop the first training at 300 epochs of the 600 total

[rare goblet]

Ayo? @charred topaz level 4 !!!

[fluid matrix]

please just refer to what we said previously or read this https://aihubdocs.github.io/en/rvc/local/mainline/

its very simple

and i would recommend switching away from mangio as its outdated and you most likely dont need to train at 40k sample rate

[hexed shard]

go to log/robie1 open the "train.log" and paste please whats written in the first lines like in here:

[fluid matrix]

he already reentered all the settings why would he want to do that

[hexed shard]

so just press train brother

[fluid matrix]

no

dont

why is training on 20?

and you sure your batch size is correct? @charred topaz

[charred topaz]

i changed it

[fluid matrix]

again, same exact settings

to 600?

[charred topaz]

it was originally 600 like i said, i stopped training at 300 epochs and saved (was saving every 25)

[fluid matrix]

also why are you using v1

[hexed shard]

i though you said he entered the exact settings

[fluid matrix]

i thought you know what you are talking about

@charred topaz did you use v1 before?

are you sure it wasnt v2?

[hexed shard]

?

[fluid matrix]

im not going to start arguing again

[hexed shard]

so dont

[sand iris]

stop

pls

[charred topaz]

ok now its the same as before, i didnt press any button, but I still didnt touch the "G" and "D" pre values

[fluid matrix]

dont just change the settings, we keep saying to use the exact same. are you sure you used v2? yes or no? if not check in logs

[hexed shard]

its already loaded

dont worry

thats the g and d

[fluid matrix]

why do you keep changing it

you cant just enter random values

exact

same settings

[sand iris]

it's v2 wait fuck that was pickachu's screenshot nvm

[fluid matrix]

well he had v1 selected before

[dusty rampart]

what

[fluid matrix]

and way different settings

@charred topaz instead of guessing, can you please open your logs and recheck?

[charred topaz]

the first image i send was worng

[fluid matrix]

the 2nd is wrong too

[hexed shard]

bro can you just do that? @charred topaz

[fluid matrix]

dont send it in here @brittle wing

also thats not great for a model

[molten pecan]

it's possible to make a model with a 10 second dataset, but it's not going to sound good

[hexed shard]

its his, i just cropped it to show him

[sand iris]

oh it's v2 then

[molten pecan]

make it at least 2 minutes. that's my personal tip

[sand iris]

it managed to load the v2 pretrains

it cannot be v1

[hexed shard]

actually with this https://huggingface.co/ORVC/Ov2Super pretrained files there is a possibility of training with 10 sec dataset, but its the minimum

[fluid matrix]

but he keeps switching his settings, they were way different before

[charred topaz]

2024-02-21 23:17:06,717 robie1 INFO {'train': {'log_interval': 57, 'seed': 1234, 'epochs': 20000, 'learning_rate': 0.0001, 'betas': [0.8, 0.99], 'eps': 1e-09, 'batch_size': 10, 'fp16_run': True, 'lr_decay': 0.999875, 'segment_size': 12800, 'init_lr_ratio': 1, 'warmup_epochs': 0, 'c_mel': 45, 'c_kl': 1.0}, 'data': {'max_wav_value': 32768.0, 'sampling_rate': 40000, 'filter_length': 2048, 'hop_length': 400, 'win_length': 2048, 'n_mel_channels': 125, 'mel_fmin': 0.0, 'mel_fmax': None, 'training_files': './logs\robie1/filelist.txt'}, 'model': {'inter_channels': 192, 'hidden_channels': 192, 'filter_channels': 768, 'n_heads': 2, 'n_layers': 6, 'kernel_size': 3, 'p_dropout': 0, 'resblock': '1', 'resblock_kernel_sizes': [3, 7, 11], 'resblock_dilation_sizes': [[1, 3, 5], [1, 3, 5], [1, 3, 5]], 'upsample_rates': [10, 10, 2, 2], 'upsample_initial_channel': 512, 'upsample_kernel_sizes': [16, 16, 4, 4], 'use_spectral_norm': False, 'gin_channels': 256, 'spk_embed_dim': 109}, 'model_dir': './logs\robie1', 'experiment_dir': './logs\robie1', 'save_every_epoch': 25, 'name': 'robie1', 'total_epoch': 600, 'pretrainG': 'pretrained_v2/f0G40k.pth', 'pretrainD': 'pretrained_v2/f0D40k.pth', 'version': 'v2', 'gpus': '0', 'sample_rate': '40k', 'if_f0': 1, 'if_latest': 1, 'save_every_weights': '1', 'if_cache_data_in_gpu': 0}
2024-02-21 23:17:09,185 robie1 INFO loaded pretrained pretrained_v2/f0G40k.pth
2024-02-21 23:17:09,462 robie1 INFO loaded pretrained pretrained_v2/f0D40k.pth
2024-02-21 23:17:45,456 robie1 INFO Train Epoch: 1 [0%]

[fluid matrix]

thanks

[sand iris]

Yeah, but that's what the log said it was, so gotta pick back v2

[fluid matrix]

those were not his logs

these are

[brittle wing]

I didn't know well, so I wanted to leave it up to the AI model production

[fluid matrix]

i saw your dm, wait

[hexed shard]

40k sample rate on v2

[junior glade]

give plz link to working rvc

[rare goblet]

Ayo? @junior glade level 1 !!!

[charred topaz]

now i put the exact same values, they are the same exepct the G and D part

[fluid matrix]

https://aihubdocs.github.io/en/

[hexed shard]

the g and d already loaded

[dusty rampart]

ima dm you

[fluid matrix]

they are tho, dont bother with the pretrains if you didnt use a different one

[charred topaz]

here, after i created robie1 and saved at 300 epochs of the 600 i wanted to train, it created this

D_23...3.pth and
G_23...3.pth

I do nothing with those?

[forest sluice]

Just keep it there, crucial for training models

[junior glade]

don't you have any problems with rvc?

[charred topaz]

taken from logs

2024-02-22 02:32:51,008 robie1 INFO loss_disc=3.495, loss_gen=3.993, loss_fm=11.068,loss_mel=18.170, loss_kl=0.781
2024-02-22 02:33:09,881 robie1 INFO Saving model and optimizer state at epoch 300 to ./logs\robie1\G_2333333.pth
2024-02-22 02:33:11,080 robie1 INFO Saving model and optimizer state at epoch 300 to ./logs\robie1\D_2333333.pth

[hexed shard]

bro just put

10 batch size, 600 (or whatever you want 300<) 40 sample rate and v2

and press train model.

[proper shale]

yea it loads those up for resuming

they're for resuming specifically xd

[rare goblet]

Ayo? @proper shale level 89 !!!

[proper shale]

AYOO

[charred topaz]

So i have to put those values? everyone else telling me not to do that

[proper shale]

put what value

on what

[hexed shard]

this is great. make it look exactly like this and press train model

[proper shale]

if you mean same settings, yes, you must keep the same settings, you can change total epochs to be trained to and the save frequency

[charred topaz]

I thought I had to put the G and D files that were created

2024-02-22 02:32:51,008 robie1 INFO loss_disc=3.495, loss_gen=3.993, loss_fm=11.068,loss_mel=18.170, loss_kl=0.781
2024-02-22 02:33:09,881 robie1 INFO Saving model and optimizer state at epoch 300 to ./logs\robie1\G_2333333.pth
2024-02-22 02:33:11,080 robie1 INFO Saving model and optimizer state at epoch 300 to ./logs\robie1\D_2333333.pth

[hexed shard]

its detecting it

[rare goblet]

Ayo? @hexed shard level 5 !!!

[proper shale]

Yeah no

Automatically detects them anyway

And plus those 2 paths are for pretrains, which don't need to be changed, so dw ;)

[charred topaz]

ok, so i do not press "Process data" button, I do not press "Feature extraction" button, just directly press "Train model" button?
Also the page says to not press "Process data" again, but doenst specify about "Feature extraction"

[fluid matrix]

bro

[hexed shard]

exactly

[fluid matrix]

just press train

nothing else

[half cove]

They are already saved in logs so you don't need to do it again

[fluid matrix]

stop pinging me or the others, i was away for a bit

will now

[hexed shard]

the preprocess+feature ex. is CREATING the log folder of the model that the training using. once the folder and files are created, you dont need to recreate it. u understand?

[charred topaz]

ah ok, just to clarify, the values shown there, I didnt change/touch them, I thought I needed to replace the G with "logs\robie1\G_2333333.pth" and the D with "logs\robie1\D_2333333.pth"

[hexed shard]

👆

[fluid matrix]

you asked this 10 times now, not even exaggerating. do you read the replies? just press train please 😭

[charred topaz]

ok ok gezz ill press train sorry i literally started this ai thing like yesterday

[ionic sedge]

Hello, I might be using the wrong jargon here, but does anybody know if there is difference in audio output between different RVC notebooks, like the AICOVERGEN UI vs Ilaria RVC Colabs.

[rare goblet]

Ayo? @ionic sedge level 1 !!!

[fluid matrix]

no its all based on rvc

just different colabs

[half cove]

Main difference is that aicovergen uses mdx models on your audio so you can put instrumental but it can be kinda iffy

[ionic sedge]

Ah, thank you. I had been using one of the old easyGUI notebooks, but it died a few weeks ago. I've been using some of the new notebooks, but had some bits of low quality output, but maybe I was just using a low quality audio source

[odd shale]

No

Don't do that.

[boreal holly]

can anyone help me set up the AI voice changer please

[odd shale]

-realtime

[azure marsh]

This interaction has expired, use the command /guides realtime if you wish to see it again.

[lament wagon]

Might as well, share this here too

The public URL is missing on the colab page

[low shard]

thats outdated

[lament wagon]

Like... only since yesterday has it become outdated?

[low shard]

Theres the updated easy gui colab

https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw#scrollTo=G0MEhFM19Vq6

[lament wagon]

Oh, thanks

[low shard]

What? The outdated easy gui rvc download improved is outdated from alot

I gave you the link that gets updated

If you want there are also other ways to train and inference

You can also check our docs

https://aihubdocs.github.io/en

There isnt easy gui on it tho

[low shard]

yw

[lament wagon]

Umm, I think this is for making voice models... do you have any for making the covers?

Or is there a different channel for me to ask for that

[low shard]

It has also inference (making ai covers), scroll down

[lament wagon]

Aight

...This is WAY too hard for me or my buddy who uses zip files

Sorry for complaining, it's just difficult

[low shard]

If you are trying to use it only for making ai covers, you can skip the training part and go only to the inference part

Would you rathsr a way to make ai covers with ui (interface)?

[lament wagon]

Yes

[low shard]

You can use the Ilaria RVC UI Colab please check out the guide!

[lament wagon]

Thanks again, hoping this works

[low shard]

It should be easier to use

Its only for inference tho (making ai covers)

You're welcome

[boreal holly]

Pipeline is not initialized.
[Voice Changer] Waiting generate pipeline...

why am i getting this

please help

[lilac glacier]

Collab RVC link plzz

[low shard]

im sorry but i dont know, please ask in #🔍│help-w-okada or #1192011222023950368

[low shard]

check out our docs, we have guides for do inference (making ai covers) and training (making) ai voice models on colab! https://aihubdocs.github.io/en

[dusk token]

Hello. I keep receiving this error in colab:
'nonetype' object has no attribute 'tobytes'
What may cause the problem?

[fluid matrix]

can you send a link to the colab you are using? looks like it didnt recognize the audio, try disconnecting the runtime and try again

think that was one of the more random errors

[dusk token]

https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing#scrollTo=9yQA9h6N2YV7

I diconnected and ran again, and still have this problem

[fluid matrix]

but if you send your colab we can know for sure

try again with this colab https://colab.research.google.com/drive/1u1brjK8IZt647UsbZuGYfW29oFM2I4tk?usp%3Dsharing&sa=D&source=editors&ust=1704303145687891&usg=AOvVaw3M9tmokG80RXF-GD1LJqCL

the ilaria colab works again but your error is one of the more random ones, easier to try that one imo

[dusk token]

Thanks, I'll try it

[rare goblet]

Ayo? @dusk token level 1 !!!

[mystic nebula]

so uh ye i have blue suit (aka Nice Godman) how do i get it to work

[low shard]

@mystic nebula so as we were saying in gen chat, can you send me the model link that you are using?

[mystic nebula]

ye bro

[low shard]

Send the model link here

[mystic nebula]

https://huggingface.co/Rigosaenz/Goodman20102017/resolve/main/Mr. Goodman (SML) (2010 To 2017).zip

[low shard]

Okay checked it and it has everything, what tool are you using for making ai covers?

Like, locally or a google colab?

[mystic nebula]

explain more clearly i dont understand

[rare goblet]

Ayo? @mystic nebula level 2 !!!

[low shard]

Did you ever make an ai cover before?

[mystic nebula]

in kits ai with Avenger Chuck but not in rvc

[low shard]

Ohh, got it, i suggest you to not use kits ai tbh, its really greedy wheb rvc is all free, do you got a good pc or do you have a phone/bad pc?

[mystic nebula]

good pc

[low shard]

U got an rtx ?

[mystic nebula]

i got omen pc

[low shard]

What processor and how much ram?

[mystic nebula]

i thought this was helping me convert the mr goodman file

what is dis

[low shard]

I need to see if your pc can support to run it locally

Else if it isn't powerful enough you have to do it online on google Colab

[low shard]

Yea thats the storage, its a different thing than ram or processor

[dusk token]

It says: "Invalid Youtube url"
In the instructions, it says I can use Drive link which I do. I have wav files on my drive, not a Youtube video. Should it work in this case?

[low shard]

Btw, check out our docs, https://aihubdocs.github.io/en , this contains guides to inference (make ai covers) online and locally

If your pc has got enough ram and good processor it should be able to do it locally, if it doesn't, just use the online way, there is everything you need in the docs

[mystic nebula]

it is a good pc tho plenty of ram

[low shard]

Yea you should be able to do it locally then

Was just asking to check yk, dont want users to waste time doing things that their pc dont even support

[mystic nebula]

link to the website

[low shard]

check in our docs, https://aihubdocs.github.io/en/, there are 3 ways to do it locally, each one has pro and cons, its very detailed and see which is better for you

[mystic nebula]

im usin applio

[low shard]

okay, if you get any error with it ask it here, check the local applio guide https://aihubdocs.github.io/en/rvc/local/applio/ and goodluck with it!

[mystic nebula]

also how do i train a ai becuase i do want a ai of Eevee From Pokemon Plush Pals

ig i better read it

[lament wagon]

Hey @low shard
Update on my whole thing

This piece of shit isn't working for me or my friend

[rare goblet]

Ayo? @lament wagon level 2 !!!

[ashen glen]

-rvc

[azure marsh]

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

❔ English FAQ

https://docs.aihub.wtf/faq

Credits: AI HUB FRANCE

✨ More guides

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

How To Make an AI Cover With Ilaria RVC

Link: Rentry
Credits: 👽 Julia (ailen2091)

[slate gate]

Hey, can anyone help me with an error if you know what the problem is

[fluid matrix]

you can upload your upload your files onto your drive and then copy the path from the folder symbol on the left

[knotty parrot]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• easyGUI, by rejects Google Colab

🤗 Hugginface Spaces

• Applio, by IA Hispano Huggingface Spaces
• Ilaria RVC, by thestingerx Hugginface Spaces
• RVC-HFv2, by r3gm Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces
• GPT-SoVITS (demo), by litagin Huggingface Spaces

More commands: -audio, -local, -overtrain, -realtime, -rvc, -help

[knotty parrot]

anyone know if there is a way to get the huggingface links to models instead of the applio?

[slate gate]

Does anyone know why would this error appear "AttributeError: module 'huggingface_hub.constants' has no attribute 'HF_HUB_CACHE'"

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[low shard]

hey are you using the hugging face space of ilaria rvc or the ui colab?

if you are using the hugging face space, you need to be sure you duplicated the space first

[azure marsh]

🎙️ Realtime Voice Changer (Local)

Guide for W-Okada's RealTimeVoiceChangerClient

Description: This one has more options and is somewhat more convenient to use than Real-time RVC.
Link: Rentry
Credits: ravencutie21

How to use real time conversion with RVC (voice changer)

Description: More accurate than W-Okada's, but requires a decent GPU.
Link: Google Docs
Credits: teas.hi

RVC TTS

Description: Text to speech with RVC
Link: Rentry

[brittle wing]

I'm using google disconnect to make my models, but what will happen whilst am training my model & then my trail ends?

[onyx ferry]

[true kelp]

ocupo ayuda

[lament wagon]

The UI Colab

[charred topaz]

how can I share a voice model I created with someone? like I just need to send them the files? what files?

[low shard]

its weird, click on runtime and then disconnect runtime and delete, retry again and if it still gives error, see in the colab what could be the error, also be sure your runtime is a gpu runtime

[low shard]

do you want to send the model ONLY to him?

it depends if he uses online or locally rvc, if he does locally, you just need to give him your (modelname).pth and added.index file zipped together

and he would be able to load it locally

if instead he uses the google colabs, you need to upload your zipped model into your google drive, set the link viewable to everyone that has it, and then share the link to him to use it

if you instead want the model to be public for anyone, check this guide in our docs

[charred topaz]

so, these 2 files? (the ones clicked)

[lament wagon]

Look, me and my friend are having the exact same issues, that should be screaming to you that this colab isn't working

[dawn nebula]

yep

[charred topaz]

ok thanks

[low shard]

be sure you are connected to the T4 free gpu runtime

i just runned the 2 cells and clicked speak and it seems all working fine

check the colab output if you get errors and tell me them

[lament wagon]

I am

It was fully connected and loaded, I went through all the same things I normally do, AND IT DIDN'T WORK

[low shard]

when you get the error, go to the colab and see if theres an output error

[worthy lantern]

rvc not working for me atm?

[cyan mesa]

https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing&authuser=3#scrollTo=7vh6vphDwO0b

Is the site broken, it doesn't need to give 2 links, what's the problem?

[low shard]

This is the outdated version of easy gui

See this #✨│ai-help message

[molten ember]

me too 😦

[lament wagon]

Okay, it works now

[wicked cave]

!howtoask I'm using the ardha27 collab CovernGen_No_UI, and I put in my voice model and the youtube link for my song cover, and it keeps saying

python3: can't open file '/content/src/main.py': [Errno 2] No such file or directory
2

what am I doing wrong?

I just wanna do free ai covers, why does it have to be so hard?

[molten ember]

Even if I pay, it doesn't work

{Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered }

[rare goblet]

Ayo? @molten ember level 1 !!!

[molten ember]

That phrase will pop up

[rare goblet]

Ayo? @worthy lantern level 4 !!!

[worthy lantern]

@lament wagon

[dry brook]

how good is Ilaria RVC TTS?

how does it compare to styletts2 and tortoise

[quiet wind]

Hey, are there any docs bout how to use Applio?

[rare goblet]

Ayo? @quiet wind level 1 !!!

[worthy lantern]

why that takes so long time for me to conver the song on Ilaria?

how to download faster?

[wicked cave]

Help me

[azure marsh]

This command has been changed to !howtoask

[odd shale]

!howtoask

[patent trellis]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[odd shale]

You're using the HF space or Colab?

[worthy lantern]

huggingface

[odd shale]

Yeah.. Stop using it, it's obsolete when there's the Colab version

[worthy lantern]

wheres the colab version

😭

[odd shale]

https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing

[slate gate]

Hey everyone, can you please share the link for a collab (where I can create an ai cover and not to train or somethig else). Thnaks!

[rare goblet]

Ayo? @slate gate level 1 !!!

[quiet wind]

Hey! Every time i try to download a model there is an error. I dont know what to do. All the links are from here:/

[quiet wind]

I've tried 5 different links

[worthy lantern]

how the new one works?

@odd shale

[odd shale]

Use Ilaria RVC

[odd shale]

.

[odd shale]

https://colab.research.google.com/drive/16LkwvFZeudTpUOsE_6bMjOq2qkxFo8Hr?usp=sharing

[worthy lantern]

how that one works?

same with the old one?

[supple ibex]

how do I upgrade from applio v2 to v3 while keeping all my files (audios, models, pretrains)

[dry brook]

how do i get links for a hugging face model?

[rare goblet]

Ayo? @dry brook level 2 !!!

[dry brook]

im a little lost

[quiet wind]

I'm using this link, but i can't download the models

[dry brook]

and the docs arent very clear

[odd shale]

Idk about that issue

[odd shale]

Exec the first cell, then the second one and wait till a public URL pops out

[odd shale]

#1175430844685484042

[worthy lantern]

EYO was much faster, ty for ur help

[odd shale]

@dry brook Click on any model on the voice models channel

Then look for a link like this:

[nimble salmon]

anyone here use voice changer client demo? i cant seem to hear the voicechanger but i can hear myself while using the app

[dry brook]

what if i want to find something specific on hugging face? do i look for the pth file?

[odd shale]

It seems that model was packed unproperly, so it's unusable.

Unless you download the .pth and the .index, give them other names, put them on a folder with same name and .zip them and upload them to your HF account

[quiet wind]

Maybe i should try doing it on the pc later

[proper jasper]

hi, how do i download the weights.gg models via links on Appio? because I try but it says: HTTP Error 403: Forbidden

[dawn nebula]

i believe that the model has been removed

[lament wagon]

Yes?

[worthy lantern]

what chainsmokers song was that?

[lament wagon]

Call You Mine

[wraith moon]

So i used Huggingface spaces to batch convert a bunch of voice files. It says "Files from /home/user/app/audios inferred successfully." and they've been outputted to a folder in the space. How do i even download it?

[slate gate]

omg thanks!!!

[odd shale]

You're welcome

[odd shale]

[halcyon oxide]

When running the rvc step 3 always gives me errors in recent time. It wasn't the same case a month ago. Did something change?

Warning:Couldn't find TensorRT

[wraith moon]

Short answer: I can't?

[rare goblet]

Ayo? @wraith moon level 1 !!!

[odd shale]

Maybe not

[halcyon oxide]

Please someone help me..

[loud spear]

Hello everyone.

[candid crescent]

Do I open the url it tells me to open so that I can launch the RVC?

[loud spear]

I'm downloaded the Magio RVC train 7z file and I'm gonna use it since I tried all the links are disconnected. Hopefully there is actually a chance for me to use this program to make AI song covers.

[royal pond]

Hi, everyone!!
An error appears when I try to download the model.
It says "JSONcodeError" and "File exists, proceed with creation of creds and client"
Could someone please help me?, Thanks<3

I did a cover the last month with the same model and now it doesn't let me, I don't understand why

[candid crescent]

Is there a reason why I can hear myself? 😭

[random garden]

I second this. Also getting the error and not sure how to fix :p

[left oasis]

wtf?

[candid crescent]

How do I fix all my audio being played?!

[rare goblet]

Ayo? @candid crescent level 1 !!!

[spare shale]

Did you install 1?

[dawn nebula]

wdym?

you can ask in chinese if you want, i can understand

[dawn nebula]

check your output settings

[candid crescent]

我的所有音频、通话中的朋友以及游戏中的所有声音都在播放。不断有回声并且很烦人

• if that works

[dawn nebula]

你用的变声器吗？

还是什么

@candid crescent

[candid crescent]

yes

Voice changer

[dawn nebula]

看看设置

give me your settings

[plain comet]

best settings for a girl voice?

[candid crescent]

Settings?

它也一直在窃窃私语

[plain comet]

can it turn my voice to really girl not some ai robot voice?

[dawn nebula]

turn off monitor

[candid crescent]

it just stopped working entirely

[dawn nebula]

its still working, its just that you wont hear urself

[candid crescent]

lifesaver i hated that sound

[dawn nebula]

also f0 det should be rmvpe

not crepe

[candid crescent]

alright

[plain comet]

? good or

btw is it fine to speak with headset mic?

[candid crescent]

it's still echoing though

[dawn nebula]

why is everyone using crepe for f0 det

[dawn nebula]

use rmvpe

[candid crescent]

like I'll say "Hello" and it'll say "hello" quieter

[half cove]

yt tutorials im guessing

[plain comet]

rmvpe?

[dawn nebula]

we need to post new ones

[candid crescent]

yes, it was said as the best setting but I guess it was a lie 🤣

[half cove]

i knew it

[plain comet]

like people voices?

[dawn nebula]

when was the video being created

[rare goblet]

Ayo? @plain comet level 2 !!!

[dawn nebula]

f0 det

[candid crescent]

2 months ago

[dawn nebula]

@half cove rmvpe was available back then?

[half cove]

yes

[plain comet]

can you please explan what is rmvpe?

[candid crescent]

I mean it's not like there are any others when I searched it up

[dawn nebula]

its another f0 detection which is better than crepe

[candid crescent]

so I just had to follow it,

[half cove]

atleast crepe tiny

[dawn nebula]

@half cove i was about to mute that person

bro

[half cove]

why

ohh

fuck

[dawn nebula]

💀

[plain comet]

s.thresh what should i put it on?

[half cove]

i didnt see it

[candid crescent]

by the way the echos are still playing, all of my ingame audio and friends speaking are heard

[dawn nebula]

use headphones, it will help alot

[half cove]

it was sapphire

[dawn nebula]

in the middle

[candid crescent]

what do you mean by this?

[dawn nebula]

戴耳机回音会小很多

[plain comet]

i sound so Asian

[candid crescent]

我已经戴着耳机，但它正在播放我的所有音频。不仅仅是我的游戏音频，还包括来自 Discord 或其他 YouTube 视频的任何声音

[rare goblet]

Ayo? @candid crescent level 2 !!!

[dawn nebula]

你把别的音频关了啊

[candid crescent]

你说的其他音频是什么意思？还有其他音频吗？

[dawn nebula]

你不是说你discord还有youtube有声音吗

你把视频关了就没了啊

[candid crescent]

啊不，我的意思是因为我想用变声器玩游戏，但我的意思是我所有的游戏音频都是用变声器发送的。普通麦克风不会发生这种情况，我在任何地方都找不到解决方案

这是我现在的设置

我被告知要把这些设置

[dawn nebula]

你看看你系统输入输出设置

[candid crescent]

有什么不合适的地方吗？

[dawn nebula]

左边的选 mic in at rear panel

右边的你平时用哪个就选哪个

[candid crescent]

右边的是我用的。现在应该停止回声了吗？

[dawn nebula]

应该没问题了

[candid crescent]

麦克风仍能听到其他人说话

但它停止了回声

[dawn nebula]

游戏里的话

你把输入改成变声器

系统的话维持现状

[candid crescent]

我的不和谐停止从虚拟电缆中接收任何声音。我应该将其连接到其他输出吗？

[hexed cave]

-docs

[azure marsh]

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

❔ English FAQ

https://docs.aihub.wtf/faq

Credits: AI HUB FRANCE

✨ More guides

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

How To Make an AI Cover With Ilaria RVC

Link: Rentry
Credits: 👽 Julia (ailen2091)

[hexed cave]

-links

[azure marsh]

🖥️ Local stuff

• 🍏 Applio, 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 GitHub
• GPT-SoVITS, by RVC-Boss GitHub

Need some help? 🤔

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.

More commands: -audio, -colabs, -overtrain, -realtime, -rvc, -help

[hexed cave]

-help

[azure marsh]

Command has expired

[hexed cave]

fuck

where's that doc

where it shows the links to download mp3-?

-guide

[azure marsh]

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

❔ English FAQ

https://docs.aihub.wtf/faq

Credits: AI HUB FRANCE

✨ More guides

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

How To Make an AI Cover With Ilaria RVC

Link: Rentry
Credits: 👽 Julia (ailen2091)

[candid crescent]

还有什么我可以尝试的吗？它似乎向我重播声音，但不在我的不和谐或游戏中播放

[dawn nebula]

你看一下你游戏的设置

[inner ridge]

does anyone know why its echoing

i have the echo thing turned off and when my friend speaks he said he can hear himself

[onyx ferry]

[candid crescent]

This is my same problem

[inner ridge]

idk why this is happening

[candid crescent]

yes, this man was helping me find solutions but i still cant fix it

[dawn nebula]

show settings

[inner ridge]

settings on the voice changer thing

?

[fluid matrix]

read this: https://rentry.co/VoiceChangerGuide

[finite galleon]

The og rvc v2 pretrains used 50 hours of audio data and RIN_E3 pretrains uses about 140 hours of audio data but again it all comes down to audio quality and some other stuff like the epochs, batch size, config settings, etc

[hot lagoon]

I did the 3. Install Mangio.
Copy this, paste it in the Terminal and hit enter:

wget https://huggingface.co/lollenape/LollenApeRVC/resolve/main/install.py

But I dont see Mangio?

[odd shale]

Refresh the folder.

And press this button

[zealous kiln]

any solution for the "JSONcodeError" problem?

[inner ridge]

anyone know how to fix the echoing problem when my friend speaks he said he can hear himself in the echo

[rare goblet]

Ayo? @inner ridge level 1 !!!

[onyx ferry]

[finite galleon]

The D and G files that you usually get from training models normally can be used as a pretrain for future models

[hot lagoon]

I did but is not showing me it

[azure marsh]

RVC Guides (How to Make AI Cover)

Documentation

🇺🇸 English (main)

https://docs.aihub.wtf/

​

Translation by country

🇧🇷 Brasil (PT-BR)

https://docs.aihub.wtf/v/brazil

🇪🇸 Spain

https://docs.aihub.wtf/v/spain/

🇫🇷 France

https://docs.aihub.wtf/v/france

🇩🇪 Germany

https://docs.aihub.wtf/v/germany

🇯🇵 Japan

https://docs.aihub.wtf/v/japan

🇰🇷 Korea

https://docs.aihub.wtf/v/korea

🇵🇱 Poland

https://docs.aihub.wtf/v/poland

🇷🇺 Russia

https://docs.aihub.wtf/v/russia

🇮🇹 Italy

https://docs.aihub.wtf/v/italy

[spare shale]

@viscid moss Sorry

Now the model has an index but it doesn't work

I found an Error Everything works

[brittle wing]

hey using applio for tts. where do i put my model? like idk what path it would be

[dawn nebula]

could u take a screenshot of ur whole screen

[rare goblet]

Ayo? @brittle wing level 5 !!!

[dawn nebula]

voice model and index file

[brittle wing]

yea but where in the applio folder do i put my model for it to show up there

[dawn nebula]

lemme check rq

[onyx ferry]

[dawn nebula]

logs\model

put both of them in

[half cove]

in the download tab of applio

it will have a option to drag and drop

put your pth then index

it will do it for you

[open flicker]

I need help

[spare shale]

?

yeah?

[open flicker]

File wont start up

[spare shale]

?

bat?

[open flicker]

yes

can we go into call

?

to show

[spare shale]

Try opening as Administrator

[spare shale]

Sorry but I can't 😅

[open flicker]

Doesnt work

I ran as admin

and it

wont open

?

[spare shale]

Maybe you can show me a Screenshot?

[open flicker]

it just flashes up on my screen right, and goes away

what do i do

[spare shale]

Hm..

What are you using?

[open flicker]

wdym

[rare goblet]

Ayo? @open flicker level 1 !!!

[open flicker]

I open that

and

it just flashes up on my screen

[spare shale]

i mean

[open flicker]

and goes away

[spare shale]

? 😅

[open flicker]

??

[spare shale]

I'm not sure, you better wait for another helper, sorry

[open flicker]

How do i open as administrator

[spare shale]

@half cove / @dawn nebula

[open flicker]

Ik how to run it

as admin

[dawn nebula]

you dont run with admin

@open flicker why ur folder only has start_http

[open flicker]

Idk

are any of yall able to hop in a call?

[dawn nebula]

where did you download the file

nah

[open flicker]

its that same file

[dawn nebula]

yeah

[open flicker]

HTTP BAT

[dawn nebula]

whats your gpu

[open flicker]

geforce rtx 3070 founders edition

the white on

e

[dawn nebula]

show me the folder name

[dawn nebula]

this is for amd gpus

bro

https://huggingface.co/wok000/vcclient000/resolve/main/MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.17b.zip

download this one above

[open flicker]

i just followed the yt

[gritty cargo]

i miss easy gui

[rare goblet]

Ayo? @gritty cargo level 3 !!!

[dawn nebula]

which yt

[open flicker]

some dudes

[dawn nebula]

well that dudes tutorial is wrong

[dawn nebula]

download this one, extract then run start http

[open flicker]

should i delete the other one

?

[dawn nebula]

yeah

[spare shale]

sweat

[open flicker]

does it effect anything ?

[rare goblet]

Ayo? @open flicker level 2 !!!

[open flicker]

on the pc

?

[dawn nebula]

nothing

[open flicker]

Or is it just some 3rd party application?

W

i appriciate it

[dawn nebula]

its just an app

[open flicker]

broski

[dawn nebula]

all g

[open flicker]

Most discord supports

dont do shit to help

yk

but yall actually take your time to help

Which is really good

[dawn nebula]

yeah ty

[open flicker]

how long yall been doing this, Just waiting for download btw

[dawn nebula]

10 months

nearly a year

[open flicker]

thats good

its still not opening

?

[spare shale]

Open as administrator

[open flicker]

how do i open it as that

or do i run

[spare shale]

Open as administrator

[open flicker]

I did that

[spare shale]

Right click and it should work

[open flicker]

Still flashes in and out

?

[spare shale]

@dawn nebula T-T

[open flicker]

Can you join call just to show

[spare shale]

I'm sorry, he still needs help

[open flicker]

and leave