✨│ai-help

[odd shale]

For example, on a colab like Applio's one

https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb

[granite bobcat]

thanks! i'll try it out later

[brittle dragon]

Pls help me

Running with the system Python.
Traceback (most recent call last):
File "C:\Users\senes\OneDrive\Masaüstü\rvc\rvcgui.py", line 10, in <module>
import torch
File "C:\Users\senes\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch_init_.py", line 148, in <module>
raise err
OSError: [WinError 126] Belirtilen modül bulunamadı. Error loading "C:\Users\senes\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\lib\fbgemm.dll" or one of its dependencies.
Press any key to continue . . .

[odd shale]

You're welcome buddy.

[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.

[brittle dragon]

I do not upload photos

[brittle dragon]

I'm trying to start rvc gui but it gives this error

[odd shale]

Are you trying to use models or train models?

If you're just wanting to use models, use Applio Colab instead.

[brittle dragon]

Can I write to you privately? I'm translating because I'm Turkish, I don't fully understand.

[odd shale]

But i'm already trying to help you here.

[brittle dragon]

Yes I have a model and I just want to make music

[odd shale]

Use Applio Colab instead.

https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb

[low shard]

if you are using https://github.com/Tiger14n/RVC-GUI thats outdated

dont watch youtube videos

whats your PC GPU?

[pastel oak]

#1159595680659157132 pinned message'lara bak, orda video tutorial vardir. Orda sorulari sorabilirsin, Enes o isleri bilir

[brittle dragon]

Yes I was trying to set this up

[low shard]

yea don't use that

i guess you are just gonna ask in the turkish channel

[brittle dragon]

thank you very much, good luck bro 🙂

Traceback (most recent call last):
File "C:\Users\senes\OneDrive\Masaüstü\Applio-3.2.2\app.py", line 1, in <module>
import gradio as gr
ModuleNotFoundError: No module named 'gradio'

Press any key to continue . . .

[odd shale]

Thanks for helping him Shadyyy my dear

[pastel oak]

🐢 🥬

[magic badge]

colab

:v

[unborn zodiac]

The voice I'm training is getting more vorbley as it cycles near 500
Is this like a mid way thing before it clears up or did I feed it bad food?

[odd shale]

🐢 ❤️

[odd shale]

Bad food indeed or it's overtraining.

[unborn zodiac]

I think it's over training
It starts to fall off after 96 cycles

What's the most sample Audio I should have before diminishing returns starts?

[rare goblet]

Ayo? @unborn zodiac level 1 !!!

[carmine hill]

whenever my training fails on applio for whatever reason, i start from epoch 1 again even though it has epoch checkpoints that are much higher

[stiff rover]

Am I supposed to link to the track it's suppoed to train off of?

[slender blaze]

Do you guys know which Colab notebook let's you merge models?

[thin stump]

https://colab.research.google.com/drive/1mHKTGH5e3SAyDSBss1KtiYRbDdQzwSMs

to save you time, I don't think you can merge 2 different models. Just the pth file

[slender blaze]

Thank you! Yeah I wanna merge 2 pth. I already do it in Replay on the inference stage but I wanna make standalone merged model

[rare goblet]

Ayo? @slender blaze level 2 !!!

[thin stump]

Great, you just upload the pth file paths you want in whatever location in the colab folder. I personally upload into assets > weights to find them

[stiff rover]

How do i make a dataset?

[thin stump]

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

[slender blaze]

Awesome I'm gonna try that. Thanks again for the help

[thin stump]

g/total first then your other one

[slender blaze]

This is exactly what I'm looking for

[stiff rover]

How do i make a PTH file?

I thought I did somethin on the WebUI of RVC, but all it did was made the voice clearer (which it did need)

[rare goblet]

Ayo? @stiff rover level 2 !!!

[stiff rover]

I am so damn confused

[finite prawn]

Does Batch Size and Index Algorithm impact the quality of a model's inferences, or is it simply just there for faster training?

[simple iris]

what are good settings for chunks (i need it to sound good)? i dont mind performance cause i have a pretty high end gpu

[knotty moth]

what do u mean "Index Algorithm"?

[finite prawn]

when creating a model in applio, in the final steps, there is an option called Index Algorithm, and has three options, 'Auto', 'Faiss' and 'KMeans'. I don't know what the difference is and I can't find any past messages or documentation about what it is

[knotty moth]

haven't used the latest applio, but "Faiss" is the normal one and "(minibatch) KMeans" is the one that would be executed if the dataset length is 1 hour/more, that happens in the mainline RVC and also the "Auto" option

[finite prawn]

thx!

[knotty moth]

that means if you run the normal "Faiss", the index file size would be around 500+ MB for 1h+ dataset, and "KMeans" would drastically reduce the file size to 30-50+ MB, and also possibly reduced indexing accuracy

[knotty moth]

the above guide explains all of that already

[stiff rover]

How long shuld I make the recording? Should I do the full 50 min limit or is 10 fine

[celest terrace]

yo where can we make the voice we downloaded say stuff

[rare goblet]

Ayo? @celest terrace level 1 !!!

[celest terrace]

or use it over our voice

im lost

[viral hazel]

is there anywhere i can get voice models for xtts ?

[brave garnet]

hamed kissed rvc voice model 💖

Made with love by Ilaria 💖

[finite prawn]

Does Batch Size impact the quality of a model's inferences, or is it simply just there for faster training?

[rare goblet]

Ayo? @finite prawn level 3 !!!

[finite prawn]

does batch size 2 result in best accuracy lmao???? i dont care about speed or file size - i just want the most realistic outputs i can get

[analog obsidian]

batch size is how the ai is learn your dataset, not a quality setting
lower is more accurate but you have risk of unstable training
to get a realistic model first you need a big dataset (30 minutes and above), with the person using their full voice range

[finite prawn]

"unstable training"?

and more accurate as in it'll produce more realistic inferences? what does 'accurate' mean in this context

[analog obsidian]

model will sound like shit, simple words

[finite prawn]

ahhh

kk thx

[analog obsidian]

batch size 8 is the most safe of all

[analog obsidian]

accurate in terms of the model sounding similar to the dataset, capturing every details of the audio

batch 8 already does that pretty good

[knotty moth]

try between 4, 8, 16

[autumn dust]

@proven hill any way to pay for unlimited gpu?

[buoyant furnace]

I just downloaded the "RVC-GUI" folder from GitHub. How on earth do I use the RVC-GUI.bat file?

[slender pewter]

Model Load Error! Check your model is valid. local variable 'embedder' referenced before assignment

what can be related to?

[pastel oak]

Is this on wokada? Are you able to tell me if this is on v1 or the new alpha version?

[tired orchid]

I got mangio running but what is Running on public url?
there is a public url apart from local url

[thick dagger]

how i can fix 504 Gateway Time-out ?

-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
• UVR5 NO UI for Google Colab, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[odd shale]

RVC GUI is outdated, use Applio Colab instead.

[odd shale]

Nope.

[odd shale]

Yep, that's normal i think

If you use Mangio locally, after executing the .bat file a window with the GUI will appear.

[low shard]

RVC GUI is outdated, what's your PC GPU? .bat files are a sequence of commands for windows, are you on mac?

[low shard]

what?

[tired orchid]

What is public url?

I don't want to share it

[zinc raft]

gradio server (rvc collab)

[low shard]

yea i just seen it myself

@zinc raft and @thick dagger check #📰│dev-updates message

[brittle wing]

hmm

[dusk sun]

anyone know how can i make the text to speech with higher pitch like singing?

[azure marsh]

📘 Kaggle Notebooks

• 🆕 Applio Notebook by Vidal
• 🆕 Applio Notebook by Shirou
• 🆕 Audio Separation by Shirou
• UVR5 NO UI by Eddy
• How to use RVC Mainline Kaggle by Cauthess
• ✨ RVC Mainline by Hina
• Original W-Okada's Voice Changer Kaggle
• Modified W-Okada's Voice Changer Kaggle

Note: Kaggle limits GPU usage to 30 hours per week.

More commands: -colabs, -hf, -local, -rvc, -help

[gloomy holly]

can anyone help i cant load https://drive.google.com/uc?export=download&id=1ESLX2FtNY4S0ZyKm-rJZ1af7yRqWArpT

[autumn beacon]

Which software will be best to use RVC models and generate TTS for short vids?

[violet heron]

Download then upload the files in whatever fork your using

[knotty moth]

you should provide a direct download link, i.e. huggingface link, otherwise reupload it

[gloomy holly]

i got it working with https://huggingface.co/spaces/Clebersla/RVC_V2_Huggingface_Version

nvm

[gloomy holly]

It did work

[rare goblet]

Ayo? @gloomy holly level 1 !!!

[tame mica]

use ilaria rvc instead lolz

oh ur using a Drive link nevermind

[low shard]

u can still do it on ilaria rvc zero

just upload the model files

[low shard]

btw i suggest u ilaria rvc zero

that one ur using is on cpu which is slow

while ilaria rvc zero runs on zerogpu, which is an A100 GPU

[brave garnet]

Ilaria RVC: CLICK HERE 🤗

Guide on how to use it: CLICK HERE 📝

Don't forget to thank Ilaria if you find it useful! 💖

[wind hearth]

"Train Feature Index" taking forever with RVC, this is normal?
(i got a single 13min wav file as the data, and its been like this for 25min now)

Output information:
(34778, 768),891
training

[rare goblet]

Ayo? @wind hearth level 1 !!!

[soft tiger]

can someone send me the thing where i can put models and use vocal audio to make the model sound like the audio

[rare goblet]

Arcane Commands

The prefix for AI HUB is /

Useful Resources

Command List
Dashboard
Support Server
Arcane Status
Premium

[soft tiger]

@odd shale

[low shard]

u mean make the audio sound like the model right?

or u mean make a model

[odd shale]

Make a model.

[odd shale]

Are you using local or colab?

[low shard]

whats your pc gpu?

[low shard]

oh okay

[soft tiger]

a rx 6900 xt

[low shard]

you cant train on that and rvc doesnt support it, you willl have to do it in cloud

[soft tiger]

ok

[low shard]

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
• Applio (ui)
• Mainline (UI)
• RVCDISCONNECTED (no ui)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
• Mainline (UI)

[soft tiger]

ive done it before

so its all good

[soft tiger]

yeah i got applio

so now i dont rememeber how to add my models to software

[low shard]

are you trying to make one or using the model you just trained on applio?

[soft tiger]

use a model i have

[wind hearth]

kaggle || RVC mainline

[low shard]

you can read this https://docs.aihub.wtf/rvc/cloud/applio-colab/#inference-

Also for inference (using model), you can also use Ilaria RVC Zero, which is an HuggingFace ZeroGPU (dont duplicate the space), that has a better gpu than colab for free, which means faster

[flint solar]

can u hel

p

[low shard]

you are using the gpu right

[low shard]

with what

[brittle wing]

is 2 gb vram ok for rvc training?

[low shard]

no

[flint solar]

im tryna resume my training but its starts from scratch

any idea why

[low shard]

where are u training

[flint solar]

first time ts happens

[flint solar]

rvc disconnected

[wind hearth]

yeah, the other previous steps ran fine, and i enabled the GPU from kaggle

[low shard]

did it disconnect out of no where?

[wind hearth]

process data, feature extr went fine

[low shard]

kinda weird ngl

[flint solar]

no i stopped the training and saved the model files

[wind hearth]

yeah idk it was 40min like that

[brittle wing]

is it even possible to train it with that vram?

[wind hearth]

i restarted ui, did stuff again but its same issue again

[low shard]

i just said no

[low shard]

mmm weird, u are using the same google account of the last time u stopped training right? bc its connected to google drive

[wind hearth]

nothing about "index" in the logs either idk

[low shard]

40 mins dataset?

[wind hearth]

noo i mean it was on this message for 40min

[brittle wing]

even with ilaria's optimizations?

[flint solar]

no i actually saved it and shared it to another g drive and started training, before it starts it gives me the previous tensorboard with all the infos, but it just restarts out of nowhere

it usually works fine

[flint solar]

yes even wit ilarias optimizations

[alpine osprey]

applio

-applio

[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

[wind hearth]

idk still nothing

[low shard]

yes, you cant, just use cloud

also, all forks have the same quality

[thin stump]

If it was jammed, i would have deleted the notebook and start over

should take you 30 seconds to train an feature index. Any longer means something got stuck in the log files like train.log

[flint solar]

@low shard

[brittle wing]

how can you get more than 4 hours in colab?

[rare goblet]

Ayo? @brittle wing level 2 !!!

[brittle wing]

plus this amount isn't even set

it fluctuates amd you get disconnected randomly

unless you pay up

[thin stump]

Use kaggle instead. 12 hours daily

https://rentry.co/RVC-Mainline-Kaggle

[brittle wing]

what should be the optimal length of a dataset?

is 15 minutes more than enough?

[thin stump]

It's more quality > quantity so if you have 15 minutes of dirty audio that's what you get

[brittle wing]

i mean, don't always have the time to look for the creme de la creme vocals

ngl

[low shard]

you can:

• Use an alt google account
• Use kaggle which gives more gpu time but its harder
• Wait till tomorrow
• Pay for colab pro

[brittle wing]

what's tmr?

[low shard]

short for 'tomorrow'

[low shard]

u should use the google account in the google colab which has the model saved

[flint solar]

no gpu for now

[low shard]

you can:

• U can maybe transfer the model files into an alt google account and use that
• Use kaggle which gives more gpu time but its harder (but ig ud start by 0 but atleast wouldnt worry)
• Wait until tomorrow
• Pay for colab pro

[flint solar]

how much gpu time does kaggle offer

[brittle wing]

is the ngrok monthly limit enough to train a model with 15 minutes of data and 200-250 epochs at most?

[low shard]

30 hours weekly

[flint solar]

u have a guide for it, cause i never used it

[brittle wing]

https://rentry.co/RVC-Mainline-Kaggle

[flint solar]

thx

[thin stump]

oh yup it's train.log. I'm not sure if your screenshot cut off that chinese mumbo jumbo. Updated in the guide

[low shard]

oh my bad he replied faster lol

[wind hearth]

i didnt cut off the screenshot, that was all i had yeah

i saw the guide had more lines and the "added..."
line at the bottom once it was finished so i figure smth is wrong

I didnt get to those lines :/

[flint solar]

how much storage do they give u

[wind hearth]

i had only these first two ones yh, even after waiting for 40min

[thin stump]

I had that problem once so I remember deleting the notebook and lowkey seething lmao

[tired orchid]

No such file or directory: 'rmvpe.pt' in mangio

[rare goblet]

Ayo? @tired orchid level 1 !!!

[thin stump]

it was train.log

[wind hearth]

ah oke

[tired orchid]

rmvpe is missing?

[rare goblet]

Ayo? @wind hearth level 2 !!!

[wind hearth]

welp i tried restarting n stuff but no success, i ended up uploading the index file that i generated previously from the RVC Disconnected colab haha

i tried colab first but quit since it kept insisting i have a limit reached x)

we'll see how it goes ig

[low shard]

20gb

[tired orchid]

Is kaggle free?

[thin stump]

Read the guide. It's too good to be true? Maybe so but it's difficult to use for first-timers

[wind hearth]

-overtrain

[azure marsh]

Overtraining

You can detect if a model is overtraining if the TensorBoard graph starts to rise and never comes back down. An overtrained model will sound robotic, muffled, and won't be able to articulate words well.

Check these resources to learn more about this topic

• Epochs & TensorBoard from AI HUB Docs
• TensorBoard from 🍏 Applio Docs

[astral jungle]

Which UVR5 model should I use for vocal separation?

[rare goblet]

Ayo? @astral jungle level 19 !!!

[wind hearth]

says MDX23C in docs ig

[astral jungle]

give me docs link

[wind hearth]

https://docs.aihub.wtf/rvc/resources/vocal-isolation/#best-models

[astral jungle]

thanks

[wind hearth]

-kaggle

[azure marsh]

📘 Kaggle Notebooks

• 🆕 Applio Notebook by Vidal
• 🆕 Applio Notebook by Shirou
• 🆕 Audio Separation by Shirou
• UVR5 NO UI by Eddy
• How to use RVC Mainline Kaggle by Cauthess
• ✨ RVC Mainline by Hina
• Original W-Okada's Voice Changer Kaggle
• Modified W-Okada's Voice Changer Kaggle

Note: Kaggle limits GPU usage to 30 hours per week.

More commands: -colabs, -hf, -local, -rvc, -help

[finite prawn]

I'm getting a single epoch every hour. Is this normal for my training settings running on an RTX 4070 w/ 8GB VRAM?

Dataset is 1 hour and 12 minutes long.
Sampling Rate: 48000
RVC Version: V2
CPU Cores: 20
Pitch extraction algorithm: rmvpe
Embedder Model: contentvec
Batch Size: 8

Save Only Latest: Off
Synch Graph: On
Save Every Weights: On
Cache Dataset in GPU: On
Pretrained: On
Pitch Guidance: On
No custom pretrain

Index Algorithm: Faiss

[analog obsidian]

disable cache dataset in gpu, that only helps if you're training a small dataset, below 10 minutes
save only latest should be on, you don't need all g and d files
make sure your audio is true 48k, all youtube audio is 32k for example, you know this by using spek or any spectrum analyzer
use 2~4 cpu cores, even if you have more than that, rvc has a bug that affects perfomance if you use a lot of cpu cores during preprocessing

applio uses rmvpe cpu so reducing cpu cores will also decrease rmvpe pitch extraction speed process by a ton

mainline has rmvpe gpu so reducing the cores will not impact the pitch extraction speed too much (pitch extraction is still going to be a bit slower if u decrease the cpu cores)

[tame mica]

the docs hasnt been updated in months

[finite prawn]

ohh shit thanks so much!!! i was so confused. lyery with the save once again

[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
• UVR5 NO UI for Google Colab, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[knotty moth]

especially when using crepe and that much cpu cores (seems like an intel i5 6P+8E), it might freeze a while and crash the application, no matter if your cpu is high end

[finite prawn]

20 was my default value

like the one applio gave me

[mystic tangle]

wahtsthe github for w okada modded

[stiff rover]

Should i get Appolio with or without pickles

[tropic plover]

Is it normal? It doesn't look right to me...

[analog obsidian]

those are mode collapses

that happens when rvc tricks itself that is learning a lot when in reality is only reading a small portion of your dataset over and over again
this happens for multiple reasons:
excessive amount of silence in the dataset
dataset lacks variety so only learns to do one thing
speech/singing is not natural, e.g. words split in half resulting in unnatural speech/singing
very low batch size + very small dataset

collapsed epochs sound very bad so you have to choose a non-collapsed epoch

some models sound fine despite collapsing a lot, some not, is hit or miss

but ideally u want it to not collapse

[tropic plover]

Thanks, that's life-saving🥲 Also it seems neither converge nor show signs of overtraining, do I still need to wait for more steps or smth is wrong?

[analog obsidian]

something is wrong, if you have collapses it means the training is unstable so probably your model didn't converged or it did converge but very bad

[tropic plover]

Guess have to remake dataset then💀 Maybe making it 'sounds' less repetitive, and prob choose a different pretrained

[analog obsidian]

use the original one, use audio labels and try to remove some repetitive sounds, that should help

most of the time collapses happens* simply because the dataset has too much silences, which is fixed by using audio labels

#1203830360878743612 message

[finite prawn]

Which Indexing Algorithm produces a more realistic model, Faiss or KMeans?

[analog obsidian]

none, what produces a realistic model is the dataset itself
faiss is what is used for index creation, kmeans is an algorithm of faiss which increases the training speed of the index file

[compact iron]

How do I make a model? I have about 11 mins of data

[knotty moth]

as explained before #✨│ai-help message

[analog obsidian]

also want to add that mainline rvc already uses a similar algorithm named mini batch kmeans for datasets that are over 1 hour long, its enabled automatically but i believe is possible to disable this as well in the configs

oh damn u already said it

lmao i just read the message

[flint solar]

why cant i train more than 1000 epochs on kaggle bro

ts suck

[compact iron]

What’s kaggle

Can you use it to make models?

[knotty moth]

most models barely overtrain in over 1000 epochs, but you can try this modified UI code to replace the one in the imjoy's root folder

[knotty moth]

https://rentry.co/RVC-Mainline-Kaggle

[flint solar]

do i just execute this separately?

[compact iron]

Thanks

[knotty moth]

replace the existing runmain.py there and then re-execute the run gui cell

[flint solar]

i cant find runmain.py

[knotty moth]

open the imjoy file manager (the File url provided in the kaggle cell output), and it is in the home directory

[flint solar]

alr i found it

thx

[dawn kelp]

i noticed that while im using it, it spikes my cpu usage but doesnt really use much gpu even tho i downloaded the one specifically for gpu usage, is this normal?

[knotty moth]

especially when you use extra more than 16k and/or index rate above 0, but anyway try this optimized fork installation https://rentry.co/forkvoicechangerguide

[dawn kelp]

the index is still at 0 for me and i assume the 16k is referring to the chunk thing?

[knotty moth]

16k extra as I said

but in the fork one, it is 2.7s extra as optimal setting

so why not go try it?

[dawn kelp]

for the fork should i uninstall the first one and install it in the same place?

[rare goblet]

Ayo? @dawn kelp level 1 !!!

[knotty moth]

you can just extract to another folder, no system's appdata and registry used at all

[dawn kelp]

im gonna give it a shot right now

the optimized one *

[azure marsh]

Documentation

AI HUB Docs

https://docs.aihub.wtf/

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• Full AI Voice Model Training Guide (local) by Christopher Villanueva
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[brittle wing]

is there any way to avoid RVC v2 Disconnected disconnecting after around 5 hours? separate from the 2-3 hours inactivity thing. It seems to just end on its own for no reason

[patent pasture]

No, there isn't because you have about 3-4 hours of using Colab's GPU per day

[brittle wing]

damn so how can I do over 300 epoch with a large dataset?

[patent pasture]

maybe use kaggle (there you have 12 hours per session)

but there are many nuances when you use kaggle so I don't recommend

[patent pasture]

or you may train your model for 300 epochs on one acc and then continue training on other acc

idk

[knotty moth]

it needs phone number verification for some needed feature

[knotty moth]

depending on the length and batch size you could predict on

[brittle wing]

okay ty

[rare goblet]

Ayo? @brittle wing level 2 !!!

[unborn crest]

I'm not sure if this is the best place to ask, but if I'm using a duplicate of Ilaria RVC in huggingface and I want to change the format of the output audio, which file should I edit? I've been just converting them manually after downloading but it's kind of a pain

[ember lily]

I don't understand why I use ilaria rvc on google colab, everything is fine, as soon as I convert the file, after about 40 seconds it says error, before it didn't do it. Can someone help me? Thank you

[unborn lotus]

I have a question, When I turn on the voice changer and all that my virtual mic/cable picks up desktop audio. Anyway tp fix that?

[proper shale]

AFAIK the default output for RVC is .wav - ig you'd have to change the code and add some ffmpeg stuff for it to convert to other file formats

[proper shale]

I'd suggest using other alternatives atp, like Ilaria RVC on Huggingface

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

[low shard]

Are you using Ilaria rvc mainline btw?

This one I mean #✨│ai-help message

Bc if it's broken Idk if Ilaria knows that

[proper shale]

Probably mic bleed, enable sup2 and increase s thresh. If that didn't solve it, maybe It's your input/output setup...

[unborn lotus]

Did it, doesnt seem to be working. I changed my virtual mic to another one called VAC and it seems like the virtual mic is the problem. Should I install another version of rvc?

[knotty moth]

you should use decent headphones instead of speakers, besides the above suggestions, turn down the system volume if necessary

[knotty moth]

what kind of the mic problem?

[proper shale]

Nah not needed if the different cable fixed it

[unborn lotus]

As of now I changed the VB Cable to VAC and it's still the same. The issue being whenever I'm in a call or anything similar it picks up their voices and other desktop audio and relays it back to my microphone and in turn they can hear their own voice coming from my mic. Like the voice changer picks up their audio and stuff

Now I know it's not the virual microphones fault but I think the voice changer picks it up

[knotty moth]

turn down the system volume and/or adjust gain?

[unborn lotus]

Alr I'll try

Should I go for 10 in gain?

Unfortunately it's still happening. Idk what seems to be the problem

[knotty moth]

input gain? it'd worsen lol

[unborn lotus]

Oh I misread it

But the resutl was still the same man

[finite pond]

!rvc

[dim solar]

is there any way to pause the model training? i need to turn off the computer for the night but dont want to lose 2 hours of training

[pastel oak]

If you are training locally with your GPU and set a save frequency, you will automatically save the G and D paths required to continue. To double check this before ending your training, go into logs > modelname > check if you have D_233... and G_233... files

If you are on colab, what colab are u using

[dim solar]

i dont think im using colab but not sure

colab is online right? im training locally

the files are there though

[pastel oak]

Then you can save at the next frequency (if you set it to 10, its every 10 epochs, if its 25 every 25 etc.)

You can either use the stop training button, but if its bugged then just close the whole console window

[dim solar]

yea i closed it since i had no option to stop it

[rare goblet]

Ayo? @dim solar level 1 !!!

[indigo torrent]

Could someone help me set this program correctly so that it doesn't sound like plastic?

[odd shale]

Do you mean W-Okada?

[indigo torrent]

RVC GUI

[rare goblet]

Ayo? @indigo torrent level 1 !!!

[odd shale]

Yeah.. RVC-GUI is pretty outdated, don't use it.

Use Applio Colab instead.

[indigo torrent]

Where can I download?

[indigo torrent]

?

[odd shale]

https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb

Nope, you don't have to download it.

[indigo torrent]

Ok

[indigo torrent]

Can you send a tutorial for this?

[wind hearth]

unlucky

bruh kaggle reset my files i dont got a logs folder anymore

guess that means i cant resume training?

i got the last .pth and .index downloaded tho

[astral jungle]

hello

[odd shale]

There you go buddy.

https://docs.aihub.wtf/rvc/cloud/applio-colab/

[latent kettle]

-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
• UVR5 NO UI for Google Colab, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[low shard]

i mean, if he has a good pc gpu he can also do it locally

or if he wants to only inference theres ilaria rvc zero

idk what @indigo torrent wants to do though

[odd shale]

Of course.

[brisk quarry]

FROG

[thin stump]

Oh, ig headless mode wasn't a bad suggestion after all . I made it seem like it's way too OP for the average use case

[wind hearth]

(un)fortunately idk what u mean

[wind hearth]

https://tenor.com/bLuSE.gif

[rare goblet]

Ayo? @wind hearth level 3 !!!

[wind hearth]

sad no gif privilege

[wind hearth]

oh damn i missed that

welp now i know

now i know the cause atleast thats good

[thin stump]

I didn't really like using it that often since It kept running off the background if I didn't stop it. I do one whole session anyways

[low shard]

@tall dew i will ping u here as u used the wrong channel

You can search rvc ai voice model at:

• #1175430844685484042
• #🔍│find-models
• https://weights.gg/
• https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice models)
• https://applio.org/models
• https://voice-models.com/

if there isnt one, you can:

• #1159289738314919936
• #1191429836321849435
• make it yourself with our docs guides https://docs.aihub.wtf/essentials/how-to-make-voice-models/

also whats your pc gpu?

[wind hearth]

aah

i did another session all from scratch, which i migh have to stop again but ill try downloading everything just in case
n then the headless mode ig

[thin stump]

everything should be saved in the /logs/ folder if you used the same notebook regardless (don't delete it at the end, just leave the page)

when the interactive session stops running

[wind hearth]

imo i did that the last time, but still ended up here

aah but i probably reinstalled after I saw it didnt let me run the UI

[thin stump]

oh ok, there we go

[wind hearth]

welp

the pleasures of first time x)

how do i listen to samples of the default tts models

[steel fog]

i have problem, everything i say in programm is ignoring, and all i hear in output is moaning

[brisk quarry]

[indigo torrent]

Why I have there error?

[fathom reef]

So, I just finished training my model. Which one is which?

Screw the D, my ngrok expired ridiculously quickly after that.

[thin stump]

What colab is this? It's based on data transfer out so if you download these huge ass G & D files you're capping

-kaggle

[azure marsh]

📘 Kaggle Notebooks

• 🆕 Applio Notebook by Vidal
• 🆕 Applio Notebook by Shirou
• 🆕 Audio Separation by Shirou
• UVR5 NO UI by Eddy
• How to use RVC Mainline Kaggle by Cauthess
• ✨ RVC Mainline by Hina
• Original W-Okada's Voice Changer Kaggle
• Modified W-Okada's Voice Changer Kaggle

Note: Kaggle limits GPU usage to 30 hours per week.

More commands: -colabs, -hf, -local, -rvc, -help

[fathom reef]

So, if I don't have a new phone number, how do I even get access to these?

[thin stump]

then you don't. You can try making a new ngrok without the phone number. But I think it's just kaggle that needs it

[fathom reef]

bruhhhhh

Screw this.

Because my GPU is not supported.

An RX 580 will not do the job well at all.

[thin stump]

Yeah, sorry about that. I was hoping someone else will write a guide on applio but they have to add images 👀 @brittle wing

[fathom reef]

Wait, hold on, hold on...

Damn.

That's mine.

[thin stump]

Great, train local

instead

[fathom reef]

So it is possible, it's just ridiculously slow.

[brittle wing]

How do i tell you that @brittle wing did the kaggle applio guide 4 days ago?

[fathom reef]

Like, taking several hours just for one model.

[brittle wing]

https://applio.org/guides/58

If we ain't talking about that, what we talking about then?

[thin stump]

hmm, ig it's good enough. Just saying, MGLR downloaded the D and G files instead of the pth files so guides have to be user friendly

[brittle wing]

He forgot to tell about exporting .pth and .index

But it's good tho

[thin stump]

it capped out his ngrok usage. Entirely his fault 🙃

[brittle wing]

tho ngrok links are the same, right?

bc of this

[thin stump]

yup, the imjoy gui was implemented for the applio kaggle recently

since it was going to be discontinued when blaise left so that got saved

[brittle wing]

you can export the pth and index models directly from the webui (model finished training > click on export model arrow > click on refresh > select pth and index > download ur files)

no need to use imjoy elfinder unless you want to upload the dataset there or download g and d files

[broken prairie]

I mean, I've also made fusion drum models

[brittle wing]

omg

I also have that same guide in Spanish xd https://docs.google.com/document/d/1MBqEYCYAVSEK09sX-DEId6-D0LIeRyDQZcyy9XpZIfc/edit?hl=es#heading=h.gfycrao9hi2p

[thin stump]

All that's missing is telling them how to package the files together (pth file and added_index) so you don't get people downloading unrelated files like they're off the leash

at the very end of the guide, it's good either way i think

[brittle wing]

I will add it

although of course, the locations are the same as in the applio notebook in colab xd

I wanted to upload it to rentry.co but the guide crashed after 5 minutes of existence xd

I am modifying the guide xd (I will also translate it into English)

[thin stump]

Yeah no worries, it's in the right location. Also don't be afraid to add tons of images if you could on there because people cannot read 🫣

I guess the difference between the mainline and applio kaggle is how much time is spent on the imjoy

[brittle wing]

I put also a part where you can replace imjoy the Finder by one modified by vidalnt

the modified one is supposed that when changing pages the folders do not disappear from the gui

also fixes some bugs when deleting files

[brittle wing]

the bad thing about the applio guides is that you can't resize the images.

[thin stump]

Oh I see. As long as you can add them in

[brittle wing]

I told deivih (deivih is in charge of applio.org) about being able to edit guides and change image sizes but he is a bit lazy and lost his pc xd

[thin stump]

Way better then if you can copy and paste it on google docs. Now this part makes some sense lol

[brittle wing]

listen https://docs.google.com/document/d/1LXaZpb446wK9gTeMiTKwbB8xQVrPMPovK7aNiE6oh-0/edit?usp=sharing

Spanish version: https://docs.google.com/document/d/1MBqEYCYAVSEK09sX-DEId6-D0LIeRyDQZcyy9XpZIfc/edit?usp=sharing

the section on knowing where the pth and index are located is short because it is very simple.

the guide on applio.org was to satisfy deivih to use the guide section xd

[fleet cedar]

whats the best male voice i cant find it on weight or voice models channel?

i hate my own voice

it sounds ass

[olive bridge]

Can I have an admins help in dms

[subtle cedar]

-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
• UVR5 NO UI for Google Colab, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[subtle cedar]

no more easyGUI? It's sad because only there I have a chance to train models because that RVC Disconnected is hard to and doesn't work to me ☹️😔😭

Pls bring back easyGUI and it'll still be free?

what happened to easyGUI tho and why it doesn't exist anymore?

is there's alternative for both easyGUI and RVC v2 Disconnected? because we can't download the zip of the models that we trained on Astra Labs bot due to restriction of Discord and the free training on Weights needs Streaks and that Streaks is so limited and we need to do AI Covers,Images or Chat there daily/everyday to gain more Streaks ☹️ can anyone pls help or answer me?

[thin stump]

It's ok. My only feedback would be not to cover creating a pre-train model as i'm sure this is something that has to be discussed with others (a bit advanced for the average person)

• Since it's separate from the applio guide, you can also condense down the pretrain list by linking this https://docs.google.com/document/d/1j9J8A8Oop9bMOHmCs3jDXzPujuD6TQ0Q396rJ0MyuIc/edit

• also reuse images from the applio guide. I don't think anyone will mind. Because I kinda got lost trying to read the text https://docs.aihub.wtf/rvc/local/applio/#training-

[final sentinel]

スマホでrvcモデルを作る方法ってありますか？

[zinc kernel]

have you managed to resolve the laugh issue ?

[pastel oak]

Laughing is noise, as of now theres no workaround for laughing

in relation to the whole dataset, a few seconds of laughing could help a little bit just like the whole topic on including breathing sounds, but it should be very small parts of it and it will not magically make laughing accurate

[zinc kernel]

i trained a model already and it sounds perfect. should i train it again over 2 .wav files of laughter ?

[rare goblet]

Ayo? @zinc kernel level 1 !!!

[brisk quarry]

how do i install a plugin

Theme Loaded: Applio
Traceback (most recent call last):
File "C:\Users\Lukag\Desktop\ApplioV3.2.4\app.py", line 89, in <module>
plugins_tab()
File "C:\Users\Lukag\Desktop\ApplioV3.2.4\tabs\plugins\plugins.py", line 31, in plugins_tab
plugin_import = importlib.import_module(plugin_main)
File "C:\Users\Lukag\Desktop\ApplioV3.2.4\env\lib\importlib_init_.py", line 127, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "<frozen importlib._bootstrap>", line 1030, in _gcd_import
File "<frozen importlib._bootstrap>", line 1007, in _find_and_load
File "<frozen importlib._bootstrap>", line 972, in _find_and_load_unlocked
File "<frozen importlib._bootstrap>", line 228, in _call_with_frames_removed
File "<frozen importlib._bootstrap>", line 1030, in _gcd_import
File "<frozen importlib._bootstrap>", line 1007, in _find_and_load
File "<frozen importlib._bootstrap>", line 972, in _find_and_load_unlocked
File "<frozen importlib._bootstrap>", line 228, in _call_with_frames_removed
File "<frozen importlib._bootstrap>", line 1030, in _gcd_import
File "<frozen importlib._bootstrap>", line 1007, in _find_and_load
File "<frozen importlib._bootstrap>", line 972, in _find_and_load_unlocked
File "<frozen importlib._bootstrap>", line 228, in _call_with_frames_removed
File "<frozen importlib._bootstrap>", line 1030, in _gcd_import
File "<frozen importlib._bootstrap>", line 1007, in _find_and_load
File "<frozen importlib._bootstrap>", line 984, in _find_and_load_unlocked
ModuleNotFoundError: No module named 'tabs.plugins.installed.BS-RoFormer-0'

[pastel oak]

Its worth a try, lmk how it goes:)

[zinc kernel]

ill do it rn, i have probably 3 laughters of the model. ill train it and let u know

[fossil pasture]

How do I figure out which Batch size is right for model training?

[knotty moth]

RVC disconnectedは簡単な方法だと思います
https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS

[fossil pasture]

also how much does the choice of Batch size affect the result?

[brittle wing]

Ok

[brisk quarry]

1 is most accurate

and slowest

[brittle wing]

about the images, I told a mod of ai hispano (server where they take care of applio) and he told me that there is no problem xd

[tidal isle]

https://colab.research.google.com/drive/1r4IRL0UA7JEoZ0ZK8PKfMyTIBHKpyhcw#scrollTo=0o9rMOIk1EEF

THIS IS NOT WORKING PLEASE HELP

[brittle wing]

which rvc is it?

[tidal isle]

I WANT TO CONVERT VOICES INTO AI SO ITS THAT WEBSITE

[little ore]

what is best way to make a ai cover rn?

[brittle wing]

use applio https://colab.research.google.com/github/iahispano/applio/blob/master/assets/Applio.ipynb

[tidal isle]

IS THERE A AVAILABLE RVC WHERE WE CAN CONVERT VOICES ONLINE

[brittle wing]

weights.gg

[tidal isle]

HOW TO USE IT

[little ore]

what to make a ai cover that is a website and not google colabs?

[brittle wing]

weights.gg

use that

[tidal isle]

WHAT IS WEIGHTS.GG

[brittle wing]

this https://www.weights.gg/es

[little ore]

a website

[tidal isle]

IS IT FREE

[little ore]

is it all free

[brittle wing]

you need to pay to make models but it is unnecessary since you are going to make covers.

[little ore]

if i want to make my own model what do i use to make?

[rare goblet]

Ayo? @little ore level 1 !!!

[brittle wing]

you have an nvidia gpu?

[little ore]

yeah but its a 4070ti i dont think its good for ai

[brittle wing]

I recommend 8gb of vram for train models

[little ore]

is there no online websites that can train?

[brittle wing]

colabs and kaggle notebooks

[little ore]

wait wouldnt colab disconnect ?

[brittle wing]

yes

[little ore]

then theres basiclly no websites to train models free?

[brittle wing]

I don't know of any good ones

[little ore]

i have one more question like there people that make ai covers that changed the lyrics but it sings like the original

how do they make that

[brittle wing]

making good models

[little ore]

no imean

[gray crow]

realtime

[little ore]

it sings like the original song

[gray crow]

thats how

[little ore]

wdym

the person itself sing it?

[gray crow]

Depends. If you mean, like song covers with changed lyrics, than they recorded them theirselfs

[little ore]

but theres no way they sound like the original song

[gray crow]

Why?

[little ore]

example this one

https://www.youtube.com/watch?v=bLTQwgT64T8&pp=ygUNc2tpYmlkaSBzaWdtYQ%3D%3D

[gray crow]

i dont know the song, but maybe they did not change the "hook"

[little ore]

what is a hook

[brittle wing]

I think they put production on the audio they cover.

[gray crow]

Maybe they just changed the parts they wanted and kept the parts original which you refer as "sound like the original"

[brittle wing]

by fl studio

[little ore]

ohh

thats hard work

so currently the best way to make ai covers is weights?

[brittle wing]

I would say applio because I have the feeling that weights will be made payable

[gray crow]

i mean the quality depends on the model. But yeah, weights is easy, but use applio instead if you have gpu

[atomic gyro]

-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

[gray crow]

even works with cpu, but slow

[little ore]

what about the one made by SociallyIneptWeeb

[rare goblet]

Ayo? @little ore level 2 !!!

[brittle wing]

never heard of that xd

[atomic gyro]

he made a video 1 year ago

[tidal isle]

IS THERE ANY WEBSITE OR THE SAME WEBSITE LIKE GOOGLE COLAB VOICE CONVERT TO AI FOR FREE

[analog obsidian]

damn the amount of misinformation here
so we start with this:
best free option to train models rn is kaggle, can do a max of 16 batch sizes and is super fast compared to colab
4070ti is already strong to train locally
To change the lyrics of the song, the people that made those sung the lyrics and inference it on rvc, alternatively they used synth v
Weights is not the best option to inference

[brittle wing]

I gave him the easiest to use

[analog obsidian]

in fp16 mode u can literally train with 4gb of vram

[gray crow]

hes a starter, he surely knows what you mean by kagg,e batch size and whatever.

[brittle wing]

kaggle is difficult to use

[gray crow]

without given a notebook for rvc training or inference... yes?

with... no?

[analog obsidian]

I pinged him so he can know the actual facts because u guys were telling him to use colab and weights despite him having a strong gpu

[gray crow]

i didnt read about his gpu, tbh

[analog obsidian]

He literally said he has a 4070ti

[brittle wing]

I recommended that he train with his gpu.

[gray crow]

still without a notebook, the keyword kaggle wont be helpful ^^

[little ore]

what bout ai covers

[analog obsidian]

Before jumping into kaggle first he needs to understand the basics

[brittle wing]

use applio

-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

[gray crow]

yes yes and yes

[analog obsidian]

http://docs.aihub.wtf/essentials/how-to-make-ai-cover/

[little ore]

OK

[analog obsidian]

[little ore]

is applio run on local gpu?

[subtle cedar]

You can train models for free on Weights.gg but you need a 5 Streaks and you need to do AI Covers,Images or chat there everyday to gain Streaks

[analog obsidian]

Use mainline, which is the original unedited rvc, is faster (way faster) than applio and less buggy

[brittle wing]

if you have gpu, yes

[little ore]

what is mainline

[analog obsidian]

http://docs.aihub.wtf/rvc/local/mainline/

the original rvc

[little ore]

oh

[analog obsidian]

applio is a fork, edited version

[brittle wing]

applio was a fork of mangio

[little ore]

is there websites besides weight that u can use ur own model to make covers

[analog obsidian]

no

[little ore]

hugging face smthing like that

[tidal trail]

colab

[analog obsidian]

ilaria rvc

[brittle wing]

-hf

[azure marsh]

Suggestions for @little ore

<:huggingface:1179800228946268270> Hugginface Spaces

• UVR5 UI, by Eddy and Ilaria Huggingface Spaces
• Ilaria RVC Zero, by Ilaria Huggingface Spaces
• RVC⚡ZERO, by r3gm Huggingface Spaces
• Applio, by IA Hispano Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

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

[analog obsidian]

Use ilaria rvc zero

[little ore]

do i have to remove the song's music by myself?

or it does it automaticaly

[brittle wing]

use uvr colab

[analog obsidian]

U need to remove it yourself, mvsep currently has the best models
uvr locally is very outdated

[little ore]

ok

thanks

[brisk quarry]

Applio is buggy???

[brisk quarry]

Faster by how much

[analog obsidian]

Used to be, not sure now tbh, but ive read is better in that regard currently
About ui speed mainline is faster because is less bloated
Training speed? Ilaria told me once that raven tested it and found that applio was a few seconds slower than mainline during training

[brisk quarry]

Yea it is a bit bugfy

But not too bad

Also wdym seconds

I meant how much faster by it/s

[analog obsidian]

Epochs take 6 seconds more in applio, or at least that was ive been told

Tbh i should try to confirm that

[brisk quarry]

But uhh

What even is the it/s

[analog obsidian]

I don’t know but if im telling u that epochs take longer, then the it/s are slower as well

[brisk quarry]

What are your it/s usually when you train

@analog obsidian

[analog obsidian]

I use kaggle for training because my gpu can’t do batch 8 in fp32 mode lol

[brisk quarry]

Use fp16 then.

[analog obsidian]

No

[brisk quarry]

Why

I run batch size 1

@analog obsidian WHY

[analog obsidian]

I like my gradients to be more stable, and avoiding NaN values as well

[brisk quarry]

I dont notice a difference

[analog obsidian]

They sound almost/exactly the same

is just training stuff that changes a bit

[brisk quarry]

Its slower

Thats all

@analog obsidian is there any benefit of fp32

[analog obsidian]

Avoiding NaN values at the start of training

[brisk quarry]

And what is that

[analog obsidian]

Gradient explosion

Means something went wrong while calculating gradients

[brisk quarry]

And what is that

[brisk quarry]

Means??

[analog obsidian]

It literally says in the pic what means

unstable network

unstable training

fp32 in rvc helps u getting more stable training

not a quality setting

[brisk quarry]

And. What is the difference in result

[static gyro]

cant rechie this logs

help

[analog obsidian]

In my test my model got better breathing sounds
besides that, quality wise the results are almost identical to fp16, with the fp32 model sounding a bit better

because what handles quality is the dataset anyways

fp32 is not mandatory or super good, it just a precision setting for precise guys like me

[quick jungle]

https://cdn.discordapp.com/attachments/1160576561636966491/1165372018414342284/warning.gif

[analog obsidian]

[steel stump]

is it possible to train a model off a single like 20 second clip?

[waxen owl]

@analog obsidian Hey, sorry to ping, just can't dm ya. I just saw a message from you saying that for realtime, microphone quality matters. Would you have any suggestions? Would rather know I'm getting a good mic for the software before spending money ^^

[rare goblet]

Ayo? @waxen owl level 1 !!!

[brisk quarry]

give it to me and ill try

[analog obsidian]

anything that is not a laptop mic should work, even a headphone mic
make sure your mouth is in a good distance from your mic, not too close, not too far, the volume should be also fine, not very loud, mic volume too loud can cause us to hear our models sounding* more robotic because we are increasing the volume of the artifacts

[red kayak]

u dont know the difference in the first place uwu

[analog obsidian]

for software i recommend steelseries noise supression, at least in my mic it does a decent job

[waxen owl]

gotcha, then I think I should be good with the one I have, thanks! I'll check out the software

[brisk quarry]

precision

speed

accuracy(kinda)

[red kayak]

yeah you know how gradients work though?

[brisk quarry]

no

also

is running batch site 16 fine

[red kayak]

yeah but usually do it on larger datasets

[brisk quarry]

would 8 be better?

or 16

[red kayak]

8 is usually better

[valid mural]

-Rvc

[azure marsh]

Documentation

AI HUB Docs

https://docs.aihub.wtf/

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• Full AI Voice Model Training Guide (local) by Christopher Villanueva
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[valid mural]

-Help

[azure marsh]

Command has expired

[vital grotto]

Is there a Implementation of RVC that has TTS to Model Voice Conversion?

[odd shale]

Applio i think.

[low shard]

RVC is natively for Speech To Speech, but 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 with RVC Voice Models (if you got a good pc):

• You can find Ilaria RVC Mainline here (no guide as of right now)
• While Applio in our docs

If you don't got a good pc you can do tts with RVC Voice Models on cloud:

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

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

• if you don't wanna use edge tts, you could try another tts ai from our tts index and use the output as an input in rvc

[vital grotto]

Tahnks!

[rare goblet]

Ayo? @vital grotto level 2 !!!

[low shard]

your welcome

[cosmic epoch]

where do i share hugginface models?

[low shard]

https://docs.aihub.wtf/extra/model-maker-role/

[low shard]

@river tiger lets talk here, whats your PC GPU?

[wheat sentinel]

-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
• UVR5 NO UI for Google Colab, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[tranquil raven]

where is the main link for rvc

The one thats good with amd gpus

[olive bridge]

So I have the vc okada how do I use the rvc for discord

[olive bridge]

how do i fix delay with rvc

i say something

then it says it like 2 seconds later

[zinc kernel]

it went bad. however, that model had only 5 minutes of audio samples because google collab. i downloaded a trainer software, Mangio, and i did 17 minutes 500 epoch. in this training i had about 18 seconds of laughter for the model, and it is still bad but noticeably better for laughing.

[subtle cedar]

Hopefully that easyGUI will be back because only easyGUI the colab that i have a chance to train because it's the easiest and because that RVC v2 Disconnected is so hard and doesn't work for me

[glad zealot]

you can try using mainline colab

it also has a gui

[subtle cedar]

Yeah but someone said only when I have a PC ya'll give me a guide from Kaggle for that

[glad zealot]

you can use kaggle or colab versions, they are the same thing, pretty much the only difference is how to upload a dataset

and need a phone number on kaggle to actually use it

[subtle cedar]

I can use the colab version of mainline as a mobile user?

[glad zealot]

yes

[subtle cedar]

And pls do the screenshot of the tutorial at least on DM to understand more how to use or train models there thanks

[glad zealot]

same guide as kaggle mainline

https://rentry.co/rvc-mainline-colab

this too

[knotty moth]

sounds like a skill issue, RVC disconnected is easier to me, but the mainline kaggle is more customizable and has more gpu session quota

[subtle cedar]

Okay I respected your opinion

[subtle cedar]

But I think I'm gonna give up to train models on Google Colabs because Weights and Astra Labs are the easiest and best alternatives to me although they're so limited but waiting for unlimited Streaks (I mean because on Weights you need to use 5 Streaks for free training) or any future new free training options/features on Weights and also Astra Labs' future website because probably we can finally download the model or zip from there since we can't download models or zips on the bot version of Astra Labs due to restriction of Discord and because most of my favorite artists are popular or semi-popular so many people already do their models since most of people are only focusing on popular artists/musicians

[obsidian beacon]

when i open the google drive link it doesnt work

how do i o this😭

[surreal bear]

how do i make it so my voice isnt heard but only the changed voice (to others) 😭

[low shard]

the easygui google colab link got even deleted, it wont be back and the creator doesnt respond or be active since months

[low shard]

what link?

do what

[fleet cedar]

pls guys help im literally getting bullied by people for my stupid accent

[patent pasture]

unluck

[knotty moth]

https://tenor.com/view/who-cares-al-bundy-married-with-children-no-one-cares-idc-gif-23009028

[subtle cedar]

Yep ☹️

[low shard]

yea, wont be ever back

[subtle cedar]

It's okay because I won't train on Google Colabs anymore because the pth file is too large to download

[low shard]

ah rip

thought u were gonna use the mainline rvc colab

[quaint dock]

hey guys, is there a website that writes down what is on the picture?

like deepbooru or how is it called

[subtle cedar]

Because people don't give me tutorial how to use that like screenshot or screenrecord tutorial because only screenshot or screenrecord tutorial helped me to understand how to use things

[low shard]

yea its because video tuts are hard to make and to maintain as there could be changes

[brisk quarry]

[low shard]

there are no updated yt tuts

[subtle cedar]

Okay thanks

[low shard]

maybe https://docs.aihub.wtf/rvc/cloud/applio-colab/ could help as there are screenshots

[subtle cedar]

But that's for Applio not RVC mainline and I already tried that colab months ago but it doesn't work to me and because probably that's only works on PC

[low shard]

Applio is an RVC Fork (basically modified version, mainline is the original one), btw yea the Applio Colab works as i tried it

[subtle cedar]

Oh

[brisk quarry]

im bored

ai voice aint fun anymore

[low shard]

learn LLMs

[brisk quarry]

what

[low shard]

ye

run Llama 3.1 405b

[brisk quarry]

low latency models?

[low shard]

Large Language Models

[brisk quarry]

eh

nah

[low shard]

text2img

Stable Diffusion

flux

[brisk quarry]

all boring

[low shard]

train a LoRA

[brisk quarry]

no

thats cursed

[low shard]

hows image generation boring

[low shard]

how

[brisk quarry]

its never really accurate

[low shard]

u need to be good at prompts and use the right models and settings

[brisk quarry]

and there we go

why tf would i want that

[low shard]

bc its funny 🔥

[brisk quarry]

time to make a useless discord bot

ok nvm

[low shard]

make a chatbot using gemini api

[brisk quarry]

idfk what that is

wasnt gemini a upscale app for phone

[low shard]

google gemini

[low shard]

what

https://gemini.google.com/app

[brisk quarry]

life is not even worth it

[knotty moth]

this could be the fastest in running Llama 3.1 405b
https://www.techspot.com/news/104482-cerebras-challenges-nvidia-new-ai-inference-approach.html

[brisk quarry]

idk what that is

[low shard]

its cerebras

do u like ai?

[brisk quarry]

the whast

everything is so useless

i live in suffer

[simple hatch]

Umm guys i want to ask

How to look queue or progress my download

[low shard]

what

[simple hatch]

my bad

[obtuse sierra]

can someone help me? i dont know what channel i should go for this but my vb audio thing is mixed up because it has the cable input in where the output devices are

and the cable output is in the input devices section

i can provide a image if you dm because i have no image perms here

[brittle dragon]

How can I model my own voice?

[low shard]

whats your PC GPU?

[brittle dragon]

4

[wraith coral]

download link pls for nvidia

[low shard]

thats the CPU

not GPU

[low shard]

of what? and what gpu do u have more specifically

[brittle dragon]

7.8 gpu

GPU 0

Intel(R) UHD Graphics

Kullanım    2%
Adanmış GPU belleği    
Paylaşılan GPU belleği    1,0/7,8 GB
GPU Belleği    1,0/7,8 GB

[knotty moth]

you should opt for one of these cloud solutions:

• RVC disconnected colab: https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS
• Kaggle: https://rentry.co/RVC-Mainline-Kaggle

[low shard]

yea thsts integrated graphics, u cant do it locally

As you dont got a good PC, its better you use cloud for training an RVC Voice Model:

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
• Applio (ui)
• Mainline (UI)
• RVCDISCONNECTED (no ui)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
• Mainline (UI)

[wraith coral]

rtx 4050

of rvc

[low shard]

alr

check https://docs.aihub.wtf/ -> RVC -> Local -> Download either mainline (original rvc) or Applio (rvc fork with some extra features, same quality) on your pc

[wraith coral]

Which is best

Applio or Mainline

[low shard]

they have the same quality

the only difference is some extra feature in Applio

but there isnt a 'best' one

[brisk quarry]

which one is the best one

[brittle dragon]

Is there a video explanation link? I couldn't find the video on the Turkish Animation channel.

[subtle cedar]

But I'm a mobile user and don't have a laptop yet so I don't think I will use Applio again and probably it's still doesn't work to me

[brittle wing]

-hf

[azure marsh]

<:huggingface:1179800228946268270> Hugginface Spaces

• UVR5 UI, by Eddy and Ilaria Huggingface Spaces
• Ilaria RVC Zero, by Ilaria Huggingface Spaces
• RVC⚡ZERO, by r3gm Huggingface Spaces
• Applio, by IA Hispano Huggingface Spaces
• AICoverGen, by r3gm Huggingface Spaces
• Advanced RVC Inference, by r3gm Huggingface Spaces
• RVC v2 Huggingface version, by Clebersla Huggingface Spaces

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

[low shard]

it works on Mobile, its on cloud, it runs on a remote google pc

[brittle dragon]

pls help me

[low shard]

may be harder because of the interface but its possible

[low shard]

no, there is no video

follow what i told u

[low shard]

.

[subtle cedar]

Okay

-colabs

[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
• UVR5 NO UI for Google Colab, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer Google Colab

ℹ️ Note

While the Colab free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

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

[subtle cedar]

But I mean for train models and idk if it's works for train models on mobile since I already tried that colab for infer models and when I infer models there it's always gets error and doesn't work

[subtle cedar]

And the training there seems so hard and I don't think I can train models there 😭

[low shard]

it works, just is harder

[low shard]

yea its hard but doable

what error did u get btw?

it shows the error in the google colab page

[subtle cedar]

No, it's on Gradio

[low shard]

yea it shows the red Error in the gradio ui, but it shows the error output in the google colab page

[subtle cedar]

Oh okay

[subtle cedar]

Sorry to say this but I think I'm actually should give up to train models on Google Colabs and because easyGUI which is the easiest colab is gone ☹️😢😭 but also because the MB of pth files are too large to download this is why I only trained models on Weights and Astra Labs nowadays because not only those are the easiest but also because you don't need to download pth files there and it's just zip

[knotty moth]

you can run rvc disconnected

[subtle cedar]

No sorry because that colab is so hard and it's doesn't work to me so stop recommending me that colab

[knotty moth]

sounds like a skill issue

[subtle cedar]

Yeah maybe haha

[subtle cedar]

I'm gonna give up to train models on Google Colabs right now and I'm tired when people always recommended or forced me to use RVC v2 Disconnected because that's doesn't work to me and it's hard to train there (I'm not to be rude or attacking anyone it's just that colab isn't work and hard for me)

I will just train on Weights and Astra Labs right now because those are the easiest to me and no need to download pth files although they're so limited because of that Streaks thingy and Astra Labs won't allow to download the model or zip because we know that Discord is so strict

[low shard]

the zip of astra labs and weights is just the pth and index compressed lol

[subtle cedar]

Yeah

[low shard]

btw weights.gg training is paid

[low shard]

unless U:

• Make a creation for 5 days streak to get a free premium training, you can repeat this
• Use referal code, getting 5 friends to subscribe via your training

[subtle cedar]

I know but you can still train models there for free but nowadays you should use 5 Streaks for free training

[low shard]

if its really too hard for u its ur choice then

[subtle cedar]

Yeah thank you for agreeing with me and respected my decision hehe

[simple hatch]

Can someone give me some recommend for a voice changer

[rare goblet]

Ayo? @simple hatch level 1 !!!

[odd shale]

A model?

[simple hatch]

Executor

[plucky vortex]

Hey everyone sry to disturb you i have some question about RVC i never used that software i wanted to know :

• Is it totally free ?
• Its working on any app or game ?
• the quality of the generate voice are good ?

Thx in advance for your answer

[simple hatch]

Well we kinda have same question

[subtle cedar]

But I think Google Colabs are easier to PC/desktop and it's just harder for mobile users because some of them especially the Gradio one gets error on mobile

[simple hatch]

I just want play dnd tho

[knotty moth]

use the non-webui one like rvc disconnected

[subtle cedar]

Okay but maybe it's hard like RVC Disconnected so no I don't think I will use that

[subtle cedar]

waiting to buy my own laptop so maybe I can use any Google Colabs for training models but I don't have a lot of money for that since laptops are too expensive and I won't buy soon for some other reason

[pastel oak]

Links can be found on the guide

Original wokada v1 (this is the og that most people use):
https://rentry.co/VoiceChangerGuide

Fork wokada (frequent updates, runs on webui instead of client app. It runs better and has better performance than v1):
https://rentry.co/ForkVoiceChangerGuide

Theres an alpha v2, but its not recommended yet

[simple hatch]

I already try but potato laptop 🥲🥲🥲

[pastel oak]

Then use fork

[simple hatch]

Ok lets give it try

[pastel oak]

I assume you mean the voice changer, which is not RVC btw. We recommend w-okada

• its free
• works on any app and game, assuming you are on a pc/laptop
• Quality depends on the voice model. A lot of models dont sound great in realtime, some do. you have to trial error

[ripe void]

any reason why my AI conversion has a lot of "voice cracks"?

[plucky vortex]

Thx for your answer

[rare goblet]

Ayo? @ripe void level 1 !!!

[red kayak]

bad quality acapella used

[ripe void]

ah ig that makes sense

[red kayak]

yeah clean it up good and dry to get it as dry as possible

[ripe void]

i filtered it from an instrumental so the best i can get isnt that great xd

[modern gorge]

hey im using VC client and i dont really know how to put models i downloaded

[modern gorge]

can someone help me w this?

[modern gorge]

thats old

for the 1.5.1 version

[pastel oak]

Same principle still

[pastel oak]

waaaaait

[modern gorge]

there are more buttons tho

[pastel oak]

u said

1.5.1

Thats 2 years old XD

[modern gorge]

yeah

i have the v.1.5.3.1

[pastel oak]

Thats 2 years old

[modern gorge]

there no one more recent

(i think)

[pastel oak]

1.5.3.18 is most recent

[modern gorge]

huh

[pastel oak]

how did u download 3.1

lol

[modern gorge]

found it in github

[rare goblet]

Ayo? @modern gorge level 1 !!!

[modern gorge]

wait so where can i find that version

[pastel oak]

https://huggingface.co/wok000/vcclient000/tree/main

directML is for AMD
onnxgpu-cuda is for NVIDIA

[modern gorge]

oh

thankss

[red kayak]

use bs roformer on mvsep

[subtle gazelle]

where I can download RVC

[unborn zodiac]

For training a Voice
How do I know my GPU is being used correctly?
It's sitting at 5% when it's training

[violet heron]

What’s your GPU and what method are you using to train

[wraith coral]

Virtual Audio Cable suggestion for RVC?

[pastel oak]

VAC Lite
https://rentry.co/voicechangerguide#virtual-audio-cables-mandatory

[wraith coral]

is this for real time

[brave garnet]

Ilaria RVC: CLICK HERE 🤗

Guide on how to use it: CLICK HERE 📝

Don't forget to thank Ilaria if you find it useful! 💖

[pastel oak]

Yes

[wraith coral]

kk thx

[low shard]

its harder on mobile for the interface but u shouldnt get an error unless u do smt wrong

[low shard]

whats ur pc gpu

[subtle gazelle]

NVIDIA GeForce RTX4070

oo GPU

NVIDIA GeForce RTX4070

[low shard]

check https://docs.aihub.wtf/ -> RVC -> Local -> Download either mainline (original rvc) or Applio (rvc fork with some extra features, same quality) on your pc

[subtle gazelle]

ok thx

[rare goblet]

Ayo? @subtle gazelle level 1 !!!

[low shard]

yw

[wraith coral]

Index search FAILED or disabled
whats this

[jagged plume]

Which tune do you use when you are using a girl voice as a girl?

If I set the tune too high it becomes Mickey

[wraith coral]

also how to fix sizzle plus the voice comes lower at start and becomes higher

[odd shale]

You must test with different tunes.

[odd shale]

The appropiate tune for your model will depend on your own voice.

[old pumice]

why does my shit sound so ai generated and yall sound better than mines

[wraith coral]

How can i make the RVC use GPU

[low shard]

whats your pc gpu?

[wraith coral]

RTX 4050 Laptop

[low shard]

what rvc are u using?

[wraith coral]

Mainline

[rare goblet]

Ayo? @wraith coral level 6 !!!

[low shard]

also i heard laptop gpus are worse than desktop ones

[wraith coral]

idk about that

[low shard]

it should use gpu automatically weird

[wraith coral]

My cpu is high but not gpu tho

[low shard]

are u inferencing or training?

[wraith coral]

its at 48 *C rn

%3 usage

im using real time

go realtime gui bat

[low shard]

oh sorry idk

usually people use wokada

[wraith coral]

thats better for realtime?

[low shard]

maybe @pastel oak could help you better about this

[wraith coral]

downloading to test rn

[pastel oak]

Well u did comment on the other help channel and i told you what to do to fix it

But for 4050 laptop i would rather recommend the fork wokada, itll give u better performance

https://huggingface.co/Shadicti/deiteris-Fork/blob/main/voice-changer-windows-nvidia-b2309.zip

[rare goblet]

Ayo? @pastel oak level 64 !!!

[pastel oak]

it runs on webui, not client app just so u know

for any other questions or settings look here:
https://rentry.co/Forkvoicechangerguide

[low shard]

is there any good sides about rvc realtime instead of wokada or its fork btw

[violet mica]

can anyone help how do i re use ai voice

[pastel oak]

Ehh not anymore

Noise gating is broken, noise suppressions are ok, anything friendly user-wise is bottom of the pack
Its only worth it if you use third party program for noise gating and noise suppression, but even then, fork wokada does that for you and has the same, if not better performance

[low shard]

yea as i thought

[pastel oak]

imo its better than wokada v1 because better quality and less delay, i can deal with the broken noise gating personally

but fork beats it

[low shard]

i dunno much about wokada but so rvc realtime beats the wokada in https://rentry.co/VoiceChangerGuide ?

[low shard]

for what?

[violet mica]

well before i used a website to do my ai voice but this website got shutdown and i cant find a new app or website to put my song with characters voice anymore

[pastel oak]

U have to differentiate:
the prebuilt packages of wokada v1 in the guide you linked are probably the worst because of the bugs and broken extra and higher delay etc.

But if you manually set up the wokada v1 environment, its just as good as fork and realtime

Its just that fork prebuilt is already as good as manual environment of v1

And the problem is, people dont know how to set up environment so 98% people use prebuilt

what makes fork slightly better than manual environment v1 is also that extra is gpu accelerated instead of cpu, else its for the most part identical in computing. only differneces are the nice new features emojikage added that make things easier

[low shard]

alright you mean ai covers, guessing you dont have a good pc u should use ilaria rvc zero

[brave garnet]

Ilaria RVC: CLICK HERE 🤗

Guide on how to use it: CLICK HERE 📝

Don't forget to thank Ilaria if you find it useful! 💖

[violet mica]

actually i do have a good pc

its just

ill try it out

oh wait its look like the old one i used to use

damn thanks you

[low shard]

whats your PC GPU?

you could do it locally so u dont have to worry about cloud limits

[low shard]

oh damn

is it even worth to update the normal wokada guide

raven made it i guess

[pastel oak]

I talked to raven briefly, they updated and added the installation method for manual env wokada

at least for nvidia

[subtle cedar]

Okay

[low shard]

alr

[velvet dew]

-rt

[azure marsh]

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

[rare goblet]

Ayo? @velvet dew level 1 !!!

[old oracle]

can a 1660 ti run real time my friends trying and its not doing anything 😭

[unborn zodiac]

GTX 1070 mobile
Using RMVPE

[violet heron]

What fork of RVC?

[unborn zodiac]

Mangio

I feel like something broke in a weird way
I ran it for 256 cycles and nothing changes though it
32 sounds the same as 256

[violet heron]

Use mainline instead

Also maybe check if Cuda works properly

[unborn zodiac]

git link plz

[cosmic basin]

How i install a w-okada

With github.com

[violet heron]

-rt

[azure marsh]

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

[violet heron]

-local

[azure marsh]

Suggestions for @unborn zodiac

🖥️ 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

[shut turret]

when I add any new model, in rvc, I get this error: Please select a model and input audio fail. Audio and model are selected but they don't work.

[unborn zodiac]

You mean original?

[rare goblet]

Ayo? @unborn zodiac level 2 !!!

[blazing tangle]

Why isn’t my voice changing software working?

[violet heron]

Yes

[violet heron]

!t howtoask

Bot not botting

Uhhh

What are you using and what’s your GPU?

[blazing tangle]

The program starts and I keep seeing something like this.

I can't send a screenshot, but when I run the program I get a white screen

and no options

[rare goblet]

Ayo? @blazing tangle level 1 !!!

[blazing tangle]

rtx 4090

[violet heron]

Make sure the file isn’t stored in OneDrive and and your opening start_http

[meager basin]

nasıl açılır

[blazing tangle]

Ok już działa, dzięki mordo 🙂

you are a chad

[tranquil raven]

Wait so how do you download rvc?

[blazing tangle]

github

[tranquil raven]

Do you have the link

[blazing tangle]

yes

[tranquil raven]

Rvc is better than w-okada with amd gpus right?

Because I have a 6900xt

also do I need pytorch installed?

[blazing tangle]

this one is the best in my opinion because I don't sound like a robot

[tranquil raven]

Is the one your referring to the Original RVC?

Or a side branch

[blazing tangle]

Orginal

[tranquil raven]

oh okay thanks

[blazing tangle]

no problem bro 🙂

[low shard]

yea it should, be sure u are using the wokada fork

-rt