✨│ai-help

[glacial pollen]

if I was you I'd either purge pip and python then repeat or use conda

I guess all is a consequence of moving to forks or rather, solutions that are too far away from mainline

[analog obsidian]

yea im gonna remove pip + python better xD

[glacial pollen]

just get clean 3.10.6 and only use the nvidia-standard bat

[analog obsidian]

gotcha tho last time i tried that and i got the first error of torch

i suppose i have to downgrade it before starting bat

[glacial pollen]

nono, don't touch the torch

oh

[white edge]

is it possible to use text to speech with rvc gui?

[analog obsidian]

no

[glacial pollen]

you're telling me it got you 2.1.2 by default? @analog obsidian

[white edge]

aight thx

[analog obsidian]

im not quite sure, look im gonna reinstall pip and then run nvidia standard

and im gonna show u what it installs

im sure it did install 2.1.2 by default

this is so confusing lol

but i'll do it again to confirm if it does install 2.1.2 or not

brb 5 minutes

[glacial pollen]

It's weird, cuda 117/118 should give you torch 2.0.1 regardless

as each torch iirc has it's own supported cuda version ( or range of versions )

might be wrong on that one but, probs not

[analog obsidian]

hmm, okay pip and python are gone now, gonna install python 3.10.6 and see

[glacial pollen]

Most likely it's due to the fact you first went for standard

get 3.10.6 and use 118 bat, ye

[analog obsidian]

oh

[glacial pollen]

no other torch fiddling

[analog obsidian]

ok ok

oh man i hope this works

xD

[glacial pollen]

tf

2.3.1

@analog obsidian cancel it

[analog obsidian]

Ok

[glacial pollen]

Imma just isolate my versions

tired of this shit ngl

brb

[analog obsidian]

[rough ridge]

i know svc is outdated but how do you create one using voice clips?

[glacial pollen]

@analog obsidian

• torch 2.0.1
• torchaudio 2.0.2
• torchvision 0.15.2

fairseq 0.12.2

there

those 3 up with cu117

so 118 should work as well

[analog obsidian]

i install them manually via pip?

[glacial pollen]

hold on, gonna get you a string

[analog obsidian]

thx

[analog obsidian]

hmm, use rvc for voice cloning

[glacial pollen]

pip install torch==2.0.1+cu117 torchvision==0.15.2+cu117 torchaudio==2.0.2+cu117 -f https://download.pytorch.org/whl/torch_stable.html

[analog obsidian]

k thanks

[glacial pollen]

tell me if it works

[analog obsidian]

ok so first im running that command in a random cmd window
then i suppose i have to run cu118 bat

is that fine?

[glacial pollen]

nah, after that is done

just do the

this

pip install -r rvc-requirements.txt

[analog obsidian]

ohh ok

[paper thunder]

isHello guys, before it was possible to use Google LCollab servers on Voice Changer, is this possible now?
Sorry for the English, I'm using a translator

[glacial pollen]

cause what batch files do is the installation of torch and other thingies and then, the reclist

[rare goblet]

Ayo? @paper thunder level 1 !!!

[glacial pollen]

Aside that, if it turns out to be working fine, guess I have to re-edit the files

[rare goblet]

Ayo? @glacial pollen level 7 !!!

[analog obsidian]

https://colab.research.google.com/github/hinabl/voice-changer-colab/blob/master/Hina_Modified_Realtime_Voice_Changer_on_Colab.ipynb

[glacial pollen]

gotta hate when known and working solutions suddenly start causing issues

[analog obsidian]

its fine, at least you know how this behave in 40xx series

[glacial pollen]

it's weird in general

Ever since I remember, standard users if it wasn't the case for 30xx users

would simply do normal installation without defining the versions

I assume when i was gone, some stuff changed with torch

or some handlings, shit no clue

[analog obsidian]

they changed something in 40xx probably

[glacial pollen]

more so with torch itself

maybe what was nightly, stopped being nightly builds

and is considered the go-2 now

no clue ¯_(ツ)_/¯

either way, will reupdate the stuff

[paper thunder]

but it says it doesn't work

[analog obsidian]

what? you should be able to run it

[analog obsidian]

oh it installed without issues now

[glacial pollen]

Neat, now to confirm the rvc working

if it detects your gpu, you're all set

[analog obsidian]

yeaaa it woorks omg

[glacial pollen]

Awesome

[glacial pollen]

that'd mean, whatever works for 30xx works for 40xx

rvc developers be like lol

Instead of saying right away that 30xx and above require XXXX

they just go exclusive for ampere

smh

[analog obsidian]

[glacial pollen]

Whatev. Gonna update the stuff. Now, enjoy the fork

[analog obsidian]

thx codename

[glacial pollen]

np, also, honorable mention to emojikage for pointing out the " > " lol

[analog obsidian]

yeah xD

[glacial pollen]

I tend to overlook simple things when explaining n multitasking

[analog obsidian]

btw one last question, your fork doesnt have autosync right? i have to manually sync it?

[glacial pollen]

Correct

Autosyncing is garbage and to be avoided at all cost

[analog obsidian]

amazing thanks

[glacial pollen]

Once I regain the motivation to continue, I'll automate it

as in, test train, acquire the stuff, reupdate files, repaste config and rerun

but when and if? we'll see

[wispy lodge]

Idk where you got that, but it doesn't depend on GPU directly. It's the minimum driver version that matters. As long as your gpu supports >=527.41 (and you have it installed), you can use cu121 without any issues

[glacial pollen]

You should go to their repo and tell them that instead

[proven hill]

not exactly a good moment for them tbh

[glacial pollen]

However one thing is sure, not all cuda version cooperate well with certain things

and vice versa

a working combo is definitely 117 and 118 with torch 2.0.1 to properly support mangio crepe and crepe

[proven hill]

isnt mangio crepe useless

[glacial pollen]

it is not

It is still the highest quality method for high quality audio

[analog obsidian]

no???

[glacial pollen]

but given how rare it is to see people with actual high fidelity audio

If you have legit clean audio and pair it with crepe, expect miracles
if you don't, rmvpe is your godsend

that's all there is to it

[wispy lodge]

Crepe works fine even with latest torch and cuda (if we're talking about torchcrepe)

[proven hill]

idk, i dont bother to port it anymore in my forks

[glacial pollen]

I am pretty sure this is not the case

If it was so simple people wouldn't have torch and cuda compatibility issues

and there wouldn't be certain criteria to be met

[wispy lodge]

I can assure you since I run it in realtime

[glacial pollen]

being 117 and or 118

[wispy lodge]

And my wokada fork uses latest torch and cuda

[glacial pollen]

Realtime is different to rvc

this is not 100% the same mechanism

[analog obsidian]

btw this downgrade shouldnt affect realtime?

[glacial pollen]

It won't

https://github.com/codename0og/rvc-realtime-voice-changer

There's a solution for that too

[wispy lodge]

Inference process is the same, except there are optimizations for realtime

[glacial pollen]

it's fully separated and independent from rvc

but shares the reclist for my fork

[analog obsidian]

oh nice, i actually prefer to use realtime_gui xD

[glacial pollen]

so you're all set just in case ✨

Yeah, it is better

[analog obsidian]

perfect, thanks

[glacial pollen]

✨ Enjoy

[glacial pollen]

I think there's no point in discussing it further as you've seen the real-scenario case of when version matters

They did get 2.1.2 along with it's respective n compatible cuda for 4090

and rvc did not recognize it

same was for some other person which required cuda 118 as 117 and above 118 wouldn't budge

lots of components rvc uses are outdated and or kept for lack of higher ver compatibility issues

Whether it is changed now or not, is another thing

( such as upping the python support to 3.11 or 3.12.x and torch changes in applio, and so on )

[brittle wing]

how do i delete everything in audacity that isnt under a label

like delete it not mute it

[analog obsidian]

select it and click supr

o

del in english

xD

[brittle wing]

no like the non labeled parts

cuz i have way too many labels to manually

do that

or is there no way to automatically do that

[analog obsidian]

pretty sure if you select export labels only, it will not export the non labeled audio

[brittle wing]

yes but i need to select which parts r good for model

too much silence gaps so

it would be easier to remove it all so i can quickly just listen back to all of it

[analog obsidian]

audio labels include silence but its a small amount and rvc can handle those

if ur getting huge amount of silences in the label then idk honestly

[wispy lodge]

Well, let it matter for rvc, I don't mind :)

[glacial pollen]

That's all there is to it

it's an rvc focused channel

[brittle wing]

no the problem is theres alot of screaming

and it messes the model up cuz i did a test w 200 epochs and it randomly gets loud

i didnt think there was that much so i just ignored it at first

[analog obsidian]

i.. told you to fix the distortion

[brittle wing]

yes i did but the original is like

the mic clips too bad u cant make it sound better w effects

[analog obsidian]

then remove it xD

[brittle wing]

yes but theres

too

much

so

[analog obsidian]

rvc cannot handle screaming, laughing well

[brittle wing]

i wanna remove all the

gaps

cuz its basically like 30 mins of silence in total

im just manually doing it it takes So long

[analog obsidian]

[white edge]

how does one add a model to applio

[analog obsidian]

http://docs.aihub.wtf/rvc/local/applio/#inference-

[ornate summit]

My RVC stopped working and now when I try to run it I get this instead. I don't know code. What does it mean? Is there an easy fix?

[analog obsidian]

hmm you didnt changed anything? probably it corrupted tho not sure

[ornate summit]

I plugged in the Mr. Krabs Laughing model from #1175430844685484042 then went to switch over to the Pyra model and the program crashed. I reopened and it's been doing this ever since.

[analog obsidian]

i suppose it corrupted, try downloading it again

[rare goblet]

Ayo? @analog obsidian level 23 !!!

[analog obsidian]

it should work

[ornate summit]

Okay, thanks!

[rare goblet]

Ayo? @ornate summit level 2 !!!

[brittle wing]

finally done i deleted 50 mins of audio in total

bro can i like

[rare goblet]

Ayo? @brittle wing level 7 !!!

[brittle wing]

remove these from audacity

theres too many

[subtle kernel]

Im kinda new to this thing so how does this work on mobile

[hot trail]

whats the amd rvc again?

[hasty harness]

can anyone help me in vc?

[lean owl]

is there a way to isolate two singers from a song

like for example if theres two people singing at the same time like a male or a female singer

oh this

because when i think of back vocals i think of lead vocals with harmonies n stuff

i might try

[cyan maple]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

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

[shrewd nebula]

does the overtraining detector function work in applio

[tame mica]

go to uhh plugin manager and disable ones u dont waant

[shrewd nebula]

what should i choose here if im using a custom model

[noble dawn]

where do i get my zip file ? im like so lost lmao

can someone kindly walk me on it ?

[violet heron]

zip file for what?

[noble dawn]

I figured it out we good 💯🫡

[noble dawn]

But thank u for the response

[brittle wing]

@proper shale you around?

this is what i downloaded

the modles folder i put myself, saw it in a video

and containes this

[neat glade]

hi um i just installed RVC today and for some reason it wont changes the voice even if i left the settings unchanged, any idea why?

[glacial pollen]

wdym it won't change voices? you're meant to pick the model ( voice ) on your own

what rvc you using

[neat glade]

yeah i know im currently using the stock one

[glacial pollen]

so, you don't know how to switch models?

or

[neat glade]

no i thought my output is playing the sounds thats not been through rvc

its already working when i streamed it, sorry for bothering

[rare goblet]

Ayo? @neat glade level 1 !!!

[glacial pollen]

streamed.. aren't you confusing real time voice changer or w-okada with rvc?

rvc =/= real time voice changer

rvc means retrieval based voice cloning / conversion

[opal wave]

best girl voice changer?

[brittle wing]

[patent pasture]

Can you remove Ilaria RVC from colab list?

and add Ilaria RVC Mainline

[small raft]

HELP ME I KEEP GETTING THIS ERROR AND I DONT KNOW WHY JSONDecodeError Traceback (most recent call last)
<ipython-input-12-8ccb95324ce0> in <cell line: 31>()
31 if os.path.exists(config_path):
32 # File exists, proceed with creation of creds and client
---> 33 creds = Credentials.from_service_account_file(config_path, scopes=scope)
34 client = gspread.authorize(creds)
35 else:

5 frames
/usr/lib/python3.10/json/decoder.py in raw_decode(self, s, idx)
353 obj, end = self.scan_once(s, idx)
354 except StopIteration as err:
--> 355 raise JSONDecodeError("Expecting value", s, err.value) from None
356 return obj, end

JSONDecodeError: Expecting value: line 1 column 1 (char 0)

[rare goblet]

Ayo? @small raft level 1 !!!

[small raft]

oh and heres a screenshot of it

I'm Stumped I honestly don't know what's causing this error

[pastel oak]

outdated colab

write

-colab

in this channel and use something more up to date

or open the docs in pinned message and select one of those

[pastel oak]

you should get used to the fact people think rvc means real voice changer

[pastel oak]

send screenshot of your app

[glacial pollen]

ofc, but that doesn't mean I won't correct them in a way

[pastel oak]

thats the wrong shortcut get the one thats start_http without the s at the end

[pastel oak]

yea its just a waste of energy hahaha

also good to see u back

[glacial pollen]

✌️

[brittle wing]

yeah i did that already, able to setup everything except the voice.. like the one im using is kinda chunky

also its not the perfect girl either.. do u have any that sound good?

[pastel oak]

look for egirl, mommy, asmr, streamer voices

[brittle wing]

got any top ones recommendations? i just wanna sound good enough to troll a group chat

like link or smth which has the top most quality and best sounding ones

[pastel oak]

Not really i havent looked at the models in a long time

mommy mae sounded good

theres been better ones probably by now

[brittle wing]

okayy

[floral tinsel]

is there gus fring ai voice?

I can't find it really

[gusty kestrel]

An error occurred

Sorry, I could not find models that match your search "gus fring"

@brittle wing, I have found 5 results that match your search!

kislottxd

Gus (Brawl Stars)

• Uploaded: <t:1704574536:d>
• Server: AI HUB
• Likes: 0
• Lang: EN

Epochs

250

Technology

RVC

Algorithm

Rmvpe

Requested by fuckkissah · Jun 16, 2024, 10:08 AM

[floral tinsel]

or at least giancarlo esposito

[brittle wing]

Nope

[floral tinsel]

that's the actor name

[gusty kestrel]

@brittle wing, I have found 2 results that match your search!

megaaziib

Giancarlo Esposito / Gustavo Fring (Breaking Bad)

• Uploaded: <t:1698697142:d>
• Server: Unknown
• Likes: 0
• Lang: EN

Epochs

400

Technology

RVC

Algorithm

Rmvpe

Requested by fuckkissah · Jun 16, 2024, 10:08 AM

[brittle wing]

@floral tinsel

[floral tinsel]

life saber

[rare goblet]

Ayo? @floral tinsel level 1 !!!

[floral tinsel]

saver

thx

[bold yarrow]

ilaria mainline rvc seems to be having trouble converting

i can download models just fine, but converting results in an error

[brittle wing]

is there a vr model to seperate backing vocals/adlibs

[gusty kestrel]

An error occurred

Sorry, I could not find models that match your search "BERKCAN GÜVEN 'BEGE'"

An error occurred

Sorry, I could not find models that match your search "BERKCAN GÜVEN 'BEGE'"

[proper shale]

i was sleeping 😭 sorry

[brittle wing]

is this the right one

[proud elbow]

yes, though not always work on adlibs

[brittle wing]

[hardy wasp]

i cant select gpu

who can help me

and i cant convert fast

[pastel oak]

how dare you sleep

[hardy wasp]

can you help me

[pastel oak]

whats your gpu name
and send screenshot

[brittle wing]

i got most of it sorted, just need a good voice model now

[hardy wasp]

gt740 m

how can send ss

[pastel oak]

nah that gpu is not strong enough

its old

[hardy wasp]

my cpu better

[pastel oak]

doesnt matter

[hardy wasp]

but its so slow

can it be fast

[pastel oak]

what program are you using

[hardy wasp]

ryc gui

[pastel oak]

use colab to run it online

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

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

[pastel oak]

use applio

first link

[hardy wasp]

thanks

how can i use it

i dont know

[rare goblet]

Ayo? @hardy wasp level 2 !!!

[pastel oak]

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

tutorial

[fierce anchor]

https://colab.research.google.com/drive/1Gj6UTf2gicndUW_tVheVhTXIIYpFTYc7?usp=sharing
I'm running this sample code and there's an error in the next part. Does anyone know a solution?

creds = Credentials.from_service_account_file(config_path, scopes=scope)

[hardy wasp]

thanks

how can i add

voice models

i dont understand

[pastel oak]

So you have 2 options

1. you have a huggingface or google drive link and like in screenshto you enter url, then click on "download model"

2. below "drop files" you can press upload box and input model pth. and AFTER you do index file

[hardy wasp]

can it be zip

[pastel oak]

for LINK, yes it SHOULD be zip

for manually "Drop files", no you have to do first pth then index

[hardy wasp]

i accitendtly download zip

[pastel oak]

if you do manually just take it out of zip

its not a problem lol

[hardy wasp]

need i copy voice models link?

i understand now thanks

[pastel oak]

👍

[hardy wasp]

i downloaded

what i need to do now

[hardy wasp]

what is the best settings to convert

[pastel oak]

the standard are already good

you can maybe increase the one thing thats at 0.3 saying adjust volume to original up to full and u can reduce index if you notice some artifacring

[hardy wasp]

ok

need i embedder model

[pastel oak]

Ehm i dont know what that is rn i havent infered in a while

[hardy wasp]

ok thanks

[odd shale]

Estoy confundido si quitaron el Hubert de Applio.

O en realidad era que se usaba contentvec todo el tiempo

Son lo mismo?

[hardy wasp]

how much pitch i need for girl voice model to boy sound

[odd shale]

what

You want a girl model to sound like a boy?

[hardy wasp]

yea

music

for

[odd shale]

Ehh.. I would suggest you to lower the pitch to -12 or beyond tho it will probably won't sound well.

You can't normally make a girl model sound like a boy

You instead should use a male model

[hardy wasp]

-12 or 12

[odd shale]

-12

[hardy wasp]

but its sound like boy now

[odd shale]

Well, in that case it's fine then.

[proud elbow]

any kind of time/pitch shift could do it

[odd shale]

It's mandatory to put DC Offset to zero on any audio Mumei?

[quick jungle]

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

[hardy wasp]

i didnt understand i m not using this program

[brittle wing]

@quiet vale is my voice convincing enough?

i did

its a voice changer btw

im trolling a gc

[proud elbow]

doing it before bs-rofo/separation models and further processings could help improving it, though just give it a try

[brittle wing]

oh

i did

[proud elbow]

you just need this

[brittle wing]

thats correct?

[hardy wasp]

ok thanks

[rare goblet]

Ayo? @hardy wasp level 3 !!!

[brittle wing]

yes

[rare goblet]

Ayo? @brittle wing level 3 !!!

[brittle wing]

@quiet vale do yk whats wrong?

me?

@quiet vale help me once done

yes

no

1 sec

whats happening lol

the girl voice is me

@quiet vale can u tell me which softwares u use also the voice pack

can u come dms?

@quiet vale gimme a call once ur free from here

[brave swallow]

@brittle wing Hi , can we get In touch Im arabic content creator
I been trying for over 30 days to clone my voice but all the things I tried failed so far 🥲

and Eid Mubarak 😄

[brittle wing]

LOL

btw

how i can help you

[fading drum]

anyone know how to resume rvc1006nvidia training?

[brave swallow]

I need for little suggestions and guidance
https://discord.com/channels/1159260121998827560/1251907012422402049

[rare goblet]

Ayo? @brave swallow level 1 !!!

[azure marsh]

Documentation

📚 All-In-One English documentation

https://docs.aihub.wtf/

Full AI Voice Model Training Guide (Local)

Link: YouTube
Credits: Christopher Villanueva

Model training with Mainline RVC

Link: Rentry
credits: Raven (ravencutie21)

AICoverGen Colab Guide

Link: Google Docs
Credits: Eddy (Spanish Helper)

Create a model with RVC disconnected (colab)

Link: Google Docs
Credits: Angetyde

🍏 Applio-RVC-Fork

Enhanced RVC Variant: Optimized Performance Through Modifications, Built upon Mangio-RVC-Fork.

[brave swallow]

the question deleted LoL

[shut crane]

Hello people,
Im trying to make an ai cover with a goku black voice model from this discord server, but de google collab wont go through with it.
creds = Credentials.from_service_account_file(config_path, scopes=scope)
this line of code is underlined in red, does anyone know what this means and how to fix it?
I've tried with other voice models, but the same line of code is allways the problem.

[azure marsh]

Suggestions for @shut crane

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

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

[shut crane]

thanks! though is there one in particular which is best?

[analog obsidian]

applio is the easiest to use

[shut crane]

Cool, thx! I'm on the webpage rn. I just press install Applio right?

[rare goblet]

Ayo? @shut crane level 1 !!!

[brave swallow]

is there a good TTS that I can use for applio ? because it uses Microsoft tts which sounds garbage

[analog obsidian]

Yea install applio, wait until cell is completely off, then start applio

[analog obsidian]

Elevenlabs

[brave swallow]

isnt this paid cloud based service?

[analog obsidian]

You generate audio there and inference in rvc

[analog obsidian]

10k characters per month free

[brave swallow]

hmm , I was thinking about XTTS then rvc with applio ( I assume applio is rvc right?)

[analog obsidian]

Yes, applio is a rvc fork

[brave swallow]

hmmm

[analog obsidian]

Try the tts you like the most, i like elevenlabs

[brave swallow]

the issue I couldn't find good Arabic TTS

if it's on English many works actually

[brave swallow]

but I been struggling with this for while .

[analog obsidian]

Elevenlabs support arabic language

[brave swallow]

I'll give it a try then

[analog obsidian]

Use multilingual model

V2

And arabic text should work

[brave swallow]

I don't want to rely on something trail

[brave swallow]

oh nice!

because the pronouns of the letters is the issue

each latter has 3 different audio based on the context

[rare goblet]

Ayo? @brave swallow level 2 !!!

[analog obsidian]

Try if elevenlabs has proper arabic pronunciation

[shut crane]

i just pressed start applio and now its downloading a whole bunch of things, is that supposed to happen?

[brave swallow]

sure I'll definitely try it

[analog obsidian]

Thats normal, wait until green check appears

[shut crane]

also something called tensorboard is on my screen

[brave swallow]

because I tried everything for the last month

[analog obsidian]

Ignore it, is for training purposes

[shut crane]

oh oke, i was kinda worried.

[analog obsidian]

Most tts are paid sadly

[brave swallow]

Oh I see

so the dream of having full local based project won't become real

💀

[analog obsidian]

[brave swallow]

btw what's the technology they use called ? for the TTS training

I don't see many information about this part

only coqui Xtts

[shut crane]

No dashboards are active for the current data set.
Probable causes:

You haven’t written any data to your event files.
TensorBoard can’t find your event files.
If you’re new to using TensorBoard, and want to find out how to add data and set up your event files, check out the README and perhaps the TensorBoard tutorial.
If you think TensorBoard is configured properly, please see the section of the README devoted to missing data problems and consider filing an issue on GitHub.

Last reload: Jun 16, 2024, 5:08:58 PM
Log directory: logs

this is what tensorboard is "saying" thats fine?

[analog obsidian]

Neither of those supports arabic

He needs an arabic one

[brave swallow]

is it same as whisper?

[analog obsidian]

Ignore tensorboard, this is not gonna affect your inference

[shut crane]

alright, its just taking a long time to process clicking the start button.

[analog obsidian]

Gpt sovits uses whisper for transcribing the audio iirc

But only supports chinese and english

[brave swallow]

yep that's the thing

[analog obsidian]

Yea it takes around 5 minutes

Oh yep this too

[shut crane]

okay, i dont understand what its doing but ill wait, though its already been 8 minutes

[brave swallow]

btw I got applio trained model(with my voice) in tensorboard in total/g/loss
reached 30 in the chart and it's not continue learning anymore what shall I do ?

[analog obsidian]

Uhmm is not supposed to be this slow

[shut crane]

maybe its because im just using a laptop and not a pc

[analog obsidian]

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

Use this guide

[analog obsidian]

Is your training finished? Pick lowest point in the graph before overtraining

[brave swallow]

I assume it reached the lowest yes because I have straight line in the bottom at 30

and yeah finished

no overtraining detected

[analog obsidian]

Now you have to hear the model to confirm youre using a good epoch

[brave swallow]

epoch = step
yes I do have one of the best epochs between them

[shut crane]

step 3 asks if im training models, but if im only trying to get the voice model that was already made im not training anything right'?

oh wait i get it now

[rare goblet]

Ayo? @shut crane level 2 !!!

[shut crane]

step 3 of inference says: Then select it in the dropdown below.

what am i supposed to do

nvm sorry

holy guacamole it did it

[analog obsidian]

Xd

[shut crane]

TYSM guys

[analog obsidian]

[shut crane]

this is insane

i got goku black singing let it go

[analog obsidian]

Yeah u choose epoch based in their step number (lowest point in graph before raising infinite)

[brave swallow]

oh yeah I know this, my question was why it stopped training

Is there any chance it hits more lower points If I do 1000 epoch ?

[analog obsidian]

But many minutes your dataset has?

[brave swallow]

the one I tested was 10 minutes

I have up to 50H of my voice

[analog obsidian]

Try batch size 8 for 10 minutes

[brave swallow]

not all clear but I could work on it

[brave swallow]

I can do 20 batch size in my gpu

I got 12 Gb ram

[analog obsidian]

You dont choose batch size based in gpu vram

You choose it based in minutes

Batch size means how much details rvc will try to learn

Lower datasets may require low batch size so rvc can properly learn them

[brave swallow]

I see.... , the more batch size leads to lower quality?

[analog obsidian]

The less it learns

[brave swallow]

got it

[analog obsidian]

For 10 minutes, batch size 8 is the most stable

[brave swallow]

I'll give it a try now then

[rare goblet]

Ayo? @brave swallow level 3 !!!

[brave swallow]

and how many epochs?

already done about 150

[analog obsidian]

Try 500

[brave swallow]

ok

let's go

[gusty kestrel]

@brittle wing, I have found 3 results that match your search!

nicsi_

Mutahar (SomeOrdinaryGamers)

• Uploaded: Unknown
• Server: AI Hub
• Likes: 0
• Lang: EN

Epochs

300

Technology

RVC

Algorithm

Rmvpe

Requested by fuckkissah · Jun 16, 2024, 4:03 PM

[harsh robin]

when downloading the model i get

JSONDecodeError                           Traceback (most recent call last)

<ipython-input-8-1d7b23896ae4> in <cell line: 31>()
     31 if os.path.exists(config_path):
     32     # File exists, proceed with creation of creds and client
---> 33     creds = Credentials.from_service_account_file(config_path, scopes=scope)
     34     client = gspread.authorize(creds)
     35 else:

5 frames

/usr/lib/python3.10/json/decoder.py in raw_decode(self, s, idx)
    353             obj, end = self.scan_once(s, idx)
    354         except StopIteration as err:
--> 355             raise JSONDecodeError("Expecting value", s, err.value) from None
    356         return obj, end

JSONDecodeError: Expecting value: line 1 column 1 (char 0)

[azure marsh]

Suggestions for @harsh robin

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

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

[harsh robin]

ty

i just want to put my vocals in and put a model on it

which one is the best for it?

@brittle wing

ty

[rare goblet]

Ayo? @harsh robin level 1 !!!

[pliant lance]

why does rvc take so long and how do i prevent the voices from crackling?

[proper shale]

If you have an AMD GPU, make sure to convert downloaded models to ONNX. Also, if this crackling is happening outside of Okada, I recommend installing VAC Lite instead of VB Audio Cable.

https://rentry.co/VoiceChangerGuide#uploading-models-amdintel-gpus-and-fixing-its-laggy-issue - Converting to ONNX

There's also the Virtual Audio Cable section as well, where it has the VAC lite download link

[brave swallow]

which one we must pay attention to ? value or smoothed ?

[hardy wasp]

hey

i got an error in

colab

[analog obsidian]

value

[brave swallow]

if its like that then means I really hit low point here

but I couldn't catch the path of it LOL

[analog obsidian]

graph is not synced

thats why it doesnt match the epochs

[brave swallow]

hmm

wait

is it something i must enable ?

this ?

[analog obsidian]

yes

epoch value should be the same as the graph values

[brave swallow]

isn't this the reason ?

[analog obsidian]

otherwise they're desync and... is not good overall

[analog obsidian]

you have to tick save only latest, save every weights, and sync graph on applio

if sync graph isnt working you have to manually sync the graph

and that can be only done before training, not after

[brave swallow]

I'm trying to understand

okay

[analog obsidian]

also save freq is too high

[brave swallow]

do NOW I'll have only the latest but how I make sure I saved the lowest one

[analog obsidian]

the best is 1 but if you want more space you can set it to 15-20

[brave swallow]

no issue at all

I'll go 5 or 10

got freee 100 gb

disk

[analog obsidian]

yea dw, keep in mind is not going to be as accurate as save 1

[brave swallow]

yes u mean for sync

ok I'll let it be one

[analog obsidian]

after enabling sync graph on applio, wait for first epoch, it should match the first step of tensorboard

if it doesnt match then

https://github.com/IAHispano/Applio/issues/405

thats how you sync graphs in rvc

[brave swallow]

let me read it

❤️

[analog obsidian]

in applio's case is
logs folder > model folder > delete eval folder, delete events file (events.out.tfevents), delete G and D

[brave swallow]

so its like reset

[analog obsidian]

then you have to open config.json inside model folder and search for "log_interval" set that value to the same value u got in your first epoch

yep

[brave swallow]

ok I'll do it step by step

[analog obsidian]

after this go to
rvc > configs > v2 > 32000.txt (in case your training in 32k) and find log_interval and to the same, set it to first step of your epoch

if you're training in 40k instead u go to v1 > 40000.txt

and do the same

after that you have to go training tab in applio, make sure everything is the same before training
DONT PREPROCESS OR PITCH EXTRACT AGAIN
USE SAME BATCH SIZE, EVERYTHING HAS TO BE SAME, WITHOUT THE PREPROCESS AND PITCH EXTRACT PART

then you can train and its gonna be sync correctly

[brave swallow]

hmmm

[brave swallow]

I think I missed this part

😄

[analog obsidian]

first try if sync graph applio option is working

its supposed to do all of that automatically xD

but for me it didnt worked

[brave swallow]

so basically

this will reset ?

or how to notice it

[analog obsidian]

you'll have to train it again

that graph is desync and u cant fix it anymore

[brave swallow]

oh ok

[analog obsidian]

is confusing the first time u do manual sync but after that is really easy

[brave swallow]

yeah I'll learn it

acorrding to this

mine is 470

[analog obsidian]

you have to see the value of first epoch

and you can only do that if you set save freq to 1

not 10

[brave swallow]

😄

ok let me try again

I'll delete those

[hardy wasp]

await disconnected error

how can i fix

[analog obsidian]

you're using colab?

[hardy wasp]

yeah

[analog obsidian]

which one?

[hardy wasp]

first one in command

[analog obsidian]

applio? colab getting disconnected is usually bad internet connection between you and the session

[hardy wasp]

yeah apollio

need i to get colab pro

its dont convert

[analog obsidian]

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

use this guide

[hardy wasp]

ok thanks

[brave swallow]

is it correct ?

[analog obsidian]

you sync your graph?

[brave swallow]

the option enabled

[analog obsidian]

see your first epoch value

[brave swallow]

0

[analog obsidian]

should match first tensorboard value

after 0

xD

[brave swallow]

24 step

[analog obsidian]

ok if your first epoch is 24, and your first value in tensorboard is 24, then you're fine

[analog obsidian]

first step in tensorboard should say 24

[brave swallow]

in tensorboard 0 then 24

[analog obsidian]

ive never seen a graph with really high values o.O usually is 30/40

[brave swallow]

😄

[rare goblet]

Ayo? @brave swallow level 4 !!!

[brave swallow]

u will find every strange thing in my pc

I could land u some of my luck if u want

LOL

[analog obsidian]

idk tbh, but if your epochs matches your steps in the tensorboard then... its fine

[brave swallow]

yeah I think it does

[analog obsidian]

other helper may know why your values are super high but im lost

[brave swallow]

man thanks for the help u legend

[analog obsidian]

no prob

[brave swallow]

went down so fast we back to 30

[analog obsidian]

did you cleaned your dataset properly?
denoising it, de click, etc

removing very long silences

[brave swallow]

yes with audactiy

[analog obsidian]

okay

[brave swallow]

• is there any other recommendation for background removal

because I have 50H needs to be cleaned

[analog obsidian]

izotope rx spectral denoise

[brave swallow]

this include human noises ?

[analog obsidian]

spectral denoise is a manual denoising, you have to select your noise profile and clean the audio yourself with this noise profile*

but i think u should use #📑│making-models

[brave swallow]

I see , I don't have consistent noise I got variant because its audio

[analog obsidian]

this channel is for rvc help related issues, you're asking things regarding making models so the other channel is better for you

[brave swallow]

sure

[analog obsidian]

uvr denoise works automatically tho it less stable than spectral denoise
(uvr denoise usually mutes the voice because it gets confused)
x minus pro denoiser is a bit better than uvr denoise

[brave swallow]

I'll try both ,I don't think UVR works with arabic

[analog obsidian]

uvr works with any type of audio

[soft tiger]

where can i get applio]

[rare goblet]

Ayo? @soft tiger level 7 !!!

[soft tiger]

local

or what ever software i need for me to use one of my models

and get a vocal audio file and make my model sound like the sudio file

what ever is out there

as well i do have amd so if thats something diffrent i need lmk

which one

yo

ill download and ill lyk when its done

[odd shale]

Se que sonaré como disco rayado pero mija, si te gusta ayudar tanto aplica para helper

(bait)

[brave swallow]

wE aRe HeRe WaTcHiNg SiLeNtLy

[odd shale]

[brave swallow]

turtle is tts ?

[odd shale]

what

tortoga tts

himiko toga tts.

[brave swallow]

does it support AR ?

ola

si

[odd shale]

que

[brave swallow]

ʍouʞ ʇ,uop I ƃuᴉppᴉʞ ʇsnɾ

[brittle wing]

https://youtu.be/Ua1mKqDzym4?si=_LS_YmBz_-cOXZCJ

would i be able to make a model w this

or is his voice too deep will it not work

[odd shale]

Nah, of course you can a model off it

If you know how to clean the audio

[brave swallow]

Eu também posso escrever na mesma língua

mesmo que eu não saiba disso

[brittle wing]

its mostly speech so it wont be too hard to clean it

[odd shale]

Yep, i guess so.

[gusty kestrel]

An error occurred

Sorry, I could not find models that match your search "smokingskul"

[spiral osprey]

pls help-me

[brittle wing]

bandicam.com

[spiral osprey]

'-' so sorry for this.

[brave swallow]

console log ?

[brittle wing]

can u make models w clips w autotune on them

[spiral osprey]

I don't know where the console log opens

[brave swallow]

F12

[spiral osprey]

I wanted to know how to use this voice model, I spoke to the owner of the audio and he said that the voice model works normally for him.

[brave swallow]

did u restart the project?

[spiral osprey]

yes.

[spiral osprey]

i try more than 3 times.

[brave swallow]

is there a cmd running

does it shows some text

[spiral osprey]

I'll try to look at the console now to see.

[soft tiger]

@brittle wing

[spiral osprey]

For me it doesn't work.

[spiral osprey]

how to open the console via webui?

[soft tiger]

anyone?

it hasent moved in the past 5 minutes

i think its done chat

ahh i see

bet

how do i upload voices?

[brave swallow]

in the browser I mean

[brave swallow]

accidentally I deleted the eval , how can I know which pth had the best quality

👀

[proven hill]

#📰│dev-updates message

[brave swallow]

does it run using rtx 4070?

[proven hill]

its online, you dont need to load anything online or local.
Its a web page that runs on an A100 80GB

[brave swallow]

oh I see, nice, and we able to do custom voices right?

[proven hill]

yes

[brave swallow]

good job

[brave swallow]

I'm just wondering how much hosting this cost

GPU expensive

[proven hill]

dont worry about that

[brave swallow]

hahaha okay

[brave swallow]

[lilac ridge]

tha't is?

[brittle cypress]

need help doesn't work on discord

[rare goblet]

Ayo? @brittle cypress level 3 !!!

[ionic yew]

i tried the model fusion on ilaria rvc but its not working

okay even uploading the index and pth files don't work either

[lost jay]

how come my skibidi toilet voice doesnt workj anymore

[proven hill]

hi!

whats the issue?

[proven hill]

uh, it was bounded to be improved but ill take a look now

if you need it not for a test, i can merge them for you meanwhile

also: are you sure both are 40k?

[ionic yew]

yes

both are 40k

most of my rvc models are 40k

[proven hill]

do you want me to merge them locally?

[ionic yew]

yeah if you want

[proven hill]

send me the pth in dms

[ionic yew]

send you the pth files

[tiny totem]

Currently training a model & when preprocessing I've noticed I accidentally included a few files with less than 1 second of audio. Are these in cluded with training, or are they thrown out? If it is I at least hope it isn't going to negatively impact the model, if at all?

[tardy lance]

I've tried using hina's and ilaria's rvc and they both give me the same error i think, something like index error -1, I've tried looking in the pinned messages and on google but can't find anything

I've done the same processes as before so idk maybe compatibility error?

[proven hill]

try using Ilaria RVC Zero

https://huggingface.co/spaces/TheStinger/Ilaria_RVC

[tardy lance]

seems to be working

i dont really understand what was wrong before, anyway thx

uh nvm doesnt work anymore baha

i click convert and it says error

[rare goblet]

Ayo? @tardy lance level 1 !!!

[proven hill]

are you sure?

maybe youre doing something wrong

[tardy lance]

nope same things as always

[proven hill]

whats the model?

just tested, works great on me

[tardy lance]

I've tried all the mori calliope models that there are

[tardy lance]

i mean yeah I'm sure it's gotta be something on my end

[proven hill]

gives me an error too with that model

maybe its a model problem

[tardy lance]

uh i see

I'll try with different models tomorrow

tbf i tried the default tts one and it was working

but that means no mori calliope ig 😦

[proven hill]

every sinle calliope model?

[tardy lance]

all the three models on the models website the one with the pride flag and the last one here on discord made 20 days ago

[tiny totem]

I gues s 5th time's the charm but I don't get why this was downloading the other models already in the space than the link i gave it

[proven hill]

thats... strange

[proven hill]

wdym?

[tiny totem]

Can't get it to do it again now but the Space kept downloading other models like Chaeryeron and qwe despite the link

[proven hill]

scroll down

[tiny totem]

Tried a diff link & it happened again ヘ⁠（⁠。⁠□⁠°⁠）⁠ヘ

[proven hill]

thats strange

i restarted the space, can you check now?

[tiny totem]

👍 I think that worked? Downloaded correctly

[proven hill]

it was probably some gradio issue, go figure

important thing is that its resolved

[tough aspen]

um how can I send a picture. having problems with voice.ai

[low shard]

bro left

L voice.ai

[quaint dock]

Hey, I just installed both of them and it couldn't find it

[violet heron]

find what?

[cinder saddle]

hey dontt mean to be a bother but how do I know which version to gett based on my specs?

[quaint dock]

cuda

it says its not available

while it is installed

[keen mica]

Are harsh P S T, all the harsh sounds good for a dataset? I meant how sometimes when you really pronounce letters like P, S, and T some wind goes out of your mouth and it creates harsh sounds for your microphone

[violet heron]

What’s your GPU

[open stag]

HOW TO FIX THIS😭

I mean, when i click the Public URL at colab, it says Gateway time out

[thin stump]

They are all consonant sounds. You should remove some of them if you think RVC will mistake that for clicking noises especially with the t and ch sounds. The harsh P's sounds like a mic thump so those particular voice lines needs to removed from the dataset

That's just my opinion and you should ask in #📑│making-models

we have de-essers for S sounds with rx10 and t de-esser

[austere fulcrum]

how to find a voice we're looking for.
not all the voices come up in the search.
I have to skip my level.

[rare goblet]

Ayo? @austere fulcrum level 1 !!!

[austere fulcrum]

oke

when I'll be level two.

[thin stump]

weights.gg or #1175430844685484042

[lost sail]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

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

[brittle wing]

What's the best way to remove Reverb/Echo on a Vocal?

[craggy brook]

what is this ?

[brittle wing]

What is "test" for

[brave swallow]

after training 500 epoch and hit the lowest point, that epoch sounds 100000% robotic why?

[rare goblet]

Ayo? @brave swallow level 5 !!!

[brittle wing]

hey sorry for asking but how can i make the ai sound more like the model itself and not like some random when im recording it

it sounds very ai-ish and its not what the voice i want

[tame mica]

in sorry but is that the Samsung volume thing lmfao

[fiery raptor]

whats up with uvr5

talking about the google colab version

[brave swallow]

[pastel oak]

did you put smoothening to 0,987?

on the left side

and any screenshots of the graphs

[brave swallow]

yes and no,

I hadn't any spikes

so I decrease it to find the lowest point

[fiery raptor]

its been processing the file i gave it for like 10 minutes now

[pastel oak]

maybe you disconnected?

[pastel oak]

Ehh you always have a lowest, even if its harder to see on the graph

[fiery raptor]

from the runtime itself or from the google drive?

[pastel oak]

i think its the "smoothened" one that shows the right number, the one next to Value and Steps

[pastel oak]

one of them not sure

[brave swallow]

not really, when its 0,987
I used to see 30 and 14 at same line

[brave swallow]

oh I see! some people told me it's the value.
I'm lost in between

[pastel oak]

It could be the value i dont know atm ahhaa

[brave swallow]

😂

[pastel oak]

whatever number fits on the Y axis

[brave swallow]

yeah

[low shard]

If you got a good pc (like good gpu) you can use the local forks, else use the online ways

[fiery raptor]

im not disconnected from neither

its just been stuck on this for like 15 minutes now

[fiery raptor]

i tried running it on both

gave me the same results

it gets to the processing file point and then it just loads infinitely

well it processed it

worked

thanks

[urban wasp]

-colab

[azure marsh]

☁️ Google Colabs

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, modded by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], credits to Eddy, Hina and Gdr for translating and fixing Google Colab
• Ilaria RVC, by thestingerx Google Colab
• EasyGUI, by rejects Google Colab
• UVR5 NO UI for Google Colab, by Eddy Google Colab

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

[rare goblet]

Ayo? @urban wasp level 1 !!!

[craggy brook]

I'm trying to open the vrc thing and this warning comes up and it won't open

[brave swallow]

is it okay for a rvc to start from 400?

[tame mica]

how did u get a Windows emu to run on your Samsung fon lmfao
genuine answer: it's just the model steps. you can check your actual Epoch count in your console i think

[brittle wing]

can anyone tell me if i should stop training?

[brave swallow]

oh I see...
this called RDP
remote desktop I use my own pc as VPS

[tame mica]

ah

[civic sail]

can anyone create my script to voice please it be a big help

[brave swallow]

script voice?

[civic sail]

i have a script for a video that i need to convert like dialogues between celebrities

[brave swallow]

did you set it to 0.999 ?

[brave swallow]

maybe you can get better results but not guaranteed, and this is what I'm trying.

in your case you will have good audio

[brave swallow]

applio , eleven labs

[civic sail]

messi and mr beast voice is not avaialable there

[rare goblet]

Ayo? @civic sail level 1 !!!

[brittle wing]

dunno what that means

oh got it.. yes its 0.99

[brave swallow]

smoothing slider

[brave swallow]

then you reached your limit I think

[brittle wing]

so at what epoch is it best quality?

step 8k?

[brave swallow]

I can't tell exactly, but it's close to 8500

[brittle wing]

ok thx

[brave swallow]

welcome let me know if that works

[proper shale]

I'd say keep going, also check kl and mel

[brave swallow]

could you tell us why, I want to understand ur pov so I use it with my model too

share experience please

[alpine lotus]

look, i'm trying my best ok?

[brittle wing]

its only 16 mins data

[proper shale]

Regardless, if you can keep training that'll make the model better, and overtraining takes a while on most models anyway

[proper shale]

1. G/total is average

[brave swallow]

so?

[proper shale]

You can't just stop training because the average is going up

When the graphs that matter more r still going

KL and Mel are how your model is actually going

[brave swallow]

oh I see

[proper shale]

KL being how different the model is from dataset, Mel being how well it's reproducing clarity compared to the dataset

There's also FM as well but it's kinda glitched for some reason

[brave swallow]

good so don't pay attention to G spikes

[proper shale]

You should still look at g/total, but it isn't your main priority imo

Helps you see what's going on in the generator vs discriminator battle

[brave swallow]

got it,

now time to understand the difference between KL and Mel

[brave swallow]

yeah I know this part

[proper shale]

Oh btw, lower = better, for basically all these graphs

[brave swallow]

yeah

where we can find all these documented, because For last month I been learning from other people

i know 0 before 3 weeks about all these tensorboard, rvc.. etc

[proper shale]

The pins in this channel, also #📑│making-models too

[charred lichen]

why all genshin models got deleted-

[proud elbow]

takedown request from the VAs

[quaint dock]

Nvidia 1660 Ti

[charred lichen]

oh :(( are they still available somewhere or i just need to search?

[proud elbow]

internet archive on huggingface/weights.gg ig?

[charred lichen]

thank u!

[proud elbow]

btw just use it personally/non-commercially, also youtube may no longer be safe place to publish with those voices

[charred lichen]

thank u for telling me ^^

[rare goblet]

Ayo? @brave swallow level 6 !!!

[civic sail]

Hey i would like to create voices like messi and mr beast to use in videos can anyone guide me how i can create those

You help would be appreciated

[brave swallow]

you can,
you need
TTS
RVC

both of them exist in applio

use applio, train the datasets of Messi voice, then use it

[proper shale]

https://docs.aihub.wtf/ - has everything you need to know

[gusty kestrel]

@brittle wing, I have found 1 results that match your search!

adicop

Glep - Smiling Friends (RIN_E3)

• Uploaded: <t:1708985901:d>
• Server: AI HUB
• Likes: 0
• Lang: Unknown

Epochs

200

Technology

RVC

Algorithm

Rmvpe

Requested by fuckkissah · Jun 17, 2024, 2:27 PM

@brittle wing, I have found 22 results that match your search!

lepifort

Drake

• Uploaded: <t:1692489600:d>
• Server: Unknown
• Likes: 0
• Lang: Unknown

Epochs

1000

Technology

RVC

Algorithm

Rvmpe

Requested by fuckkissah · Jun 17, 2024, 2:31 PM

[proper shale]

Does Command Prompt say anything or does the GUI window just freeze?

That is quite strange - what version of the changer are you running?

Uhhhh, you can't really share datasets here

[brittle wing]

thats not the dataset

[proper shale]

But I suppose you can pitch down the vocals you wanna convert later and it'll work fine

[brittle wing]

its just audio files im gona use

[proper shale]

A 44 megabyte mp3 tho 😭

[brittle wing]

[proper shale]

Imma delete just for the sake of safety really

Will work fine I guess

[brittle wing]

im trying to delete it and its not deleting

did u delete it already and its glitched on my discord

[rare goblet]

Ayo? @brittle wing level 1 !!!

[proper shale]

Yea I got it deleted alr

DirectML or CUDA?

Hope that's not a GPU related issue

Quite strange. I'd probably use version 17b then, to see if it's a new version issue

[brave swallow]

Rvc training can become faster if there's more vram or GPU cpu?

[proper shale]

If it's your GPU yeah you should use CUDA

If you have an AMD card, DirectML

[proper shale]

CUDA cores are what matter, more of em means it's faster to process things

Ofc VRAM also matters as well, because batch size

Possibly...

[brave swallow]

is it cheaper/faster to train on cloud ? I have rtx 4070 and Use applio

[proper shale]

Anything before that?

[proper shale]

Yeah no, Local ftw

[proud elbow]

your 4070 is still faster than colab T4 (around 3050/3060 performance but 16 gb vram)

[coral glade]

when i open the app and setup all my settings, it just doesnt work???

what i mean is when i speak i hear my normal self and not the ai

[pseudo shore]

it says syntax error when downloading pytorch for the voice changer-- why?

[zealous zodiac]

Where do I download this?

[violet heron]

is passthru on?

[violet heron]

Download what?

[zealous zodiac]

Nvm I found it

[coral glade]

no ill enable it

[rare goblet]

Ayo? @coral glade level 1 !!!

[violet heron]

Don’t

It should be off

green not red

[coral glade]

should start be green or stop?

[lilac dawn]

are the servers down?

theres an error on google collab

[coral glade]

idk its still not working?

[violet heron]

Send a screenshot of your settings

[coral glade]

ok

[violet heron]

What colab are you using?

[lilac dawn]

uhhh

[rare goblet]

Ayo? @lilac dawn level 1 !!!

[lilac dawn]

iked

the one with the discord link

[violet heron]

Send a link

[lilac dawn]

ig

[brittle wing]

hi how do i train a ai voice model

[violet heron]

-guides

[azure marsh]

Suggestions for @brittle wing

Documentation

AI HUB Docs

https://docs.aihub.wtf/

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• Full AI Voice Model Training Guide (local) by Christopher Villanueva
• Model training with Mainline RVC by Raven (ravencutie21)
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[brittle wing]

thanks

do u got a video

[lilac dawn]

https://colab.research.google.com/drive/1-TnZWYnea9YvXnLocXwpo5oqSd3oCAbd#scrollTo=7vh6vphDwO0b

[violet heron]

Outdated

Are you trying to train or make covers?

[lilac dawn]

aigh

i wanna make covers

[violet heron]

Scroll down

[coral glade]

ok

[violet heron]

What’s your GPU?

[coral glade]

idk

my graphics?

[violet heron]

Yes

[coral glade]

ok it sucks but nvidia quadro k600 💀💀 (oof)

[violet heron]

Use the google colab version instead

[coral glade]

wheres that?

[rare goblet]

Ayo? @coral glade level 2 !!!

[violet heron]

https://colab.research.google.com/github/hinabl/voice-changer-colab/blob/master/Hina_Modified_Realtime_Voice_Changer_on_Colab.ipynb

[coral glade]

thx

[violet heron]

Remember to read since you might have to do some stuff

[coral glade]

ok

[brittle wing]

for some reason it doesn't let me send images

[vernal radish]

getting "Invalid username or password." on https://applio.org/models/download/1230191826800214108

[proper shale]

This is.... strange - I've never seen this issue before.

Probably use https://rentry.co/RVCRealtimeGuide then, might work out better

[proper shale]

Model has probably been privated, try to find other ones

[vernal radish]

Ok

do yk if the nvidia quadro p5000 supports model making

?

[brave swallow]

I see

[brave swallow]

hmm

[brave swallow]

I see

but in case I use 5 accounts at once 💀

won't this Become faster?

and accelerate

it seems more like crypto mining, just training randomly to hit lowest point

I hope I'm wrong

[proper shale]

I'd say probably, since it has CUDA cores and a good amount of VRAM

[proper shale]

Yeah use RVC Realtime then. Very strange issue

[rare goblet]

Ayo? @brittle wing level 2 !!!

[cinder saddle]

what does that mean