✨│ai-help

[glacial pollen]

I mean, the calculation of time it takes per your target epoch is fine and most likely correct, that's just simple math

but what I try to say is, you can't calculate how many epochs is good for your model

it is a live; train n evaluate process

and you stop when you should stop ( the training ), afterwards, you evaluate your results and get the most optimal epoch

[modest abyss]

Yeah makes sense
Here's I guess a possibly easier question to answer. In the tutorial it says the "Train model" button will become a "Stop training" button, but mine never changed and pressing it again doesn't do anything. Any clues for that?

[glacial pollen]

Depends on what you use

there's many forks of rvc

and there's also og / mainline rvc

afaik, mangio was the fork to have that stopping button and maybeeee applio

but I am not sure as I do not use nor used applio

so dunno it's elements

[modest abyss]

I'm going off of the AI Hub Docs > RVC > Local > Mainline

[fading bolt]

Is it common lower chunk setting to experience difficulty with T, P and F sounds? a big issue for me is "Oof" cutting out or being missaid by the ai voice

[glacial pollen]

afaik ( unless something's changed as I do have my own rvc fork based on somewhat a lil outdated rvc mainline ) mainline didn't have it

I can be wrong yea, but if you don't see it, means it's not there

and I'd assume that's the case as fumiama's mainline version doesn't have it either

and it's what we deem as " updated " mainline from one of the devs

[modest abyss]

Is it possible that your dataset has little of those sounds? I remember making a Mickey Mouse sound a year ago and it had those issues but it was because Mickey barely said those things in the audio I used

[glacial pollen]

They mean real time voice conversion or w-okada

cause chunk is a setting used in there

[modest abyss]

Ah my bad

[glacial pollen]

but yea, in case you wanna stop manual and still have ability to resume the training

do so by closing the training console ( spam the close / cross button ) HOWEVER

make sure you're not doing it during " saving optimizer state "

or any other saving; including that of generator and discriminator files ( G, D )

• keep the log folder and most recent weight / epoch model
  (( oh yea, and same session parameters. You'd only input the name's model [ same as model's folder ] - just do not process anything as you'd do for the first run. just hit train ))
  that's all

[wispy lodge]

Chunk doesn't really affect that. Most likely because you're using noise suppression. If you use w-okada, sup2 can cause that

And you should export uploaded models to onnx

Also, #🔍│help-w-okada for wokada questions

[autumn talon]

What is that one website that uses text to speech and is free and has all the famous people on it

[modest abyss]

Alright, thanks! I'll try and tinker with the settings to see about the speed I suppose, but if not I'll just have it going for however long

[glacial pollen]

yuh

[glacial pollen]

Despite 1060's reputation of bad ish fp16 performance, you can still try to use it

I suppose it all comes down to " performance gains on that gpu isn't worth the accuracy tradeoff "

[autumn talon]

Can I help me

[rare goblet]

Ayo? @autumn talon level 1 !!!

[glacial pollen]

butttttt, if even some lil gains ( I suppose ) satisfy you, you can try

[autumn talon]

U

[glacial pollen]

Not quite free atm, as I am busy rewriting code and updating my fork

tho, what's the issue? if it's something quick thennnnn

[autumn talon]

What is that one website that uses text to speech and is free and has all the famous people on it

[modest abyss]

I don't mind the wait since I don't need to do my project IMMEDIATELY but I was just curious since it used to be faster. Either way thanks!

[glacial pollen]

To be honest, I am not quite sure what you mean. haven't heard of any site like that

Unless it's elevenlabs you have in mind ( doubt that's the thing but, check the site and yea

[shut abyss]

do i upload the onnx file in place of the pth file?

[autumn talon]

I have been on it

Idk what it is tho

[stray mortar]

anyone konw why my audio is echoing and glitchy?

[rare goblet]

Ayo? @stray mortar level 1 !!!

[glacial pollen]

Then I possibly can't help you

[fading bolt]

Thank you I’ll look into it!

[rare goblet]

Ayo? @fading bolt level 2 !!!

[proper shale]

might be mic bleed or something, enable sup2 and echo, increase s thresh till its gone

[wispy lodge]

1000 series has like really bad fp16 performance, so it will actually be devastating. Iirc the ratio is 1 fp16 flop to 64 fp32 flops (at least that's what it was for 1080)

[autumn talon]

Dos ai hub have a site?

[glacial pollen]

oh shit

then ye, not worth it

[wispy lodge]

1600 series has 1 to 2, so it's still bad

[glacial pollen]

The more you know as they say.
Glad you're telling me that as I am an ex amd user so, nvidia is still a fresh topic for me

[stray mortar]

still there, it changed a little but its still there

[glacial pollen]

You gotta ask someone else, not responsible for it nor using it

maybe ping some mods or helpers

[stray mortar]

and it cuts out slightly whe nechoing

[modest abyss]

Yeah my current GPU works fine for most games, trying to save up for an RTX 30 series

[wispy lodge]

You can replace pth file, yeah

[low shard]

ai hub has docs https://docs.aihub.wtf, guides for using RVC & TTS

[stray mortar]

anyone konw why its still bugged for me? eching and glitchy, cutting out, idk how to explain it

[glacial pollen]

3060 12 gig is really nice - no regrets whatsoever
be it alan wake 2 or ai lol

[shut abyss]

what does the index slider do? sorry if im asking too many questions

[glacial pollen]

It's accent / voice unique features / intonation / articulation and so on
( in a big simplified form - speaking )

[modest abyss]

Yeah I have 3060 as the one I was saving for. Though, this is before I knew how much the gigs mattered (I'm slowly learning more about PCs), so I might wanna make sure I have the right one linked

[low shard]

Maybe you are talking about https://elevenlabs.io/pricing , but has only 10k Characters (not voices) free per month, there are other free ways to use text to speech (tts).

You can use tts with rvc models, some forks such as ilaria rvc mainline & applio have built in tts (using Microsoft Edge TTS to make a generated tts audio, which i suggest you to choose a tts model that is the same gender and language of the rvc model you wanna use, and then convert it with rvc)

If you wanna do tts locally (if you got a good pc):

• You can find Ilaria RVC Mainline in #📰│dev-updates message
• While Applio in our docs

If you don't got a good pc you can do it online:

• Ilaria RVC Zero (Running on A100 GPU, free fasted rvc on cloud) and the guide

• Use Applio Colab (with google colab T4 free daily limit gpu)

• if you don't wanna use edge tts, you could try another tts ai in https://discord.com/channels/1159260121998827560/1212368971127590922 and use the output as an input in rvc

There's also GPT-SoVits which is way better but has only english, japanese & chinese, ofc u can ask me abt it

[glacial pollen]

Yea, I personally have this one:

But if you can get the triple fan, yeyy

But then, can't complain

[shut abyss]

Thank you

its working really good now after all the tips ❤️

[modest abyss]

Gotta make sure it can even fit in my case, looks pretty hefty
Also gotta make sure I have the correct fans n stuff cuz my room isn't the coolest

[glacial pollen]

Hmmm, it's not that big actually

[modest abyss]

Interesting
Idk, I'll be sure to look into it all once I get the money, I've got other stuff I have to pay before then so ye
Anyways I'm gonna stop having general chatting in the help channel, hop you have a nice one lol

[glacial pollen]

temps are quite good, at least on my side. but then, I have 6 fans and undervolting so

[glacial pollen]

Yea, true. You too man

[stray mortar]

anyone know what to do if i download a model that is rvc v2?

[burnt fable]

how do i stop hearing myself

[proper shale]

disable monitor

and set output to the cable

[proper shale]

just upload it, it shall work just fine

[stray mortar]

im still having the issue from prior with the echoing and it cutting out a little, if you know wnya fixes? thats why i was asking if theeres a specific way to apply a RVC V2 model

[proper shale]

what's ur settings rn

send a screenshot

[stray mortar]

one sec

[rare goblet]

Ayo? @stray mortar level 2 !!!

[stray mortar]

sent ss

could it be my cpu?

since my laptop is weak

[proper shale]

oh yeah that's probably the issue

you do need a decent dedicated gpu

[stray mortar]

goddamnit

anything i can do with what i have?

like limit apps n etc

im not a tech guy so

[proper shale]

Well, not something you can download

You can try this tho: https://colab.research.google.com/github/hinabl/voice-changer-colab/blob/master/Hina_Modified_Realtime_Voice_Changer_on_Colab.ipynb

[stray mortar]

so what is that?

is there like a yt vid on this so i can look tommorow?

[proper shale]

colab, runs code on cloud

and has a good gpu

so yeah

[stray mortar]

i havent a clue what any of that meant im not a tech guy 😅

could you maybe dm me the details on what to do? since im heading off to bed rn anyway

[proper shale]

https://docs.google.com/document/d/e/2PACX-1vTIceEcBfS6Zqolv_QEysrFfI_EJikPxozWptP_EjkpLVl-l-gdo-ijBonQMTviAHEYm5emmd9k9TdC/pub best thing you'll get tbh

[stray mortar]

and thanks for everything, you are appreciated

[proper shale]

basically, runs on the cloud, you get a good graphics card, so it runs welll

[chrome orbit]

im using applio rvc to train my own model and its taking about 1hr per epoch, thats too slow right?

[rare goblet]

Ayo? @chrome orbit level 2 !!!

[proper shale]

how long is your dataset and what's your batch size

[chrome orbit]

dataset is 2hrs 10 mins

batchsize is 6

[proper shale]

honestly, can be expected

[chrome orbit]

oh alr

damn finna take me years

[brittle wing]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[glacial pollen]

You shouldn't go for 2 hours

[chrome orbit]

wym

[glacial pollen]

What I mean is, it is not diffusion, it is hifigan

you should stick to max, I'd say, 50 minutes

but even that is a slight overkill

What I'd actually suggest is selectively pick the most diverse sections of your set, cap it at 45 - 50 mins

batch size 16, 14 or 12

or if diversity isn't as neat, stick perhaps to 8 or 10

[analog lava]

how can i send pics

how can i actually convert an ai cover step by step (first time srry)

[violet heron]

The plus button

[violet heron]

Read the docs maybe that were linked to you

!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.

[chrome orbit]

ight ill try it

thanks

[glacial pollen]

Np.
If the model's convergence doesn't go well, decrease the dataset's size gradually / and or experiment with smaller and higher batches ( increase or decrease by 2 or 1 )

Good luck ~ ✨

[elfin zealot]

Has anyone tried using vast.ai to rent a gpu to run rvc for training? I'm looking into it bc rvc disconnected on google colab just isn't cutting it with how much it disconnects

[edgy iron]

hey i need help

which one do i use too sound better?

[urban mesa]

can i use this set up for valorant

[knotty moth]

3060 12 gb is a good choice unless you can stretch your budget for more performance

[urban mesa]

do i need to change anything

;-;

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[knotty moth]

again check the settings

[urban mesa]

its all fine

;-;

[knotty moth]

so whats ur issue? not getting smooth valorant performance with ur 3060?

[urban mesa]

i mean i cant hear myself in val

;-;

so im not sure if vc worked

i cant either hear anyone in val

[knotty moth]

check the audio input/output in your valorant settings

[urban mesa]

oh ty

it worked

luv u :3

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for AMD GPUs

Don't forget that your models needs to be converted in ONNX!

F0 Det.: rmvpe_onnx (suggested for all series)

7xxx XT cards: 112-128 chunk | +16384 extra
6xxx XT cards: 128-192 chunk | +16384 extra
5xxx XT cards: 192-256 chunk | +8192 extra

RX 580: 192-256 chunk | +8192 extra
RX 570: 192-256 chunk | +8192 extra
RX 560: 256-384 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[silent stratus]

i have a really good deep voice carti data set that is 13 min and i want to train it. i was training at 400 epochs on rvc disconected colab but it syas the vram is all used up and its been doing that for like 2 weeks becasue of the limit. when i train the dataset locally it takes 16 min for 1 epoch becasue i have a gtx 1650. is there any other way to train or can someone train it for me and send it to me?

[timid heron]

hello. I'm Korean. Is there a free AI voice cloner tts API that can be used in Python?

[modest abyss]

Tinkered with settings and got it all the way down to 8 hours for the full train yippee
Hopefully it comes out good cuz I've spent the whole day on this 😭

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[lone coral]

hey guys i m looking for a voice effect similar to the one in this video please:
https://youtu.be/z3JXFAsXFBs?si=p1H2owUhPh24EBs7

any suggestions would be greatly appreciated

[brave garnet]

RVC Colabs and Spaces

⠀

Google Colabs

⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.

AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.

RVC Disconnected
To train new voice models, by Kit Lemonfoot.

EasyGUI
The OG interface, by Rejects.
⠀

Made with love by Ilaria 💖

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

RVC Colabs and Spaces

⠀

Google Colabs

⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.

AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.

RVC Disconnected
To train new voice models, by Kit Lemonfoot.

EasyGUI
The OG interface, by Rejects.
⠀

Made with love by Ilaria 💖

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for AMD GPUs

Don't forget that your models needs to be converted in ONNX!

F0 Det.: rmvpe_onnx (suggested for all series)

7xxx XT cards: 112-128 chunk | +16384 extra
6xxx XT cards: 128-192 chunk | +16384 extra
5xxx XT cards: 192-256 chunk | +8192 extra

RX 580: 192-256 chunk | +8192 extra
RX 570: 192-256 chunk | +8192 extra
RX 560: 256-384 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[brave garnet]

RVC Colabs and Spaces

⠀

Local Forks 🖥️

⠀
Mainline RVC
Original project, suggested for advanced users,
by the RVC-Project team.

Applio
Simplified, suggested for all, by the Applio team.

RVC Studio
Simplified, suggested for all, by SayanoAI.

Mangio-RVC
Simplified, may not be supported anymore, by Mangio621.

AICoverGen
Simple yet great way to make covers, by SociallyIneptWeeb.

Replay
From the greators of weights.gg, excellent product for everyone.
⠀

Made with love by Ilaria 💖

[gusty socket]

Hey, I've got this old model I've been using with Okada for some time now. Apparently, okada is deprecated and I wanted to migrate to RVC. Can I just put that model there?

[pastel oak]

Open model_dir folder and search for the model there

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[digital quail]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[low shard]

Yep it won't work because weights.gg requires to login to download the model so none cloud rvc would work to download them like huggingface models

To use weights.gg models, you need to login, download them on your device, rename the .index file into 'added_IVF296_Flat_nprobe_1_model_v2.index', then upload them into applio like in the pic https://docs.aihub.wtf/rvc/cloud/appliocolab-img/3-modelupload.png in the download tab

@nocturne mural why does it say downloaded sucessfully when downloading weights.gg models can't be downloaded as they need login tho?

i remember before it said before when there was an error in downloading a model

[pulsar anchor]

How to get rid of static noises in my ai voice model?

smh

Should I train my model in applio?

or maybe....

[rare goblet]

Ayo? @pulsar anchor level 6 !!!

[pulsar anchor]

Weights.gg is not the good place to train it

[knotty moth]

yeah applio or mainline RVC, or some colab spaces are better option

[pulsar anchor]

Are there any tutorial for this?

[knotty moth]

right in the pinned guide https://docs.aihub.wtf/

[pastel oak]

under RVC -> Local -> Mainline
or Cloud

[subtle olive]

what should i do, when i open the start_http it opens the cmd for a few seconds and then it closes

[pastel oak]

https://cdn.discordapp.com/attachments/1159290161683767298/1159607580184821760/issue.mp4?ex=668bb577&is=668a63f7&hm=003d2aa54b71cee4b08a4bea78776489facadc36eb929b4b8599019cdd12fdfd&

[sick sluice]

why does my voice changer sound like shit

also it lags

[pastel oak]

What GPU do you have

[sick sluice]

gtx 1660

[pastel oak]

Send screenshot of your voice changer

[sick sluice]

is it not strong enough

[brittle wing]

best UVR5 settings ?

[rare goblet]

Ayo? @brittle wing level 6 !!!

[sick sluice]

here u go

[subtle olive]

and it should stay like this for a while and then it should start typing something and thats the installation right

[pastel oak]

You are using the new alpha version thats unstable

Use the normal stable one

[sick sluice]

from where

[pastel oak]

https://rentry.co/VoiceChangerGuide#download-nvidia

check out the entire guide while ur at it

[sick sluice]

ok

thank you

[brittle wing]

@pastel oak hey shad

[pastel oak]

Should

[pastel oak]

Default are fine the differences are minimal

[brittle wing]

thank you

[subtle olive]

can i dm you? i cant send images here

[rare goblet]

Ayo? @subtle olive level 1 !!!

[pastel oak]

You can send pictures now

in here

[sick sluice]

what affinity should i put?

[pastel oak]

Did you unzip the folder

[pastel oak]

unselect everything and use 2 only

[subtle olive]

yes

[pastel oak]

but why are you doing that anyway download the voice changer first 💀

[sick sluice]

is this good?

[pastel oak]

yes

[pastel oak]

What GPU do you have

[subtle olive]

gtx 1660 super

[sick sluice]

gtx 1660 chads here

[subtle olive]

well i dont have it xd im trying to fix it for my friend

[pastel oak]

U also downloaded wrong version

https://rentry.co/VoiceChangerGuide#download-nvidia

[sick sluice]

hey shad is this a good version?

[subtle olive]

so this one for 1660 super?

[pastel oak]

Ye

[pastel oak]

Ye

Select the GPU

[sick sluice]

okay

[pastel oak]

f0 det: rmvpe
chunk: 192 maybe 128 if good
extra: 8192

input microphone
output line 1 (step 6 on the guide download if necessary)

[subtle olive]

the blue thing?

[pastel oak]

click on version 18a gpu cuda

[regal trout]

can i hear myself

[pastel oak]

?

[sick sluice]

thanks shad its a lot better now

[subtle olive]

btw why cant i set crepe

only crepe_tiny

[pastel oak]

crepe bad

why do you have these though arent you on the 1660 super

[subtle olive]

no

it was my friends issue

[pastel oak]

what do you have

[subtle olive]

gpu?

[pastel oak]

ye

[subtle olive]

rx 6600xt

[pastel oak]

rmvpe_onnx

best

[subtle olive]

bet

thanks dude

[pastel oak]

you can try crepe_tiny aswell, sometimes its better it depends on the voice model

[subtle olive]

and this is set right?

[distant turtle]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[rare goblet]

Ayo? @distant turtle level 6 !!!

[pastel oak]

If your audio picks up a lot of noise from background you can use that on the full right yes

if not then dont

[subtle olive]

what about the in, out and index?

i have the egirl realistic model from here

[pastel oak]

in increases volume of your own mic going into the voice changer
out increases volume of your voice changer
tune is pitch so if u are man and use egirl you have to play around with this until it sounds realistic, usually 10 - 12 something will sound good but it depends on model and your voice

index leave at 0 thats the accent of the voice model

[subtle olive]

oh okay thanks

[rare goblet]

Ayo? @subtle olive level 2 !!!

[subtle olive]

the voice changer is kinda laggy for me

[sick sluice]

how do i deal with it detecting noise from the room and making weird sounds

also the delay is long

[pastel oak]

enable sup2 and move s.threshold to the right

[pastel oak]

you can try chunk 128 or go lower

[sick sluice]

im already on 128

[rare goblet]

Ayo? @sick sluice level 2 !!!

[pastel oak]

there will always be delay

[sick sluice]

oh alright

[glossy spindle]

not really help with rvc but i'm doing /usermodels and it says the application didnt respond

[sick sluice]

why doesnt it sound as good as the youtubers

like its kinda obvious its AI

[pastel oak]

Not every voice model is good for realtime

search for asmr, streamer, egirl, mommy

[sick sluice]

which is good

ok

[subtle olive]

shad how should i set the extra?

[pastel oak]

16384

u also have to always export every model to onnx as an amd gpu user

youre better off using emojikages modified wokada version to bypass all that ngl amd gpu on normal wokada sucks

[subtle olive]

bro what

[pastel oak]

XD

[subtle olive]

whatt is onnx

like

where is it

[pastel oak]

when you upload a model

there will be a grey button called export to onnx

next to save setting

have to click that, then it generates the onnx file, then you upload that onnx file the way you upload a normal voice model

[pastel oak]

now you upload, then when you select it theres a button called export to onnx

you know what just download the other voice changer pls

LMAO

[subtle olive]

like something here?

[pastel oak]

No

go back now

select the model like you would start using it

but theres a grey button on the mid right above chunk

[subtle olive]

oh i see it

export to onnx

[subtle olive]

which bruh

[pastel oak]

https://github.com/deiteris/voice-changer/releases/download/b2100/voice-changer-windows-amd64-dml.zip

[subtle olive]

then i dont have to export to onnx?

[pastel oak]

correct

it also runs better

[subtle olive]

alright

when i start the voice it takes 100% of my cpu

[pastel oak]

yea thats a bug with normal wokada, it doesnt happen on this new one i sent u

as i said, amd gpus and normal wokada are not great

[subtle olive]

yoo it only has 450mb

thats great for my 20mbit internet

[pastel oak]

wait until you know it installs some files through the command window when you first launch it...

[subtle olive]

cant wait dude

https://media.discordapp.net/attachments/964758239780675605/1181423304087711855/togif.gif

:/

[sick sluice]

shad

what about these settings

[subtle olive]

shad wtf i do now

do i run the mmvc server app

?

[pulsar anchor]

Did I do it right?

My first model was 80 epochs

now I will do it again at 500 epoches

I am a newbie

[sick sluice]

how much SR should i put

[wispy lodge]

Just MMVCServerSIO, yes

[tidal trail]

check your datase path

[pulsar anchor]

huh

[tidal trail]

it should be assets/datasets/datasetfolder

do not put it on a zip

[pulsar anchor]

...

oh

okay

[sick sluice]

why doesn't my razer seiren mini mic not appear when i choose server

it is there on client

[pulsar anchor]

Like this?

[tidal trail]

also 3.2.1 is avaible and do not put in D drive root, should be D/foldername/Applio 3.2.1 or or for more security in C/foldername/Applio 3.2.1

[pulsar anchor]

Do I also need to update Applio too omg

smh

[sick sluice]

nvm its there

[pulsar anchor]

Are you serious.

[subtle olive]

tf do i open now shad

[sick sluice]

can soemone answer this

[pulsar anchor]

Then what's the point of me installed Applio like that ;-;

sigh

How am I gonna update Applio btw?

[subtle olive]

@pastel oak

[tidal trail]

yes, make sure your audio do not contain spaces

[pulsar anchor]

whatthe fuck

you mean

file's name?

[tidal trail]

boyfriend voice dataset, just put boyfriend

[pulsar anchor]

Ah

hold on

Well shit

[wispy lodge]

You just run MMVCServerSIO and it should open the browser with voice changer UI

[pastel oak]

start 0.2 end 0.8
trancate 300

rvc quality LOW!!! important else your pc dies for no difference xd

[patent trellis]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[pulsar anchor]

I downloaded the 3.2.1

but the zip file is 3.2.0

Is this right?

[tidal trail]

https://huggingface.co/IAHispano/Applio/resolve/main/Compiled/ApplioV3.2.1.zip?download=true

[pulsar anchor]

oh ty

[subtle olive]

shad how do i know which extra

[rare goblet]

Ayo? @subtle olive level 3 !!!

[subtle olive]

in this voicechanger

[pastel oak]

Try 4 or 5

shouldnt matter too much

[subtle olive]

why do i hear myself but it doesnt work on discord

[patent trellis]

how do i seperate the music and vocal from a mp3 file

[brisk oasis]

Is rx 5700 capable of running this programm without it lagging?

[pastel oak]

did you download virtual cable

[subtle olive]

yes

should i switch this to server?

[pastel oak]

no

[subtle olive]

to make it work on discord etc

[pastel oak]

select Line 1 on output

[subtle olive]

what

so back to client?

so its like this now

[pastel oak]

Output: line 1

[subtle olive]

i mighht be blind gang

btw how much extra on the 1660 super

[pastel oak]

Then you didnt install virtual cable

[pastel oak]

8192

https://software.muzychenko.net/freeware/vac470lite.zip

VAC by Muzychenko you should run setup64 after extracting the zip to a new folder not the 64a just the regular 64, you do not need admin for it, and you do not need to restart your computer.

[warped star]

help

[subtle olive]

i did

[pastel oak]

Then you should be able to select Line 1 on your output

[subtle olive]

in this should be line 1?

[pastel oak]

on OUTPUT

yes

[pastel oak]

cable input

[subtle olive]

alr

and it should work now?

yoo it works now

its fine

should i use crepe on gtx 1660?

[wispy lodge]

No

Use rmvpe or fcpe

[subtle olive]

and do i use the thresh or some of the three noise options

[subtle olive]

shad why is it lagging for my friend when he talks on discord and when he tries it on the voice changer for himself it sounds fine

[pastel oak]

does he also use CABLE?

its shit

do this

https://software.muzychenko.net/freeware/vac470lite.zip
VAC by Muzychenko you should run setup64 after extracting the zip to a new folder not the 64a just the regular 64, you do not need admin for it, and you do not need to restart your computer.

[subtle olive]

like as the output in the voice changer?

[pastel oak]

and do this

open task manager hit details

right click audiodg.exe and set Priority to HIGH

right click it again and press set affinity and only have an EVEN number CPU core set, 2 is what I use. AKA everything unchecked except CPU 2

[subtle olive]

he does

so this will eliminate the lags in discord vcs?

[pastel oak]

ya

[subtle olive]

alright thanks

he cant find audio dg in task manager

oh its here

hes a retard

after he checks only cpu 2 it should work now?

[rare goblet]

Ayo? @subtle olive level 4 !!!

[pastel oak]

should

[subtle olive]

it does

and how do we set it into valorant

like what input do i put

and output

[pastel oak]

select Line 1 as your input

on valo

[subtle olive]

so

the

virtual cable

[pulsar anchor]

What is this?

Embedder Model

@tidal trail

[tidal trail]

use contentvec

[pulsar anchor]

oki

What about this?

btw

nvm

it should be 4

I did do everything right for training model

Why I feel like nothing happened?

?????

Help

What is this error? @tidal trail

I did everything correct

but tensorboard doesn't appear like this

[tidal trail]

the link is above

[pulsar anchor]

This is my tensorboard

Why it keep saying no path found but not start to generate with eposch?

Is it becuz I updated Applio so it will caused like this?

[pulsar anchor]

What link

[tidal trail]

to enter the tensorboard

[knotty moth]

try tensorvenv.bat from the guide https://docs.aihub.wtf/

[pulsar anchor]

The tensorvenv

hold on I can't find it

oh nvm

found it

WHY THE FUCK IT TAKE 8 MINUTE FOR 1 EPOCSH

DON'T TELL ME IT USES WRONG GPU

and great

the new update of Applio doesn't allow us to change GPU setting

What a fucking suck ass update

[burnt fable]

Guys ever since i got this and the cable thing my mic has echo now

how do i fix

[plain fulcrum]

Hey guys I was wondering if it's possible to download models from weights and continue training them with my own dataset?

[tidal trail]

advanced settings

[pulsar anchor]

Nope

It doesn't exist

wait

[tidal trail]

yes

[pulsar anchor]

Oh I see it

I'm blind

plz fkin help me lmao

[violet heron]

applio moment

[violet heron]

No because they don’t provide any of the training data

[pulsar anchor]

I'm trying to make my first model

[rare goblet]

Ayo? @pulsar anchor level 7 !!!

[pulsar anchor]

Nothing happened at all

0 Eposch appear

Zero

It do nothing

[plain fulcrum]

So I can't use the pth file If I can get it?

[violet heron]

You can

[plain fulcrum]

Oh

[violet heron]

You said “train” not infer

inference is making the ai cover

Training is making the model

[plain fulcrum]

No no I'm trynna add more data to the already made model

Cus the model isn't that good and I was wondering if I could like improve it

[violet heron]

If you know the frequency the model was made in you can merge 2 together

Or you can ask the creator if they still have the training data assuming they made it on RVC

[plain fulcrum]

So I can make my own model and combine it with the model from weights?

[violet heron]

Yes

But you need to know what frequency the model was made in

[pulsar anchor]

WHY IT TAKES 49 MINUTES TO TRAIN 1 EPOSCH

WHAT THE FUCK?

I fucking hate this shit.

[violet heron]

And train your model using the same one

[violet heron]

What’s your GPU

I would just go with original RVC instead

[violet heron]

Might be why

[pulsar anchor]

Urgh

Then how can I gonna easily make AI Voice Model for free?

Weight.gg is cool but it's kinda mid

[violet heron]

-colab

[azure marsh]

Suggestions for @pulsar anchor

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[plain fulcrum]

Oh no I was thinkin of training stuff with my CPU but after seeing this idk😭

[violet heron]

You can use a google colabatory instead

[pulsar anchor]

How to use Google Collab?

I've never use it in my entire life lol

[violet heron]

You run each cell and read

You only get a few hours of runtime but it works depending on your dataset

[plain fulcrum]

Can I use it with my CPU

I don't have a gpu

[violet heron]

google colab runs off a googles powerful APUs in the cloud

So it doesn’t use your resources

[plain fulcrum]

Oh that's nice lol

But u said it only runs for a few hours

So wouldn't that be an issue

With training

[violet heron]

How long is your dataset?

[wispy lodge]

Free Colab provides ~4 hours of GPU time

For me it was enough to train for 500 epochs

[pulsar anchor]

What is this?

[plain fulcrum]

I don't have one right now but I think I could get like a 30 min clean dataset

[pulsar anchor]

You see

[wispy lodge]

With 8 minutes dataset

[pulsar anchor]

when I try to use colab

this happened

[violet heron]

Use half of that

It should fully train if you train for 300 epochs

[burnt fable]

Guys ever since i got this and the cable thing my mic has echo now
how do i fix

[rare goblet]

Ayo? @burnt fable level 1 !!!

[plain fulcrum]

And I could do the 2nd half after that right

[violet heron]

Combining the models?

[plain fulcrum]

Yeah that's what I'm trynna do, my theory was I could add my dataset in 10 min intervals so I don't have to worry about the colab runtime, just gradually adding to the original model

[violet heron]

Model fusing is alr kinda weird so not sure if that would work

Why do you need to add onto the weights model

I’ve trained models in a single colab runtime before

[misty monolith]

im thinking about training my own model and im brand new to this type of stuff, how much space will it take up on my pc to do this type of stuff

[violet heron]

Good few GBs

Shouldn’t really take any more then like 9 depending how much models you have loaded or what your saving

[misty monolith]

hmmm that's actually good, thank you so much

[pulsar anchor]

Soo is this Applio but colab?

What should I do next?

[violet heron]

Train the model

You will need to upload stuff onto the colab tho

[pulsar anchor]

Upload stuff to colab?

huh

LIke this?

[sick sluice]

how much S.R do i put in server?

[violet heron]

Never used the applio colab

[thin stump]

https://docs.aihub.wtf/rvc/cloud/applio-colab/ they have a whole guide here

[violet heron]

Yeah should read the guide the other user linked above

[plain fulcrum]

Just trynna make the model better, if there's a way I can just make a model myself then that's better but I thought if there's already a model that's pretty good why not just build on top of that

[rare goblet]

Ayo? @plain fulcrum level 3 !!!

[plain fulcrum]

And the fact that I only have a CPU I thought I wouldn't be able to train anything fully on my own

[violet heron]

You won’t be able to train anything on cpu period

[plain fulcrum]

Wait what

Um

So I'm cooked?

I trained 1 model with like a min of data with 50 epochs once just as a test and It did it work

It just sounded horrible but it worked

[urban flint]

what does batch mean

[violet heron]

Like is said you can use colab

[plain fulcrum]

Oh so I can train models

[violet heron]

1 minute isn’t enough and 50 epochs would undertrain that

[violet heron]

Yeah just using google colab

[plain fulcrum]

Yeah I know the result was just as bad as I expected I was trynna see if I could train anything

[plain fulcrum]

Alright one more question, how do I know what epoch number to go for with my model

Say I have 30 mins of clean audio

To make sure the model is AS GOOD as possible I don't care how long it takes

[sick junco]

can someone answer this question, im using RVC GUI for windows and whenever i convert my WAV its not actually playing it in the windows Media Player, its just stuck at 0:00, and you can still slide the bar to move forward in teh recording but its not actually playing the AI converted W

[violet heron]

RVC-GUI is outdated

-local

[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

Credits to Faze Masta and Antasma for compiling these links.

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

[sick junco]

i got it working now i just had to use a different media player

i used audacity

[violet heron]

gl with raspy audio tho

[sick junco]

whats a better one

w okada cant convert audio files right, you can only speak into the mike

at that time

i want to record on my phone somehwere else and import the file into w-okada

@violet heron

[violet heron]

Mainline RVC

[sick junco]

can you send me the link

and how do i get it

[violet heron]

-local

[azure marsh]

Suggestions for @sick junco

🖥️ 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

Credits to Faze Masta and Antasma for compiling these links.

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

[violet heron]

aka original RVC

[sick junco]

oh alr

thanks

are there any tutorials on how to install it

[violet heron]

Go to releases, download the version depending on your GPU, unzip, run the bat file

[sick junco]

so this one

[rare goblet]

Ayo? @sick junco level 2 !!!

[violet heron]

If you have a AMD GPU yes

[sick junco]

alright

ty

[violet heron]

Just a little note RVC does use Cuda so it won’t use your GPU

[sick junco]

thats fine

[violet heron]

So you can’t train with it either

[sick junco]

i just want not raspy audio

[violet heron]

Use that then

Also make sure you use clean vocals with no music

[sick junco]

also do you know how to tell if a model is accurate or not, like how accurate it is to the thing its trying to impersonate

[sick junco]

i will

[violet heron]

The samples provided by the creator

And also just testing

[sick junco]

okay

[violet heron]

It’s also good to adjust the index ratio since that controls accent strength

[worthy lantern]

why this take so long 💤

[sick junco]

Oh thats really good because I'm trying to make an AI drake song and my accent is very different from his 😂

[distant turtle]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[distant turtle]

why this error?

[brave garnet]

Voicechanger Settings (Okada)

⠀

Settings for AMD GPUs

Don't forget that your models needs to be converted in ONNX!

F0 Det.: rmvpe_onnx (suggested for all series)

7xxx XT cards: 112-128 chunk | +16384 extra
6xxx XT cards: 128-192 chunk | +16384 extra
5xxx XT cards: 192-256 chunk | +8192 extra

RX 580: 192-256 chunk | +8192 extra
RX 570: 192-256 chunk | +8192 extra
RX 560: 256-384 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[brave garnet]

Voicechanger Download (Okada)

⠀
Download for Nvidia GPUs
Version 18a cuda

Download for AMD GPUs
Version 18a directml

Download for Intel GPUs
Version 18a directml

Download for Mac
Version 17b Mac
⠀

Made with love by Ilaria 💖

[rain pasture]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[empty raptor]

How would I easily make backing vocals for my ai covers please help

[cedar badge]

question, how do i fix the problem with high tones voices in a cover

[hollow wasp]

how do I go about training a model to my own voice?

[low shard]

https://docs.aihub.wtf/essentials/how-to-make-voice-models/

[primal vector]

Hey can anyone help me try to find a good Vocal Remover for backing vocals? I'm trying to make an a.i. cover of Jump Up Superstar but I can't seem to get rid of the backing vocals in this section(s)

[hollow wasp]

@low shard hey I downloaded the tensor thing but it doesn't do anything after I started it

[rare goblet]

Ayo? @hollow wasp level 1 !!!

[hollow wasp]

anyone knows how can I train an already available voice model to my own voice, like an index file.

[tidal trail]

UVR-BVE-4B_SN-44100-1

in uvr colab or mvsep

[primal vector]

It didn't seem to work. Does the HuggingFace version work? Cause that's what i'm using

[rare goblet]

Ayo? @primal vector level 7 !!!

[tidal trail]

you mean uvr ui

technically they are the same

[primal vector]

I think so? It's this one.

[tidal trail]

#📰│dev-updates message

but i think they are the same

[primal vector]

I'll try it rq

[primal vector]

I can't find UVR-BVE-4B_SN-44100-1 but I see 13_SP-UVR-4B-44100-1

[tidal trail]

well i think uvr no ui has it

[primal vector]

Could you send a link?

[tidal trail]

https://colab.research.google.com/github/Eddycrack864/UVR5-NO-UI/blob/main/UVR5_NO_UI.ipynb

#📰│dev-updates message

[low shard]

uhh are you training locally? Sorry but it might be better you ask someone else for local things as i do it only online

[brave garnet]

RVC Colabs and Spaces

⠀

HuggingFace Spaces 🤗

⠀
Ilaria RVC
EasyGUI port with some improvements, by Ilaria.

RVC-HFv2
Applio port, by r3gm.

AICoverGen
AICoverGen port, by r3gm.

Advanced RVC Inference
Extended version of the GUI with advanced settings, r3gm.
⠀

Made with love by Ilaria 💖

[pulsar anchor]

@tidal trail So rn I am using Applio Colab

Should I tick this?

GPU Setting

[tidal trail]

Leave as it is

@pulsar anchor

[pulsar anchor]

Then I shall only do this

[tidal trail]

Yep check that

[pulsar anchor]

Alright

START TRAINING

Doing good so far?

[pulsar anchor]

Where is my index file

T_T

[flint dew]

hola necesito ayuda

[pulsar anchor]

@tidal trail What is need at least one array to concatenate??

[flint dew]

i need help with step 2 pls

[rare goblet]

Ayo? @flint dew level 1 !!!

[flint dew]

Does anyone know what I can do about this error?

🤔

[warped ravine]

For some reason when I click train feature index it just says error in the output information

Has anyone else encountered this issue???

[violet heron]

What’s your GPU

[violet heron]

https://media.discordapp.net/stickers/1250559937650757672.png?size=160

[warped ravine]

Says nothing, just Error

[rare goblet]

Ayo? @warped ravine level 1 !!!

[warped ravine]

RTX 3060

[violet heron]

Did you close anything like the cmd (you shouldn’t have)

[warped ravine]

Nope

[violet heron]

Did you enter your dataset path?

[warped ravine]

Yep, even had the issue of putting spaces in the name but corrected it

[violet heron]

Any special characters in your name?

[warped ravine]

Nope

[violet heron]

Is RVC located on the drive your booting from

[violet heron]

Also make sure the GPU index matches your 3060

[warped ravine]

Yes

[warped ravine]

Yeah, it's a Laptop I don't have other drives rn

[violet heron]

Wait did you download the INFER_TRAIN version of Mangio

[warped ravine]

Yep

[violet heron]

Any error in command prompt?

[warped ravine]

RuntimeError: Error in __cdecl faiss::FileIOWriter::FileIOWriter(const char *) at D:\a\faiss-wheels\faiss-wheels\faiss\faiss\impl\io.cpp:98: Error: 'f' failed: could not open C:\Users\Alienware\Documents\RVC\Mangio-RVC-v23.7.0_INFER_TRAIN/logs/DeadpoolEspañollatino/trained_IVF460_Flat_nprobe_1_DeadpoolEspañollatino_v2.index for writing: No such file or directory

[violet heron]

Try training the index

[warped ravine]

That's the issue, the error comes up when I click train index

I say make an Index, and it says can't make index cause there's no index

[violet heron]

Try restarting Mangio, reinput everything how you had it then try again

[warped ravine]

Alright

Thanks

[violet heron]

Hopefully that works

[violet heron]

If that didn’t work it might be the accent in the models name

[last marsh]

how do you do ai voices with text?

i have ntts

how can i do voice

[knotty moth]

the path may be too long, try move the installation to, e.g. D:\Mangio-RVC

[distant turtle]

-local

[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

Credits to Faze Masta and Antasma for compiling these links.

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

[night rune]

Hi, is there any way to train models not on my pc?

[knotty moth]

-colab

[azure marsh]

Suggestions for @night rune

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina 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
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

More commands: -audio, -hf, -local, -kaggle, -overtraining, -realtime, -rvc, -help

[night rune]

Did Colab stop blocking? (sorry, I haven't been in this field for a long time)

[rare goblet]

Ayo? @night rune level 1 !!!

[knotty moth]

wdym?

[night rune]

A year ago, after 5-15 minutes of work, colab just ended the session.

[wispy lodge]

There's no such problem anymore. At least not with existing Colab notebooks

It might end session sometimes if you upload files, but during training - no

[knotty moth]

it's now up to 4 hours of daily GPU limit

[night rune]

Help?

Now I realized that I forgot how to make models on easyGUI, huh

[night rune]

Okay, I figured it out myself

[astral jungle]

Only upload audio files, not zip files

=)))

[night rune]

Yeah, I've already realized that

[astral jungle]

Okay

[strange reef]

is there any rvcs that could help with vocal removing?

or maybe even help with doing only lead vocals?

[knotty moth]

no, RVC cant do that thing, but you can try a tool for that purpose https://huggingface.co/spaces/TheStinger/UVR5_UI

select a model like this

[ornate ore]

i have a question

in Voice ai. some models have been trained more than once but mine says trained1 uses 75

[rare goblet]

Ayo? @ornate ore level 1 !!!

[ornate ore]

can i train it more than once? and if so,how? ive been stumped on this for awhile

[knotty moth]

don't waste your money on that outdated & paywalled app, instead I'd recommend mainline RVC or Applio whose link can be found in the guide https://docs.aihub.wtf/

[turbid rampart]

hi i've apparently been using an outdated version of the easygui colab for ai covers, is there a newer version i can use? thanks

[pastel oak]

#✨│ai-help message

"inference" is ai covers

[proper sapphire]

How should you prevent voice cracking when making covers

[turbid rampart]

tysm

[knotty moth]

for the cloud inference solution, consider this https://huggingface.co/spaces/TheStinger/Ilaria_RVC

[proven hill]

https://docs.google.com/document/d/1YbXcLFPaGjhOdG5NFkK3QrucCEpHZBwFUxkeMO8aB18

[mild hornet]

Hey guys, can anyone share the latest Google Collab link of RVC client?

[brittle pelican]

['infer/modules/train/extract_feature_print.py', 'cuda:0', '1', '0', '0', 'C:\\Users\\\Downloads\\RVC1006Nvidia\\RVC1006Nvidia/logs/AlvaLorenz', 'v2'] C:\Users\\Downloads\RVC1006Nvidia\RVC1006Nvidia/logs/AlvaLorenz load model(s) from assets/hubert/hubert_base.pt ['infer/modules/train/extract/extract_f0_rmvpe.py', '1', '0', '1', 'C:\\Users\\Downloads\\RVC1006Nvidia\\RVC1006Nvidia/logs/AlvaLorenz', 'True'] no-f0-todo move model to cuda no-feature-todo ['infer/modules/train/extract_feature_print.py', 'cuda:0', '1', '0', '0', 'C:\\Users\\Downloads\\RVC1006Nvidia\\RVC1006Nvidia/logs/AlvaLorenz', 'v2'] C:\Users\\Downloads\RVC1006Nvidia\RVC1006Nvidia/logs/AlvaLorenz load model(s) from assets/hubert/hubert_base.pt move model to cuda no-feature-tod

Why it say no feature ? :"))
I training my model's index file

[hasty nimbus]

I got a juice world ai but where do I use it to change my vocals to him

[dry shard]

my pc is good but the voice is choppy asl and delayed

[pastel oak]

What GPU and send screenshot of the app

[dry shard]

4060

the voice kinda sounds like its tryna make some words even if i just scream into the mic idk

and all the voices sound the same as if theyre tryna be japanese or sum

[ornate ore]

if i have any questions,can i ask you for help?

[knotty moth]

https://dontasktoask.com/ just ask here

[dry shard]

dawg can u respond please 😿

[pastel oak]

Wrong version of wokada

https://rentry.co/VoiceChangerGuide

get nvidia

[glacial pollen]

4060 8 or 16gb?

[dry shard]

8

[dry shard]

wtf is a wokada

[glacial pollen]

lol

that's what you use my dude.

w-okada

Either way, refer to the char

[dry shard]

i used a tutorial and downloaded the version the tutorial said to smh

i aint into ts

also ty

[glacial pollen]

Sometimes it's better to find the source rather than referring to outdated tutorials

or perhaps those that have too much of misinfo

[dry shard]

thats sigma of you, my skibid rizz

[glacial pollen]

✨

If the performance will still suck for you, try my fork of rvc's native voice changer ( just let me know

[pastel oak]

Follow the guide I linked

[glacial pollen]

Also that ^

[sick junco]

i need an AI voice conversion tool where i can drop my audio file and then it will convert it

not live audio conversion

[brittle pelican]

I need help

[brittle pelican]

and my GPU is "NVIDIA GeForce GTX 750 Ti", GPU information said "Unfortunately' there is no compatible GPU avaible to support your training"

[tidal trail]

Your dataset path is wrong maybe, it should be /content/drive/MyDrive/dataset folder (not zip)

Or you spent your daily free gpu

[dry shard]

its good except that it has echo

how do i get the fork

[glacial pollen]

but like echo echo, a legit echo / delay

or something that sounds like " echo "

if it's the former, it's the model's fault

[dry shard]

when i say a word, it comes out like 5 times

[glacial pollen]

show me your settings

chunk etc

[brittle pelican]

I have question, why I cant send the picture? ;-;)

[glacial pollen]

Hmmm, Polish user?

but aside that, you're quite competent or nah?

You can first try to experiment with settings until you move to my fork

as it does require some work to get working, so better exhaust your options first

oh

lmao, my man @dry shard

why you have cpu selected

🤨

You'll obviously get stutters and such if you run it in cpu mode

[dry shard]

it was set by default

mate

[glacial pollen]

Default is default, doesn't hurt to read

[dry shard]

ok

[glacial pollen]

Intuition bruh, why else you'd need a gpu ( and so, use the chart for

No hate but like, there's too many people that do not read properly / not follow guides 1:1

and then mention issue X1 or X2

[dry shard]

also i gotta mention that the voice does not sound like the model preview at all

every model sounds kinda the same

[glacial pollen]

The " model preview " you're referring to is most likely from weights.gg

their " preview " unless changed recently ( I haven't visited the site ) is bs crap

doesn't show the true model's potential

that's 1 thing, 2nd thing is, you gotta " fit " into the speaker

[dry shard]

its not really the bad part of it not sounding like the preview, its every model sounding the same

[glacial pollen]

impersonation is an important aspect

Aaaaand

setting the pitch is also important

if you're a male trying out female model for instance, you typically raise the pitch by 1 octave

so 12 semitones ( up / positive )

[dry shard]

done that already

[glacial pollen]

then that's the model's fault

if it's not up to your expectations

[dry shard]

also the gpu and cpu settings sound the same

[glacial pollen]

do you have ur vc routed externally?

able to pass the sound to discord

or if you can record it and post here

I'll be able to tell more

Gotta hear it to say exactly what's the issue

also, before you set it to gpu, you should stop the program

[dry shard]

atp dawg its not finna work i tried it on 2 different pcs and aint nothing is working its just bs and if nothing is gonna work yall probably finna blame my pc

[glacial pollen]

set it to gpu just then, save the settings and ideally, restart the thing

Another thing is, which release you downloaded?

There's different versions per system configuration

https://huggingface.co/wok000/vcclient000/resolve/main/MMVCServerSIO_win_onnxgpu-cuda_v.1.5.3.18a.zip?download=true

Confirm if that's the one you use

onnxgpu-cuda

[brittle pelican]

Codename, can you help me ?

[glacial pollen]

mh?
what is it

if it's too lengthy, I'll pass as I am busy

if it's short, sure, Imma try

[brittle pelican]

My GPU is NVIDIA GeForce GTX 750 Ti, but when I go to RVC WebUI to train index file, the GPU information said "Unfortunately, there is no compatible GPU available to support your training."

[rare goblet]

Ayo? @brittle pelican level 1 !!!

[glacial pollen]

I mean... 750 ti

[rapid knot]

can some1 help me? the voicechanger does not seem to be working i have plugged everything correctly (including the input output) and assigned them correctly but still doesnt work

[glacial pollen]

I wouldn't be surprised if its not supported at all

Do you have all that's required installed?

pytorch with corresponding cuda

[rapid knot]

oh i needed pytorch?

[glacial pollen]

Not talking to you m8, context.

[rapid knot]

oh mb ig

[brittle pelican]

yes

[glacial pollen]

@brittle pelican

run this

should output you whether it's available or not ( or rather, whether it's being detected as viable or not )

open up cmd wherever you put the check script and in cmd:

py check.py

pro tip

you can open cmd in folders like so

via address bar

also, you can check ( again in cmd )

pip show torch

if it has +cuXXX in it

you're good and something's up with your rvc

[brittle pelican]

it said "'pip' is not recognized as an internal or external command,
operable program or batch file."

[glacial pollen]

and the previous one?

the check script

[brittle pelican]

wait me

[glacial pollen]

alr

[brittle pelican]

C:\Users\ num_gpus = torch.cuda.device_count()
'num_gpus' is not recognized as an internal or external command,
operable program or batch file.

C:\Users> print(f"Number of GPUs: {num_gpus}")
Can't find file (fNumber of GPUs: {num_gpus})

C:\Users>
C:\Users> for i in range(num_gpus):
i was unexpected at this time.
C:\Users> print(f"GPU {i}: {torch.cuda.get_device_name(i)}")
Can't find file (fGPU {i}: {torch.cuda.get_device_name(i)})

C:\Users> print(f" - Memory Allocated: {torch.cuda.memory_allocated(i) / 1024**2:.2f} MB")
No file to print

C:\Users> print(f" - Memory Cached: {torch.cuda.memory_reserved(i) / 1024**2:.2f} MB")
No file to print

C:\Users> else:
'else:' is not recognized as an internal or external command,
operable program or batch file.

C:\Users> print("No CUDA-capable devices detected.")
Can't find file (No CUDA-capable devices detected.)

C:\Users>
C:\Users>if name == "main":
The syntax of the command is incorrect.
C:\Users> check_pytorch_cuda(

[brittle wing]

Will I experience any voice delay improvement on rvc if I swap out my gtx 1660 super with a rtx 3060

[glacial pollen]

Ofc

[brittle pelican]

here

[glacial pollen]

You did not run it properly

it has to be run via python

py check.py

is all you write in the cmd

Like I showed you on the screenshot

1. Open up cmd in a folder where you put the python script

py check.py

like so ^

and you press enter

do not copy paste it to commandline ( content of the script ) no, just follow instructions

[glacial pollen]

I can get quite low delay on my 3060

( not using w-okada tho but native rvc's vc )

I'd say, for majority of use-cases

I can get it to around half a sec to ( in worst scenarios ) 1 or 1.2 sec (( tho, that did not happen to me ))

[brittle wing]

0.40-0.45 sample length already works well with my gtx 1660 super

[pastel oak]

i use 0.13 on 3060 ti

3060 can probably do 0.15-0.18

[brittle wing]

0.35 would just be maxing out my gpu memory

[brittle pelican]

okay, it said: 'py' is not recognized as an internal or external command,
operable program or batch file.

[glacial pollen]

actually, imma check the max for 3060

wait honda

[glacial pollen]

uhhh, then maybe try " python "

instead of " py "

I am now starting to wonder, how the f you even run rvc

if you've no pip nor proper py configured

[brittle wing]

Does low storage on ur ssd with the OS installed impact rvc's performance or nah

I have like 10 gigs of storage left

[glacial pollen]

it does

if you're close to literal topping of your ssd

the write / read speeds will get hindered

[brittle wing]

I filled this 1 tb ssd with too much crap

[glacial pollen]

most basic config really, works really well

256 ssd for os, 1tb ssd for games or daily use programs, 2 tb ( hdd ) of storage for media / archive

[brittle wing]

Is that why sometimes my rvc voice sometimes stutters out of nowhere

And I have to restart rvc to fix the problem

[glacial pollen]

Stutters?

[brittle wing]

Ya

[glacial pollen]

the inference output should not stutter and is not related at all to your drive, whatsoever

if you mean the voice changer* and not actual rvc ( cause a lot of people are confused with naming

then that's hard to say

as I haven't tested it in such an env

[brittle wing]

The voice changer

[brittle pelican]

I think this

[glacial pollen]

then it's rtvc

not rvc really

[brittle pelican]

I haven't open it yet

[rare goblet]

Ayo? @brittle pelican level 2 !!!

[glacial pollen]

either way, I can get the sample length to even as low as 0.10 or so

tho stability in a long run is another thing

can safely assure it can handle at least 0.13 - 0.17 just fine

tho, if you're fine with 1660, Ig? stick to it

but you won't regret 3060 at all ( pick 12 gig one.

[analog obsidian]

Sorry for interrupting this hehe
But does running realtime in fp32 makes any quality difference? Assuming we are using a fp32 model as welk

[glacial pollen]

No because it doesn't work like that

the fp16, fp32, bsfloat 16, tf32 etc etc, it is all precision modes for training, the models resulting from the training work the same everywhere, regardless

[brittle wing]

What if I were to use the voice changer for like 1 hour in constant talking

[glacial pollen]

should work? not sure man

[brittle wing]

Will there be any stability drops or nah

[glacial pollen]

I haven't used it on daily basis, just for fun few times for few mins max

but I assume, as long you pick value without under or overruns ( stutters )

you'll be fine

[brittle wing]

I use rvc pretty seldomly

[glacial pollen]

Well, your situation is quite weird and messed up.
if I were you I'd actually start everything from scratch. Properly updated and installed
but this is a lengthy process and I won't guide you through everything 100%, just give you more or less sumup of things you gotta do

[analog obsidian]

Ah ok, i asked that because modded wokada lets you choose between fp16 and fp32 mode
So is just there for 166x and other non fp16 capable gpus i guess

[brittle wing]

Sometimes I cant get rvc to run with my overclocked gpu

It just shows an error saying "ILLEGAL MEMORY ACCESS"

On rvc

[glacial pollen]

Then that's about the w-okada itself

fp32 will be much more intensive and def not worth it, w-okada wise

unless you got some beast setup, I wouldn't advise it

[analog obsidian]

Thanks for telling me this i was a bit confused

[glacial pollen]

Because you should not use overclocked hardware for machine learning

in which absolute stability matters the most

if you want performance in ML, get a new gpu

[brittle wing]

Damn

[glacial pollen]

no tinkering will help you sadly

a slight undervolt is fine tho ( my case )

[brittle wing]

Guess I'll wait for the 5090 to be released

[glacial pollen]

why bother

[brittle wing]

I wonder which sample length that gpu can withstand

[glacial pollen]

to get 5060 or 5070 with awful performance:price ratio?

you'll be better off with 4090 unless 5060's confirmed to be actually good

[brittle wing]

Money isnt really an issue for me rn

[glacial pollen]

In that case ye, wait for 5090

¯_(ツ)_/¯