#✨│ai-help

1 messages · Page 212 of 1

brittle wing
#

oh, sorry

serene musk
#

-rvc

karmic oliveBOT
hallow wasp
#

-rvc

karmic oliveBOT
warm oriole
#

Is there any local alternative for suno ai ?

low shard
stoic viper
#

Hey everyone,
l have enabled the overtraining threshold in Applio so that training stops automatically if no improvement is detected.
In this case, do I still need to use TensorBoard to monitor the training, or is the threshold enough on its own to prevent overtraining?

crude flame
#

tensorboard is much better

#

even more so if you have the avg graphs

stoic viper
brittle wing
#

i don't think so, g/total isn't rising up

#

keep training until g/total starts rising up and never goes down

fair ivy
#

Help I can't seem to download okada the online way

low shard
#

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-reocrded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

edgy tangle
brittle wing
#

and when the overtrain actually starts

#

(you know, g/total rising up forever)

edgy tangle
#

5 minutes per epoch

simple ore
#

training for more may not bring any improvement and can actually make things worse

edgy tangle
#

I think so, my dataset is ≈1h

brittle wing
#

what i want to suggest the user is train 1000 epochs, but stop when g/total actually overtrains

simple ore
#

g/total may not even go up

#

but the model still go to shit and lose all the knowledge from pretrain

brittle wing
#

yea on long datasets overtraining is rare to see

simple ore
#

5min audio, 1700e - lost all ability to sing

brittle wing
brittle wing
#

1700 epochs is crazy for a 5 minute dataset

simple ore
#

but it did not become overtrained = robot voice

#

and total g did not go up

brittle wing
#

maybe the model will sound shit at some point

simple ore
#

no, it was fine for speaking

brittle wing
#

oh nice then, maybe the dataset is just speaking

#

that's why it may be bad at singing

simple ore
#

it could sing at 500e, not could not at 1700

#

the trained model retains pretrain features, overtraining in most common sense is training a model so much it forgets the previous training

brittle wing
simple ore
#

more epoch = more of learning from the dataset, more of losing previous knowledge

brittle wing
#

and that's why models lose ability to sing, right?

simple ore
#

pretty much. the ability to generate higher harmonics was pushed out by some other realignment

brittle wing
#

interesting about the realignment thing in the harmonics

#

wdym by realignment anyways?

simple ore
#

I dont know specific parts of the model that change this way, there are many millions of parameters responsible for the waveform generation after all

#

inference uses speaker latents and noise to generate a predicted spectrogram

#

so with overtraining it fails to make one with higher harmonics

hallow thistle
vale oak
#

Why is it changed in Weights to get songs on YouTube??? It is way easier and idk how to get songs in audio files 💔

hallow thistle
vale oak
#

Is it gonna come back?

hallow thistle
#

I don't know, but better look out for more information in Weights' Discord server. Although you won't be able to submit any YouTube link there on Weights, you can still do this with an already AI converted track that used an audio from YouTube to convert.

Screenshot_20250219_205603_Chrome.png
#

These two tracks were made after the removal of YouTube link feature on Weights, but can still see the YouTube icon marked on both.

edgy tangle
#

I'll create another dataset then

hallow thistle
fast scarab
#

Hey everyone,
I think my model is overtraining because my loss/g/total is increasing sharply after 46k steps.
Should I stop training now, or is there something I can do to fix this?
Also, if someone could help me understand this better, I would really appreciate it because I'm a bit lost.
Here’s my TensorBoard screenshot for reference.

Capture_decran_2025-02-19_152027.png
simple ore
#

increase sharply is not the increase from 34 to 34.5

fast scarab
simple ore
#

this is not scalars tab

#

you're showing Time Series

fast scarab
simple ore
#

how hard is to click 'SCALARS' ???

hallow thistle
#

That doesn't look like what the Scalars tab should be. The button is supposed to be orange, not gray.

fast scarab
hallow thistle
simple ore
#

loss?

#

anyway, gradients look like crap, fm going up from the start, not good at all

urban wasp
#

-collab

#

-colab

karmic oliveBOT
# urban wasp -colab
📒 Google Colab Notebooks
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, -realtime, -rvc, /help
fast scarab
simple ore
#

how big is the dataset?

fast scarab
simple ore
#

batch size?

fast scarab
crude flame
simple ore
#

even to 4

#

FM has to go down

fast scarab
crude flame
analog obsidian
#

its like a slow rising

fast scarab
#

I'll try that, thanks for your help!

edgy tangle
#

I just optimized my dataset and the time per epoch went from 4¿ minutes to 2 minutes

craggy brook
#

Why is there this error?

ho.png
knotty moth
karmic oliveBOT
knotty moth
#

you should use one of these newer ones

#

-rvc

karmic oliveBOT
craggy brook
knotty moth
#

do you think rvc could have loras?

tame mica
gloomy lynx
#

guid

low shard
#

Realtime voice changer for calls?

#

this is the wrong channel then

#

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-reocrded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

#

so, you just want to use it on pre-recorded audios right?

#

then let's use this channel, what's your pc gpu?

#

that's too weak, it could run locally on CPU if you got enough ram and good enough cpu, but it would be extremely slow anyways so not worth to run locally

#

Train (make) RVC Models on cloud:

  1. Prepare the Dataset
  2. Setup RVC:
    Choose a cloud way to use RVC,
  • Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
  • Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
  • Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
  1. Be sure to know about the tensorboard

Google Colab = Easier but risk of getting disconnected
Kaggle = Harder but way more gpu time
If you are looking for the easiest way and for free, is using https://weights.gg which ofc uses RVC

RVC Inference (use models) on pre-recorded audio on Cloud

You can use either:

#

Via cloud, you will run it on a remote good pc

molten fog
#

why is only the eval model in my tensorboard on lightning ai?

simple ore
#

eval is the folder for logs

low shard
#

You can:

  • Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
    • Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
    • Mainline: The original RVC
  • Cloud (remote good pc, easier and faster than ur PC but it's limited):
    • Ilaria RVC Zero: fastest and simplest that you can get for free
    • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
    • Applio Colab: max 4 hours daily, not granted, of GPU

Easiest possible (automatically separates vocals & instrumentals) : weights.gg
easiest cloud: Ilaria rvc zero
easiest local: Applio

warm glacier
#

thanks for the information

low shard
quaint tusk
#

Hey, everyone, can I ask for help on an error here or is it on the making-models channel?

#

It happened when trying to reload a model in a new session, in colab applio

simple ore
#

@jovial pollen if you need something, ask here

jovial pollen
#

For refinegan

#

Since the other ones doesn’t work due to being old

#

Of course if u don’t have anything im sorry for disturbing then

simple ore
jovial pollen
jovial pollen
#

overall it sound amazing!

#

on the low and the high the voice doesn't break anymore

simple ore
#

it is a model made with a very small dataset, 5:30s

jovial pollen
#

(maybe too much to ask 🙏)

simple ore
#

no

jovial pollen
edgy tangle
#

Is this overtrained?

gleaming chasm
#

Is there any way wo make the AI voice sound more crisp by doing some adjustements to your mic?

analog obsidian
analog obsidian
# edgy tangle

graphs aren't very accurate in showing overtraining, you just have to hear the epochs

#

its possible for a model start to overtrain even when the g/total graph is still going down
the model forgets most of what it learned from the pretrain if you train it for too long
overtrained models are pretty obvious, the model sounds robotic/disorted and it struggles to inference any audio, tho every model overtrains differently, some overtrained models still are able to do some stuff despite forgetting things

crude flame
analog obsidian
#

i did it with the jeff model
after e64 the model forgot how to sing

simple ore
#

well, mainly dont run 1000+ epochs

edgy tangle
mossy crest
#

Hello, I want a voice similar to children chorus. Is there any?

brittle wing
#

Hey all, been outta the loop since the initial ai boom and need to brush up my knowledge.
Is the RVC client by w-okada still the best tool for the job, or have people moved to something else by now?
Thanks.

#

Ah, I see there's guides sections in here, I'll poke around a bit.

knotty moth
agile storm
#

guys, is there any working rvc training colab?

low shard
low shard
neat glacier
fast scarab
#

Hello,
I set the batch size to 4 as recommended, but now the graphs seem to be stagnating. Here are the charts:
I’m not sure if this is normal or if it indicates a problem. Could someone tell me if this is okay or if I need to adjust something else?
Thanks for your help!

knotty moth
fast scarab
neat glacier
fast scarab
neat glacier
#

what does your dataset look like? is it nice and clean, and how long is it?

fast scarab
neat glacier
#

some basic questions

  • what's the learning rate
  • what pretrain (cuz that's a thing now)
#

that is some INSANE levels of mode collapse

low shard
craggy brook
fast scarab
simple ore
#

i mean.. 2 steps per epoch?

#

you f'd up preprocess/extract features

knotty moth
#

there's no way it's 0.1 steps per epoch, perhaps it means it reached 12k epochs

knotty moth
neat glacier
knotty moth
neat glacier
#

i remember the original Ilaria RVC had that as an option

#

also how come we can't change activation functions?

#

rvc has not aged well

crude flame
neat glacier
#

what happened between interval 10 and 12?

#

looks like a collapse to me

knotty moth
neat glacier
#

that's interesting

#

gelu should theoreticall be better than relu?

knotty moth
neat glacier
#

no access

knotty moth
#

activate the ai testing role here

image.png
neat glacier
craggy brook
ember hollow
#

@low shard

#

When I try to put a mic

knotty moth
simple ore
low shard
craggy brook
remote karma
#

How do I download RVC V2?

low shard
remote karma
#

4060 ti and i just wanna troll with a female voice

low shard
remote karma
#

Yes

low shard
#

then this is the wrong program and channel

#

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-reocrded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

remote karma
#

Sounds good thanks

craggy brook
woeful cave
#

is applio inference not working?

craggy brook
coarse crest
#

does anyone know how to create your own RVC model?

fast scarab
#

Hey everyone,
I finally managed to get the training to work! However, I’m having trouble locating the checkpoint files in RVC. I want to use the best checkpoint I found, but I can’t seem to find where it’s saved.
Does anyone know the exact folder or path where RVC saves the checkpoint files? I checked a few places but no luck so far. Any help would be appreciated!

simple ore
#

for mainline RVC

chilly slate
#

#✨│ai-help #🔍│help-ai-art Hey I was trying to use voice model to change my ai assistant voice but couldn't do it bcz of some error is there anything who can help?

fast scarab
hallow thistle
#

!howtoask

patent trellisBOT
# hallow thistle !howtoask

How To Troubleshoot AIHC_WaitWhat

__**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.
opal hill
#

heya, what does crepe hop do for RVC f0 method crepe?

grand pecan
#

Hi guys quick question, is this overtraining / converging? From the point I think it's the lowest it has started going kind of flat. Ty
https://ibb.co/RT4S1HLY

simple ore
pastel oak
stone dawn
#

Hi! Is there an RVC that works for conversion voice only?

craggy brook
simple ore
tame mica
graceful patio
#

Can anyone help me on this

#

So im training a model that has

#

a dataset with about 10-15 mins

#

and its taking 40 steps per epoch

#

on a batch size of 8

#

is this a good sign or a bad sign?

#

using refinegan btw

#

using noobies base train 44k sample rate

jovial pollen
graceful patio
#

its training

graceful patio
jovial pollen
graceful patio
#

its an artist

#

that im training on rn

jovial pollen
graceful patio
#

i can ping you when results are ready

#

probably needs another hour or so

#

because of the steps its taking a while (to make progress)

#

well right now i am actually at 2k steps for every 50 epochs

#

let me see if i can get a sample right now

jovial pollen
graceful patio
# jovial pollen many thanks!

hmmm seems like it needs more training but im still gonna send you a sample (this is a rapper model / lil uzi vert) and you can hear some progress but i heard that it needs like way more epochs to get an actual really good result

jovial pollen
graceful patio
#

hm its not robotic you can tell

#

its just very like airy weird

jovial pollen
#

yeah weird

graceful patio
#

its only 50 epochs rn

jovial pollen
#

that explains haha

graceful patio
#

heard you need to reach 200-500 to hear any noticeable difference lmao

#

since this one is more advanced

jovial pollen
#

lets see around that then

graceful patio
jovial pollen
earnest knoll
#

might be a stupid question, but how do I have applio generate the index file at a specific epoch once i've found the lowest g/loss value on the tensorboard, or do I have to restart training and stop at that epoch to generate the index?

royal marsh
#

Hello anyone have problems with w-okada like gpu goes to 100% usage and freeze totally the game you playing?

opal hill
low shard
#

elaborate:

  • your pc gpu
  • what guide link did you follow
  • a screenshot of your wokada

in #🔍│help-w-okada

red kayak
# graceful patio

thats very likely ur dataset messing with the learning process. are you using studio recording or are you using uvr isoalated vocals

graceful patio
#

cleanest though

#

its a lil uzi vert model in the making rn

#

i can send you a sample

red kayak
graceful patio
#

of the dataset

red kayak
#

yeah

graceful patio
#

minor double vocal (but not included in all dataset)

#

i heard it just cleans it out its very minimal

simple ore
#

shit quality

graceful patio
simple ore
#

bad isolation

red kayak
#

well there are so many issues with this

graceful patio
#

thats what should matter right?

red kayak
#

you have backing vocals in there

#

thats bad

#

causes vocal doubling and confuses the model lots

#

2nd theres instrumental bleed

#

also very bad practice for training

graceful patio
# red kayak you see

and just to mention i dont have izotope 11 so i cant clean it as good as usuing ai isolation

#

nor dont want to crack it bc of virsues and stuff

#

so what would be best in ai isolation

#

to clean it

red kayak
#

a lot of frequencies between 3000 and 6000 were butcher, which may impair perforamance and training stability

image.png
jovial pollen
#

refinegan is also very sensitive to dataset

red kayak
red kayak
jovial pollen
graceful patio
red kayak
#

you even have left over reverb residues which is bad ofc

image.png
graceful patio
#

heard its better than the old fox joy model

red kayak
graceful patio
red kayak
#

and by manually silencing that part

graceful patio
#

i do hear oddly some clipping tho is there any way to fix those artifacts

#

i get that sometimes

#

like that mild clipping

red kayak
graceful patio
#

im hearing it in this audio you sent me / has some sort of clipping in it like bleed?

#

kind of like a distortion

#

to your sample

#

like very low

red kayak
#

no no, as you fore mentioned, thats distortion

#

clipping is different

graceful patio
red kayak
#

and this type of distortion is usually to the loudness of the vocals + the instrumental removal which damages the vocals

red kayak
graceful patio
# red kayak not really no

any way you can walk me through my cleaning process just quickly to see if there is anywhere wrong? i use mel reformer for vocal, and then dereverb with either fox joy or the mel reformer dereverb, de-echo if needed, denoise , and then if theres any backing vocals i use melband karakoe / then open audacity, apply a noise gate/trunacate silences and then normalize

#

what you seen (the sample i sent) was what i used and did

red kayak
#

u can just do mel roformer and de reverb

#

not fox joy though

crude flame
#

Anvuew mel dereverb v2 is the best btw

graceful patio
#

is that reccommended

#

if i want super clean vocals

red kayak
#

it'll help with annoying clicks, left over reverb residues, noise removal and more

crude flame
red kayak
#

hook my mans up

low shard
#

fuck yes the "funds" 🙏

crude flame
graceful patio
graceful patio
#

thank you for these funds

ruby vector
#

is it broken? because there's nowhere to import voiceclips

earnest knoll
ruby vector
#

oh

earnest knoll
graceful patio
#

do better?

#

then the normal model itself?

red kayak
#

not really

#

thats over kill

#

and wont really help

graceful patio
#

since its combined with the best models

red kayak
#

especially for creating models

#

mel band has a lower sdr than bs roformer but it isnt always better

#

always look at the spectrograms and compare

ruby vector
#

or maybe i'm just autistic and couldn't understand it

graceful patio
#

how does this sound

#

ill unsend

#

its just a sample of what i have

#

it does have minor echo residue

low shard
#

@simple ore btw does Applio work with Python 3.12.3? asking because I'm updating the termux guide for ubuntu24.04 which doesn't support 3.10 anymore, I'm git cloning and running the run-install.sh

#

seems like it works until it gets to numpy 1.23.5, so I feel like there's a dependency problem

ruby vector
#

do you have to use a vocal only version of a song for an ai cover in weights.com?

simple ore
#

pyenv

low shard
#

doesn't run-install already make a virtual environment btw?

low shard
ruby vector
#

oh i already found it out

simple ore
low shard
#

alright, thanks

simple ore
#

there are some differences, not sure what needs to change

low shard
#

gonna use pyenv

low shard
#

after that the installation didn't work so I dunno if other packages would need other changes too

simple ore
#

should work fine with numpy 1.26.4, <2.0

low shard
#

should

simple ore
#

I can try installing python 3.12

#

but...

brittle wing
#

when i try to use it on discord then test it i cant hear myself can someone help?

silver dirge
#

I created a song with my own voice a long time ago but I forgot it, it's been a long time, how can I do it now?

silver dirge
#

where can i train my model

vagrant bolt
#

Hello, i have trouble with the voice training, every time i try to use it there is an error : AttributeError: 'FigureCanvasAgg' object has no attribute 'tostring_rgb'
Can someone helps me with it ?

weak cipher
#

guys how to add model +tts

tough shuttle
#

how to trainnn

simple ore
low shard
#

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-reocrded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

#

elaborate:

  • ur PC GPU
  • the guide link u are using
  • a screenshot of ur discord and wokada settings

in #🔍│help-w-okada

low shard
silver dirge
#

ryzen 5 5500

low shard
low shard
#

That's a CPU

silver dirge
#

right

low shard
# weak cipher guys how to add model +tts

There are different Text To Speech (TTS) AIs:

GPT So Vits: RVC isn't as good as GPT So Vits for tts, but gpt so vits (few shot tts, which means needs just a lil training for models) can't use rvc models (and viceversa), and its only limited to: english, chinese & japanese, if you wanna check gpt so vits instead, read https://docs.ai-hub.wtf/tts/gpt-sovits/

Freemium 11labs: An easy way to do TTS is https://elevenlabs.io/, you can't use RVC model on this but its a mostly premium easy way for good quality TTS

FishSpeech: FishSpeech is a 0 shot (no explicit training needed) TTS, if you got a good pc you can use it locally else use their site

With RVC Models:

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

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

silver dirge
#

rx 6600

low shard
low shard
# silver dirge rx 6600

You can:

  • Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
  • Cloud (remote good pc, easier and faster than ur PC but it's limited):
    • Ilaria RVC Zero: fastest and simplest that you can get for free
    • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
    • Applio Colab: max 4 hours daily, not granted, of GPU

Easiest possible (automatically separates vocals & instrumentals) : weights.gg
easiest cloud: Ilaria rvc zero
easiest local: Applio

#

I would suggest you Applio with Zluda

silver dirge
#

thanks

low shard
simple ore
low shard
low shard
silver dirge
#

I forgot everything, I need to learn it again sometime. Is there a video you can recommend on the internet that would make it easier?

low shard
#

The only updated guides are the written ones, which I have sent you the link

simple ore
#

if the video is 6+ month old, it is likely outdated

silver dirge
#

sad

low shard
weak cipher
#

My computer is good enough but I want to do it on the web only.

hallow thistle
weak cipher
weak cipher
hallow thistle
weak cipher
#

need to find someone who can help me

#

when i finish i will build it through expo into apk

hallow thistle
#

Don't know about this one. I'm only here for basic RVC the audio changer and W-Okada the realtime audio changer steps.

weak cipher
#

or github

#

cvt to cdn

hallow thistle
#

Um.

#

Firebase is a cloud-based database API, that's what I know.

weak cipher
#

what if i put the full song on github will it exceed the limit?

low shard
#

Do you really need the tts model to run locally?

#

you could maybe try adding edge tts api, It works on python so should too on js

weak cipher
low shard
#

yeah I don’t use js

hallow thistle
low shard
weak cipher
weak cipher
hallow thistle
low shard
weak cipher
low shard
#

it won’t work for realtime

weak cipher
low shard
#

it would run on the phone CPU literally

hallow thistle
#

Don't expect any AI to run that fast on mobile smartphone locally.

low shard
#

Yes it would work on cpu, but it’s going to take time

weak cipher
#

maybe try deploy on web and run?

hallow thistle
#

But if you host the service on cloud, and make it as a hybrid-web application, that would work.

low shard
#

I used RVC Applio on Termux on my honor 90 lite, it took 69 secs to inference 8 seconds

weak cipher
#

make it into a web and display it on your phone=))

hallow thistle
low shard
weak cipher
low shard
# weak cipher yes bro

if it’s going to run on cloud (remote good pc), rather than the phone’s power, then yeah it would work

weak cipher
hallow thistle
#

But you'll have to pay the cloud service for that.

weak cipher
#

free

hallow thistle
#

Yes, you can make a space on Hugging Face.

low shard
low shard
# weak cipher free

huggingface gives just CPU for free, it wouldn’t be that good for realtime

weak cipher
hallow thistle
#

But if you wanna have GPU for that, you'll still have to pay for that one.

knotty moth
low shard
weak cipher
knotty moth
weak cipher
knotty moth
chilly slate
#

#✨│ai-help hello is anyone here who is good in setup and use of RVC i was trying to use but i cannot i don't know how can anyone please help me its urgent

low shard
chilly slate
# low shard Elaborate your issue

rvc when i input my sample voice model.pth and index.index it shows error of missing hubert file downloaded and aaded then statted showing another like that

low shard
hallow thistle
chilly slate
low shard
chilly slate
low shard
#

Also you have permission to send SS, is there a reason why you want to do it in DMS?

hallow thistle
low shard
chilly slate
chilly slate
low shard
chilly slate
hallow thistle
#

You can screenshot the folder of it.

#

I can identify the RVC program by folder name.

low shard
#

I mean even the UI is recognizable

hallow thistle
chilly slate
chilly slate
chilly slate
hallow thistle
karmic oliveBOT
chilly slate
low shard
hallow thistle
#

Damn, I guessed it right. The RVC GUI is too old now.

low shard
#

RVC GUI is another fork made by t1g3r

#

This is Mainline RVC

chilly slate
low shard
# chilly slate

Show a screenshot of the error too, also is it a public model which you can send the model download link?

hallow thistle
#

Not the same one as RVC GUI, but still sounds like it.

low shard
hallow thistle
chilly slate
low shard
chilly slate
hallow thistle
#

Rmvpe model file is missing?

#

Are you running RVC from an installed Python and not compiled-portable one?

chilly slate
chilly slate
hallow thistle
#

That's so far messed up.

chilly slate
chilly slate
hallow thistle
#

Typical folder path for the already compiled RVC isn't supposed to be inside an installed Python folder, which is found in C:\Users\"your username"\ or Program Files, unless you're trying to develop a fork RVC by yourself.

low shard
#

you can delete that mainline you got

#

Your Nvidia GPU is good enough to do inference (use models) locally (on ur pc), not the best to train (make models) even if still possible

You can:

  • Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
    • Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
    • Mainline: The original RVC
  • Cloud (remote good pc, easier and faster than ur PC but it's limited):
    • Ilaria RVC Zero: fastest and simplest that you can get for free
    • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
    • Applio Colab: max 4 hours daily, not granted, of GPU

Easiest possible (automatically separates vocals & instrumentals) : weights.gg
easiest cloud: Ilaria rvc zero
easiest local: Applio

#

the local links I sent are guides for the precompiled version, if you want to do it locally

hallow thistle
#

You can't say you know everything on how to install a Python program when you struggle to get it to work by hands. boohooh

chilly slate
chilly slate
chilly slate
hallow thistle
#

!howtoask

patent trellisBOT
# hallow thistle !howtoask

How To Troubleshoot AIHC_WaitWhat

__**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.
weak cipher
#

too slow chat guys =))

image.png
hallow thistle
#

Now make it to speak in English. skullfacedistorted

weak cipher
#

I don't know if coqui still supports it guys.

low shard
chilly slate
low shard
weak cipher
low shard
#

I told you to uninstall the one you got, and to choose one of the working links

weak cipher
#

trash meta

low shard
#

I also explained you the difference between them

chilly slate
chilly slate
hallow thistle
weak cipher
low shard
hallow thistle
chilly slate
weak cipher
hallow thistle
low shard
#

also it has some limits like it can't laugh well

chilly slate
low shard
#

I'm not sure what you mean exactly with emotional expression though

low shard
chilly slate
hallow thistle
chilly slate
crude flame
low shard
weak cipher
chilly slate
low shard
low shard
# chilly slate so ?

so you can choose any

Easiest possible (automatically separates vocals & instrumentals) : weights.gg
easiest cloud: Ilaria rvc zero
easiest local: Applio

#

local = runs on your pc

weak cipher
#

don't know how to remove its limit or it has some error ❓

image.png
low shard
#

cloud = remote good pc

#

as you got a gtx 1650, cloud will be faster, but you will have limited gpu time on cloud

chilly slate
low shard
chilly slate
low shard
#

you would either use google colab or kaggle, which are basically remote good pc borrowed by google that run the code

#

it still runs the same RVC program

#

the quality doesn't change if you use it on cloud for free or locally

#

what changes is if you pay for cloud it will be even more faster

chilly slate
#

how much time limit of this google cloud?

hallow thistle
#

The quality of voice model depends on how you train it. The settings, audio dataset you use.

low shard
#

this is the free limit

hallow thistle
chilly slate
weak cipher
#

using hugface and no gpu how long will it take =)))

hallow thistle
#

Even if I had no experience of development in the past, I still know how to install RVC and any other Python related program locally. nails

crude flame
weak cipher
hallow thistle
#

Despite how weak my laptop is, RVC and Applio even worked, but took hours to finishing a single audio.

crude flame
low shard
#

it's your choice

chilly slate
low shard
chilly slate
low shard
chilly slate
low shard
chilly slate
low shard
chilly slate
weak cipher
#

why not free boohooh

low shard
#

TTS = Text to Speech

#

STS = Speech to Speech

#

LLM = Large Language Model, like chatbots

chilly slate
weak cipher
low shard
#

they have also a Speech to text integration

#

but you wouldn't be able to use it with RVC models

#

I don't really use SillyTavern though

weak cipher
low shard
weak cipher
weak cipher
weak cipher
low shard
#

it won't work with rvc models

#

tts = text to speech

#

sts = speech to speech

weak cipher
#

and to stt for another free cpu service?

low shard
#

a good stt is whisper-v3-large

low shard
weak cipher
#

with just one model i think it will be ok

weak cipher
low shard
low shard
weak cipher
#

english only

low shard
weak cipher
#

oh see

weak cipher
#

I see him using this much capacity with coqui

image.png
chilly slate
chilly slate
low shard
warm garnet
#

i need help

low shard
low shard
patent trellisBOT
# low shard !howtoask

How To Troubleshoot AIHC_WaitWhat

__**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.
low shard
#

Please, elaborate your help request so helpers can help you

warm garnet
#

everytime i try to open start_http it wont let me open

low shard
#

for example, tell:

  • your pc gpu
  • the link of the guide you're following
  • what's the issue
low shard
#

that file is only on that program

#

youtube tuts are old

#

don't follow them

warm garnet
weak cipher
warm garnet
#

12 months ago

chilly slate
low shard
# warm garnet 18a

that's an old version of the original wokada, it has worse performance and worse quality

weak cipher
low shard
#

also this is the wrong channel

warm garnet
#

oh wrong channel

#

so where the right one

low shard
# warm garnet oh wrong channel

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-reocrded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

pls tell ur pc gpu in #🔍│help-w-okada

chilly slate
low shard
weak cipher
chilly slate
weak cipher
weak cipher
chilly slate
weak cipher
chilly slate
weak cipher
chilly slate
weak cipher
#

and they used coqui llma to make it speak and convert text to speech

chilly slate
weak cipher
#

exactly that's what i wanted to do but it works on web brud

#

i think you know how to do

chilly slate
weak cipher
chilly slate
#

uh can create 3d model with it

weak cipher
knotty moth
weak cipher
chilly slate
weak cipher
chilly slate
weak cipher
#

I mean I want to be like him bro =))

knotty moth
weak cipher
knotty moth
weak cipher
knotty moth
weak cipher
#

Either way, it has to convert speech to text and vice versa.

weak cipher
#

I only have 1 64g bar left lol

knotty moth
potent helm
#

Is the a Applio Collab that's able to use the new hifigan pretrains?

weak cipher
low shard
fast scarab
#

Hello, I have a question about using Kaggle for training. If I turn off my PC, will the training process stop, or does Kaggle continue running the notebook in the cloud? Is there any way to keep it running even if I close my computer?

weak cipher
#

too slow

image.png
stone wyvern
#

can someone please help me ?

weak cipher
stone wyvern
#

yeah kinda

low shard
patent trellisBOT
# low shard !howtoask

How To Troubleshoot AIHC_WaitWhat

__**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.
low shard
#

please elaborate your request

stone wyvern
#

i dont know how to use these AIs to change my voice into my fav voice model

low shard
#

also what's your PC GPU

stone wyvern
#

pre-recorded

stone wyvern
#

i dont have a dedicated GPU

low shard
#

Train (make) RVC Models on cloud:

  1. Prepare the Dataset
  2. Setup RVC:
    Choose a cloud way to use RVC,
  • Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
  • Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
  • Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
  1. Be sure to know about the tensorboard

Google Colab = Easier but risk of getting disconnected
Kaggle = Harder but way more gpu time
If you are looking for the easiest way and for free, is using https://weights.gg which ofc uses RVC

RVC Inference (use models) on pre-recorded audio on Cloud

You can use either:

stone wyvern
#

idk how to use these

#

...

low shard
#

you just have to read it

stone wyvern
#

ook

low shard
#

if you just want the easiest way ever possible to do inference, try weights.gg

stone wyvern
#

and what about the hugging face one ?

#

i tried that once

#

that was good

low shard
stone wyvern
#

the reason why i need AI is

#

that

#

i want to change my pre recorded vocals into some famous rappers or singers

low shard
stone wyvern
#

okk

weak cipher
#

hugface gives 100GB for free, so if I upload 100GB of photos and transfer to CDN, will it violate their policy?

echo oasis
#

hey guys do you have any recommended free site for vocal isolation?

tame mica
trim dune
#

How to use voices zip to convert into ai voice

brittle forge
#

what to do in Realtime Voice Changer when I speak, my voice is choppy?

stoic viper
#

Hi,
I’m having issues using custom pre-trained models on Applio with RVC. Default models work fine, but with custom models (like KLM5 RefineGAN), I get this error:
“The parameters of the pretrain model such as the sample rate or architecture do not match the selected model.

low shard
low shard
low shard
stoic viper
# low shard did you use the main branch?

Yes, I’m using the main branch and the latest version of Applio (3.2.8 bugfix), but I’m still facing the issue. Default models work perfectly, but with custom pre-trained models like KLM5 RefineGAN, I get the error:
“The parameters of the pretrain model such as the sample rate or architecture do not match the selected model.”

low shard
#

are you doing it locally or on colab ui?

stoic viper
midnight crater
#

Hello, I couldn't find the answer to this question searching through the chat history, so hopefully this isn't a repeat question with a really obvious answer. I'm trying to do a song cover of a female voice singing Fuck Her Gently by Tenacious D. Most the song goes just fine, but at the end when he rises into his falsetto the model becomes staticy and robotic. I'm using the base RVC program, and I've tried messing with the different settings with no improvement. Any advice or guidance is welcome.

vagrant marsh
#

0%? I guess something is wrong

Screenshot_2025-02-22-23-16-09-053_com.android.chrome.jpg
#

What are these percentages on the epoch?

#

Btw

tough shuttle
#

how do i train from one of my save points

green shell
#

where do i get rvc

hallow thistle
hallow thistle
hallow thistle
molten fog
#

what sampling rate would i use in this case?

brittle wing
#

-colab

karmic oliveBOT
# brittle wing -colab
📒 Google Colab Notebooks
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, -realtime, -rvc, /help
knotty moth
molten fog
#

so i cant tell what im actually at

weak cipher
molten fog
knotty moth
molten fog
weak cipher
chilly slate
#

#✨│ai-help will anyone help me in setup of Applio and then integrate it with ai assistant?

carmine siren
#

-kaggle

karmic oliveBOT
# carmine siren -kaggle
📘 Kaggle Notebooks

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

More commands: -colabs, -hf, -local, -rvc, /help
weak cipher
#

Guys I have skill issue, where can I get hugface api?

molten fog
#

is this ot or keep going?

thin edge
#

how to get higher "khz" for dataset ?

#

im stuck with 15 khz voice

simple ore
thin edge
simple ore
#

????

thin edge
# simple ore ????

You know... just forget about it,

What happens if I increase the gpu batch size ?

simple ore
#

i mean.. where did you find 42KHz?

thin edge
simple ore
#

there's 32k.. or 44k

#

not 42k

thin edge
simple ore
#

as for batch size - larger the batch size is, larger the stride it makes each step

#

could reach the goal faster, but can also miss it

thin edge
simple ore
#

depends on the dataset size

thin edge
simple ore
#

yes

thin edge
#

6 and half minute

simple ore
#

4

thin edge
#

and i find this too, what should i do ?

colab.PNG
simple ore
#

you're using some old software

thin edge
chilly slate
#

will anyone help me in setup of Applio and then integrate it with ai assistant? @low shard

hallow thistle
low shard
chilly slate
hallow thistle
chilly slate
chilly slate
low shard
low shard
weak cipher
low shard
#

?

weak cipher
#

I have tts and stt, now how do I create a chat bot?

chilly slate
low shard
honest dew
#

yo

low shard
molten fog
#

where in this would be considered the overtraining point?

#

im thinking at that third notch before it indefinitely rises but im not sure if im correct or not

low shard
crude flame
simple ore
molten fog
simple ore
#

show fm and mel charts too

molten fog
#

lmfao i accidentally refreshed my kaggle page so i mightve just botched the training altogether but no big deal tbh

simple ore
#

fm ooof

#

that's way too much. what's dataset size / batch size ?

tropic phoenix
cold tinsel
#

ello. is a batch size of 8 for 15min of data a good place to start?

lone temple
#

is there any tool that allows phonetic tts with rvc models?

analog obsidian
vestal cloud
#

how to i get Unwa's big mel roformer beta 4 on uvr?

simple ore
lone temple
knotty moth
simple ore
#

a non-LLM based TTS model can use phonetics to produce the audio, although it will be relatively bland

#

unless the model uses both phonetics and an LLM/GPT engine

knotty moth
hallow thistle
crimson depot
#

One question. Is index file mandatory for IA covers?

glossy grove
#

@crimson depotno it's not. Sometimes it can help, sometimes it can hurt.

crimson depot
glossy grove
#

sometimes

crimson depot
#

For diferent language

weak cipher
dusty ivy
#

-overtrain

karmic oliveBOT
royal marsh
#

anyone know how to change hz in steel series engine?

image.png
paper finch
#

While trying to change the voice with Harvest, RVC shuts down, and Python says 'press any key.' When I press a key, it closes without giving any error code.

simple ore
#

and python crashed

distant tundra
#

Hi, is there any software like RVC but it can clone a voice just with some seconds of a wav file?

tame mica
#

seed-vc does that

#

though it's really not that good

distant tundra
distant tundra
distant turtle
#

-colab

karmic oliveBOT
# distant turtle -colab
📒 Google Colab Notebooks
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, -realtime, -rvc, /help
chilly slate
low shard
chilly slate
low shard
#

hyperlink is a blue text, that when clicked will redirect you to the link

chilly slate
low shard
chilly slate
distant tundra
stoic viper
#

Hey everyone,
I’m training a voice AI model with a 35-minute recording, but I heard that using too many Epochs can make the voice sound robotic.
Any advice on how many Epochs I should use to keep it sounding natural? Also, if anyone can explain how Epochs really work, I’d appreciate it!
Thanks!

distant turtle
#

-colab

karmic oliveBOT
# distant turtle -colab
📒 Google Colab Notebooks
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, -realtime, -rvc, /help
polar tendon
tame mica
low shard
nocturne pine
#

Is RVC GUI still viable?

low shard
#

Don't follow yt tuts

#

What's ur PC GPU and what do u want to do

brittle gazelle
#

Hey does rvc runs in the 5000 series ?

simple ore
#

there's pytorch, but no torchaudio

fast scarab
#

Hey, when should I stop the training? It looks like it's starting to overfit. Thanks!

image.png
distant turtle
#

-colab

karmic oliveBOT
# distant turtle -colab
📒 Google Colab Notebooks
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, -realtime, -rvc, /help
stray raven
#

yo dudes getting a rtx 5080 card soon, i understand from previous messages here that its not working with the voice changer atm? is thats still the case?

simple ore
#

still no updates

knotty moth
crimson depot
#

Local CoverGen????

#

I need one

proud hound
#

-hf

karmic oliveBOT
glass igloo
#

Share a link to the official github of rvc2 and maybe there is some modern tutorial how to install and train the model? I found a tutorial video on youtube but there is a link to github where the last update is 2023.

low shard
#

Tell your PC GPU and what you want to do first

glass igloo
# low shard YouTube tuts are very old

I've written a story set in the witcher world, and I'm currently making a youtube video where I voice characters with voices from the game. I have trained models for so-vits-svc, but they work badly, with speech defects. I want to train models for RVC2 in hopes that it will work better. I have a GTX 1080 video card.

low shard
#

Your Nvidia GPU is good enough to do inference (use models) locally (on ur pc), not the best to train (make models) even if still possible

You can:

  • Locally (runs on your pc so the speed depends on that, you will have to set it up with the guides):
    • Applio: A fork of RVC with some extra features like Applio TTS, kinda faster and simpler but same quality tho
    • Mainline: The original RVC
  • Cloud (remote good pc, easier and faster than ur PC but it's limited):
    • Ilaria RVC Zero: fastest and simplest that you can get for free
    • Weights.gg: Partnered with AI Hub, lets u do them easily but u may be in a queue
    • Applio Colab: max 4 hours daily, not granted, of GPU

Easiest possible (automatically separates vocals & instrumentals) : weights.gg
easiest cloud: Ilaria rvc zero
easiest local: Applio

hallow thistle
#

SVC is too old now.

glass igloo
low shard
slim mauve
#

Hi guys! Do you know how can I create my own voice model for RVC-Gui? Do I have to take a 10 minutes recording and the renaming it with the .pth extension?

low shard
#

delete it

#

don’t use youtube tuts

#

and no model training isn’t just renaming the extension

#

what’s ur pc gpu

slim mauve
#

Thank you! My gpu is NVIDIA GeForce RTX 3060

low shard
#

With Applio/Mainline you can do both training and inference on pre-recorded audios, they are more updated RVCs

slim mauve
#

Thank you!

low shard
distant tundra
#

Hi, where should I put the .index file of a model (using RVC here)

zinc raft
#

-colab

karmic oliveBOT
# zinc raft -colab
📒 Google Colab Notebooks
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, -realtime, -rvc, /help
distant tundra
#

ok I installed applio, dunno where to put the models

simple ore
next karma
simple ore
low shard
hallow thistle
knotty moth
cold tinsel
#

hmm, which sampling rate is best to choose in applio?

Hozdata.wav.png
hallow thistle
#

Either 40000 or 48000, but 48000 gives better quality.

knotty moth
cold tinsel
#

ok ty. i think i’ll go with 40k

simple ore
#

32k probably

#

but you can go with 48 as well

knotty moth
#

consistency is more desirable, not ideal to mix up from multiple sources that may have different cutoffs, but at least you can go 32k

simple ore
#

different room / different reverb

glass igloo
#

Hi. Can I change the dataset when training a model? For example, I originally had a file 30 minutes long and I want to change it to 1 hour. How do I do that? Do I just change the file in the dataset folder or do I have to do something else?

final arch
#

hey im having a problem that i havnt gotting into since today, it seems like my inputs get filtered alot when i convert into ai, doesnt matter which voice im running into. Anyone with some good settings to help me out with?

unique rock
#

How do you enable the Use RefineGAN options in Applio colab? + KLM 5?

glass igloo
#

Hi, can you please tell me where the audio file is saved after conversion? In the interface there is an option to download the generated file but it takes too much time, I would like to be able to work directly with the generated audio.

#

It would be even better if you could specify a place to save the converted file automatically.

white bough
#

Is anyone else having problems with the Applio ColabUI? It cannot find the gradio module. It was working an hour ago

#

Traceback (most recent call last):
File "/content/program_ml/app.py", line 1, in <module>
import gradio as gr
ModuleNotFoundError: No module named 'gradio'

simple ore
#

virtual environment ded

#

reinstall

white bough
simple ore
#

yes

white bough
#

What should I do to make it work? How do I "reinstall"?

simple ore
#

i guess?

#

someone else may know for sure, I dont use colab

#

the error you got is a missing requirement

white bough
#

Oh I did reinstall. Looks like it's a Colab problem. Not sure

simple ore
#

I assume when you discronnect from colab it deletes some installed stuff or something

distant mulch
#

is there any way to make ai covers with my phone

lucid glacier
#

Help

#

The collab named "CoverGen_NO_UI_v2_en.ipynb" appears with this section: The installation takes about 3 minutes, if it takes much longer ping me at AI HUB

Pitch_Change:
12

#

What I do?

#

@cyan hare

simple ore
#

uv is a replacement for pip install

white bough
#

Sorry deleted my message, but indeed replacing !uv run by !python works, although I see some other warnings now. But it seems to work

#

It works for now. So that's good enough for me ^^

lucid glacier
#

El Colab denominada "CoverGen_NO_UI_v2_en.ipynb" aparece con esta sección: La instalación demora alrededor de 3 minutos. Si demora mucho más, envíeme un mensaje a AI HUB

Cambio de tono:
12
¿Qué hago?

void ravine
#

my colab's not workiing

#

does anyone have a working google colab link

shadow moth
white bough
#

I replaced uv run by python in the cell and it worked

shadow moth
#

Thanks a lot!

weary peak
#

When starting to train a model I get an error at 5.3 seconds no matter what I run it own, I can seem to get an error code either. To be fair I'm stupid, idk if it has to do with pytorch and python dont work 100% of the time on my PC

simple ore
distant hamlet
#

anyone getting this on mainline colab? it was working perfectly fine just this morning 😑

Screenshot_2025-02-26_at_3.54.51_PM.png
weary peak
simple ore
#

there should be output in the console window with the error

weary peak
#

is colab an easier process, found one error msg AttributeError: 'FigureCanvasAgg' object has no attribute 'tostring_rgb'

simple ore
#

old colab

formal wind
#

is kaggle or applio no ui better?

tame mica
#

kaggle would be better but if you need the latest branch use noui colab

woeful cave
#

applio isnt working?

simple ore
#

if you mean colab, it seem there's an issue with requirements install

woeful cave
#

aww man

simple ore
#

replacing uv with python may work in the install cells

formal wind
tame mica
#

like

#

the branch that could use the non hifigan vocoders

formal wind
#

I have no idea what tf any of that means kar

#

Ill try my best to figure it out ig

tame mica
#

oh 😭

#

like basically the latest version of applio

formal wind
#

Thank you lol 🙏 (Mb for being stupid)

tame mica
#

nono dw its fine xd

weak cipher
#

Hey guys have you used Whisper Tiny yet?

frozen ledge
#

-colab

karmic oliveBOT
# frozen ledge -colab
📒 Google Colab Notebooks
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, -realtime, -rvc, /help
tame mica
proper mountain
#

ModuleNotFoundError: No module named 'gradio'

what should I do? it's on Hina's Mod AICoverGen. it was fine last night 😌

white bough
#

From what I understand, the command "uv run" was supposed to call "python", but it does not anymore. So replace all the "uv run" instances by "python" and check if it works

tawdry spade
#

Is there a solution for Mainline Google colab?

low shard
glass igloo
#

Can anyone know why in applio, when I start the continuation of model training, the epoch time increases several times? When I start training from the beginning the epoch lasts about a minute, but if I finish training and then continue, the epoch takes 5 minutes.

tawdry spade
unique rock
#

What error does Applio Colab present at this time?

fading relic
#

how do i stop delayed voice

vapid mantle
#

-colab

karmic oliveBOT
# vapid mantle -colab
📒 Google Colab Notebooks
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, -realtime, -rvc, /help
low shard
# fading relic how do i stop delayed voice

This is the wrong channel

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-recorded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

Tell your PC GPU, the guide link you're using and a screenshot of your wokada in #🔍│help-w-okada

hallow thistle