🧬│ai-chat

[gray rover]

my head's gonna pop if I try to tinker 44 more

[chilly lake]

i'm more surprised by mrf hifigan with 44k clearing the top is no time at all

I have a feeling this is because the upscaler does not align to the top edge, so no mirroring there

7x7 = 4.9khz lines

[gray rover]

overal I'm a lil skeptical with 44.1

as much as how unconvenient working with 48 is.. it seems less brain devastating than 44

So if you manage to utilize it well ( n get better perf than 48 ), Imma congrat you with 200% honesty

[gray rover]

as for that, let's hope it ain't the case

[chilly lake]

i'll leave it running and check tomorrow

[carmine temple]

Hello 👋👋

[icy ocean]

hola gente una consulta

soy peruano y quisiera saber si hay algun emulador de voz q de miedo osea una voz asi como para hacer cosas turbias y amenazar

[grim terrace]

If AI uses samples to identify speech patterns and adjust accordingly, does that mean the human voice can do that too

[limber ember]

Thank you for accepting me into your group.. My name is Unyime Johnpaul from Nigeria.

[gray rover]

no, it uses convolutions, feature maps and various feedbacks from discriminators and not strictly samples

simply saying, no lul, a human voice can't do that as we ain't machines supported by raw math

[grim terrace]

oh ok

[gray rover]

thoooo if you mean legit identifying speech patterns and adjusting in a human sense, yea, ofc
brain is capable of that

But that's for really any repeated and learned tasks

[twin schooner]

Hi you ai? Im not ai

[cyan crag]

Hi everyone! I'm so interested in AI songs and can't wait exploring more! but I'm new for discord, so I'll be so grateful if anyone can help me with it!

[solar torrent]

I'm not a chatbot. I just like to talk.

[paper blaze]

I am certainly not a chatbot powered by Hermes-3-Llama-3.1-8B-Q4_K_M language model

[covert lake]

Imagine being a gguf quantized model

[paper blaze]

https://media.discordapp.net/stickers/1302564223242473522.gif?size=160&name=fingerfromhitshowbreakingbad

[solar torrent]

https://tenor.com/view/nuh-uh-beocord-no-lol-gif-24435520

[fallow plinth]

hi guys

[honest atlas]

hi

[blazing birch]

一个利用AI使生活变得更高效的人

[covert lake]

Yuh uh

[covert lake]

Hello

[covert lake]

Wsg

[solar torrent]

Please speak English in #🧬│ai-chat, or speak Chinese at a Chinese chat.

[errant sphinx]

https://tenor.com/view/gently-open-the-door-robot-leaves-me-no-choice-gif-14368586

[dense pivot]

Que tenia que ver el ser peruano xd

[icy ocean]

Jajajaaj lo que pasa es que hay varios que hablan ingles y otros idiomas

Tu sabes cómo poner otras voces?

[stiff orchid]

Greetings, I'm here to catch up a bit on AI voice matters. Been quite a while since I even glanced at it.

[covert lake]

@dense pivot @icy ocean talk english here, or use #🌍│español

[covert lake]

RVC v2 didn't change much in quality:

• Applio (fork, which means modified version) made it sligthly faster
• And there are community pretrains (pretrains models are models used as a base for training, which is what there was already, just now there are other ones that help mostly for other languages),
• RVC v3 is never gonna get released, maybe an unofficial fork by codename to help it but it's still in the works

For wokada (using rvc models in realtime for calls), there has been way better performance with the deiteris fork

[stiff orchid]

Hmmm... The only real uses I can think of atm would be for TTS/"Jarvis" type things, but I could definitely train some voices I have in mind.

[covert lake]

RVC isn't made for TTS (Text To Speech)
RVC is STS (Speech To Speech)

Ofcourse you can make a TTS audio with any TTS, and then use it as an input in RVC tho (Just like how Applio does this built in with MS Edge TTS, which is good for multilingual and quality but needs internet and doesn't have emotions)

[stiff orchid]

Thanks, I recall using RVCv2 a little way back.. No worries about dumbing things down for me too much, I'm out of touch on audio but I like to think I have a decent grasp of AI/ML in general.

[covert lake]

Idk what's "Jarvis" type of things tho

[stiff orchid]

Ahh, speech to speech. there we go, yeah, it's been ages like I said. Bogged down in the SD world too long.

[stiff orchid]

real-time chatbot/assistant kinda thing, but I want to catch up in general.

I do recall changing a few songs to use different voices, amusing.

[covert lake]

If you were looking for TTS, you can technically still do it with RVC tho

But if you're looking for a TTS program, there's 2 types, i will say the best ones:

• few-shots: GPT-SoVITS
• one-shot: F5 tts, and Fish Speech

[covert lake]

Btw Flux has better quality than SD
-# But is also more resource demanding

[stiff orchid]

Oh, excellent! Thanks, you saved me a good hour on github 😆

[stiff orchid]

https://tenor.com/view/i-know-star-wars-han-solo-gif-14764428

(check my profile)

[covert lake]

I did an index if you want

Also a noticeable good one-shot tts is XTTS2, but it's never going to get updated unfortunately

[stiff orchid]

@chilly lake I see they'll let anybody in here...

[covert lake]

I do remember there was a fork called AllTalk tts but never tested it ngl

[covert lake]

Idk what u mean by that

But Noobies is really smart, he also did play alot with TTS

[covert lake]

oh lol

[stiff orchid]

He's on my server too.

[covert lake]

-# btw embed perms at level 5

[covert lake]

Ohh 😭

[paper blaze]

no piper tts

[covert lake]

I don't remember hearing about that

[stiff orchid]

Yeah the last time I was looking at voice or TTS was at least 6mo ago, as I recall it was a bit in flux at the time so I've got catching up to do. I'm so addicted to the voice of Twilight Sparkle that I just can't stay away any longer. 😊

pretty sure I tried piper

[paper blaze]

tl;dr medium quality, very light and fast on inference

can be used even on raspberry pi

[stiff orchid]

ah yes "back in the day" of October 2023

[covert lake]

this ?

eh

@chilly lake what do u think about piper tts?

idk if to add it to the index

[paper blaze]

yea

[chilly lake]

havent tried.. 22050 does not look attractive to me

I tried it with a streaming voice dialogue AI pipeline

[stiff orchid]

Just after something relatively lightweight and easy to use that I can just pipe things to

[chilly lake]

listen to voice - asr using whisper - microsoft's llm - piper - play response from ai, loop

it was lightweight enough for close to realtime

[covert lake]

how was the quality tho?

[stiff orchid]

whisper is a bit heavy duty, there's something vlad and I use on a side project that's... some google thing, it does speech recognition quite well. I'll have to check my convo about it...

[paper blaze]

not sota for sure, you can see some examples here https://rhasspy.github.io/piper-samples/

[chilly lake]

yeah, whisper does mistakes, and now there are a lot of complaints from the medical community that use it to trascribe doctor's notes and it writes random garbage

[covert lake]

@atomic vector explain the meaning of 'sota'

[atomic vector]

hey nick088, 'sota' stands for 'state of the art', which basically means the best or most advanced in a particular field or technology. in the context of your conversation, alexeyartemiev is saying that the examples on the rhasspy github page aren't the most advanced or the best in terms of text-to-speech systems.
-# AI-generated responses may be inaccurate; please verify important information.

[covert lake]

oh

[chilly lake]

I dont remember, it was not expressive, it was just tokens to audio thing, as I recall

[toxic frigate]

What is this chat made for everyone

is there any explainaiton

while ago i joined here so just wanted to ask quickly

[covert lake]

low quality (1st) vs medium

[paper blaze]

i've been using it for twitch chat tts for quite some time, it's not as good in all fields related to quality of a tts as many alternatives, like speech clarity. i think it should be considered to use mainly if you have hardware constraints or low-end PC

[covert lake]

yea.. i think it's good only in being fast & lightweight

[chilly lake]

buhhh

[covert lake]

yes there isn't a level requirement anymore

[chilly lake]

fast yes, quality = shit

[toxic frigate]

No i mean whats the purpose of the server

[covert lake]

do u think it's even worth it to be added in my index? idk if to do that tbh

how many people are there that need only realtime fast but lowest quality ever

[covert lake]

AI,
Mostly RVC & Wokada,
But there are also other #1159513888199540817 about other AIs, like mine for FaceFusion, Termux & SuperPrompt

[chilly lake]

there are other very fast tts that use eSpeak

[covert lake]

eSpeak?

[chilly lake]

the most common phoneme to audio engine

[covert lake]

oh lol

thx

[covert lake]

Yea ig i'm not gonna add it

u can't say 22k is high quality 😭

[chilly lake]

regular speech does not require more, but it just does not sound good

it looks like the generator is undertrained

[stiff orchid]

"If it's good enough for a landline rotary phone it's good enough for me!"

[chilly lake]

at least th medium one with the same mirroring lines we saw with rvc

[covert lake]

-# in case it wasn't clear it used espeak

[covert lake]

[covert lake]

on low-end pc u could just use Edge TTS on cloud atp

[chilly lake]

run my test thru it

[covert lake]

I forgot the requirements for Fish Speech or F5

maybe F5 is lighter

[covert lake]

?

[chilly lake]

crap, one sec,need to fix that meow

weird quotes

eSpeak shits the bed on combined words and 'necromancers' 🙂

[stiff orchid]

I created a custom context menu item so I can almost instantly create edge-tts speech. Just gotta type in the text and the filename and it's done in a second or two.

[covert lake]

Yea

Piper looks to me a local & worse version to edge tts
Good only and specifically in speed

[stiff orchid]

It's a shame edge only had the fixed voices it had, likely censored too, can't say I ever tried...

[covert lake]

this is medium quality italian

looks like an english speaker tryna learn italian for the first time that talks too fast

[stiff orchid]

oh wait, it's NOT censored.

[covert lake]

it can say curse words

[stiff orchid]

I just assumed it was, did a test and there's not much of a limit there

I use that creepy victorian english sounding girl voice. everybody hates it

[chilly lake]

hello mirroring

[stiff orchid]

Pretty sure no english speaker ever did THAT

[south nova]

hey, are there any performance changes from w-okada 1 -> 2?

or just support for beatrice 2, and some added settings

[covert lake]

I mean it sounds bad tho

[covert lake]

there isn't even any high quality voices

with high i mean 22,5Khz

[carmine wadi]

hi

[covert lake]

all voices are mosty low & medium

some are even x_low which is even lower

[carmine wadi]

what are we talking about today?

[covert lake]

TTS

[carmine wadi]

TTS ?

what does it stand for?

[covert lake]

Text To Speech

[carmine wadi]

ok

[covert lake]

we were testing Piper TTS, it's fast and very lightweight
but shit quality lol

[carmine wadi]

i know some thing about CPT

[carmine wadi]

kool

[covert lake]

actually found just one

eh

[carmine wadi]

...

i did'nt understand anything

[chilly lake]

yeah, does not sound like a good expressive read

[stiff orchid]

As a monolingual english speaker, it definitely sounds like Italian, but I'll have to draw the line there.

[chilly lake]

just a robot that does not breath somehow making the sounds and not running out of air

[stiff orchid]

oh please tell me there's a way to do Latin, that could be fun

In nomine tenebrarum, evoco te, daemon potentissime.
Ex silentio aeterno advenias, ex abyssum ignis surgas!
Da mihi auxilium tuum, da mihi potentiam tuam.
Praesentiam tuam hic desidero, o princeps umbrarum.

TTS demon summoning

[queen kernel]

what is this ??

[quartz roost]

Fastest growing ai voice

[queen kernel]

name ??

[quartz roost]

...

Really?

[queen kernel]

idk what it is sorry

[covert lake]

I'm italian, I know italian & english

[stiff orchid]

I kinda figured 🤔

[elder willow]

what tool are you guy using for covers? forgot the one i used before resetting my pc, i didnt note the link

[covert lake]

It does sound italian but italians don't speak as fast as english people do

and the model sounds robotic lol

[covert lake]

For AI Covers, there's RVC

What's your PC GPU?

[stiff orchid]

Weird, I had the impression that Italian was faster than english.

[queen kernel]

kasane teto??

what it can do??

[stiff orchid]

But then again, where you were raised greatly impacts the speed of your speech

[queen kernel]

is it free @quartz roost

[quartz roost]

https://youtu.be/fNZE46C5V78?si=ynIKWfPgJfSs_0fs

There's a free version

[covert lake]

Unless you mean dialects,

I don't think so

[quartz roost]

And a paid professional version

[queen kernel]

is it the thing that convert midi into voice ??

[quartz roost]

She can sing anything

[molten river]

fr

[queen kernel]

oh crazy.... i saw this thing in Ace Studio

[molten river]

fr

[queen kernel]

how do i download it ??

[quartz roost]

Ace studio sucks

[queen kernel]

it is paid

[quartz roost]

Most of the LEGAL TO USE ai voices are paid

You know, consent and all that shit

[queen kernel]

how do i download that teto thing ??

[glass shore]

hey

[chilly lake]

you mean likeness rights

[quartz roost]

You need a license to use someone's voice legally. A written agreement.

This has been established for decades

99% of the voices here are illegal

[worthy coyote]

what is going on in south korea 💀 💀 💀

[quartz roost]

Martial law

[chilly lake]

yes, the transfer of likeness rights

you can use my voice in exchange for x

[quartz roost]

There's no flexibility on this. Either you get a contract or you can't use the voice.

There are some free voices to use with terms of use posted.

[chilly lake]

voice, photographs, video, etc

[quartz roost]

Often the voices are given different names, and you must use that identity instead of the voice actors real name.

[glass shore]

does anyone here know how to do videos

[proper relic]

KOREA WAR

[wise quiver]

😊

[icy ocean]

#🌍│español ayuda

[elder willow]

yo whats up lol

[heady adder]

wattt sup mann

[elder willow]

how is it going

[heady adder]

bad

[elder willow]

oh

why?

[heady adder]

ım shoudl learnıng science ın netherlands

[elder willow]

oh, im good cuz free from school early

starting december and now i finished school

[heady adder]

oh yeahhh

[heady adder]

Do you know a free site that has a live AI voice changer?

[elder willow]

site? no, but you can use cloud gpu as you probably dont have a good pc, give me a sec

[rare sorrel]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[elder willow]

choose the second one @heady adder, i personally recommend it

[heady adder]

that say'You didn't start this interaction. Use the command for interacting with it.'

[elder willow]

https://colab.research.google.com/github/deiteris/voice-changer/blob/master-custom/Colab_RealtimeVoiceChanger.ipynb the colab, https://rentry.co/ForkVoiceChangerGuide the guide

ask in https://discord.com/channels/1159260121998827560/1159290161683767298 btw

[heady adder]

u are so nıce mannn

[elder willow]

ty :p

[heady adder]

now ı can troll my froendss

[elder willow]

yea

[elder willow]

what

[arctic tinsel]

some idiot parked an armored vehicle at the intersection.

[elder willow]

lmao it got deleted

[night lake]

Sounds like an amazing time

[somber kindle]

by me

poopyhead

[elder willow]

ok and dont call me a poopyhead

[somber kindle]

nuhuh

poopybutt

big fat stinkface

[arctic tinsel]

Not really. My McDonald's delivery was late.

[elder willow]

i dont want to have a real beef with someone, stop it lmao

[night lake]

What! Ok time to roll up and show him you stand on business

[arctic tinsel]

[somber kindle]

if you take me seriously and start a beef with me over me calling you a "big fat stinkface" then maybe you need to deal with your own problems before going on discord

[night lake]

Does Korean McDonald's taste like cardboard too or is it actually edible

[elder willow]

i hope you're pranking by calling me those nicknames, simple as that

[somber kindle]

i was

horsin around 😎

[elder willow]

[arctic tinsel]

what do you expect? McDonald's is the same no matter which country you go to.

[night lake]

Better tasting food 😭

[arctic tinsel]

[night lake]

Ik Japan has some fire meat, China is China and idk what Korea has

[arctic tinsel]

bulgogi burger

[night lake]

What

[arctic tinsel]

korean bbq burger.

but it doesn't taste like bbq

[night lake]

It looks.... Like....

A burger

Is the sauce what makes it special?

[night lake]

Ew why is it brown

That sauce looks disgusting

[arctic tinsel]

sweet soy sauce

[night lake]

Huh

[arctic tinsel]

well.. talking about the taste of burgers during the first martial law in 40 years... Fantastic.

[somber kindle]

i hate albino people

looking like a fucking mayonnaise jar

[night lake]

Yeah, so what's your favorite burger

[somber kindle]

burgaer

[night lake]

What do you think of the baconator

Do you have Wendy's?

[minor blade]

🐢

[arctic tinsel]

pop eyes

[minor blade]

I love Taco Bell's crunchywraps.

[somber kindle]

Yeah if you mean Wendy's nuts hit your face

[minor blade]

🐢

[somber kindle]

Yummer

[arctic tinsel]

[somber kindle]

pig

[timber quartz]

doeso someone here know how to create PTH files for rmc voice changer

[covert lake]

wdym?

[covert lake]

What's your pc gpu?

[covert lake]

It's better to check what's ur pc gpu first than doing it cloud, cloud has time limit

[minor pendant]

is there a better quality realtime voice thing than using google colab and the trial gpu

using the cpu option after the trial
is up runs like garbage

[covert lake]

You need to train a model with RVC

What's your pc gpu and what are you using right now?

Be sure to NOT follow yt tuts

[covert lake]

quality depends on the voice model

Maybe you meant delay

[covert lake]

Did you check what's ur pc gpu first?

[minor pendant]

yes it wont let me use anything but the ones it provdes

i have a good pc that shouldn't be an issue

i have a 4070 super

[minor pendant]

yeah not the model itself but any other reliable ways than google colab, so i dont gotta keeep running the trial

[covert lake]

Yep that's really good

-rt

[rare sorrel]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[covert lake]

1st guide

its the Wokada (program that uses RVC, Speech To Speech, models in realtime for calls) Deiteris Fork (fork means modified version), this one has the best performance

[covert lake]

Yea you shouldn't use Google Colab when you got a good pc

Cloud services are only when you got a bad one

[minor pendant]

i just got this pc and thats what i used before

so if theres something way better then im open for ideas

[covert lake]

Wokada, especially the deiteris fork, is the best program

[covert lake]

Yea, cloud services (that run remote code on good pc) (like colab and kaggle) are only for people who don't got a good pc

It's really not worth it to use those when you got a good pc, as you would be limited by only the around 4 hours of GPU of colab that aren't even granted

[minor pendant]

it gave me an hour every 12 hours

it was bad

and i couldnt run it and play a game at the same time

[covert lake]

yea the gpu time is random for the free tier

you can get disconnected in 2 hours a day, the other day 4 hours, the other day 1 hour and an half

There is kaggle which is harder and requires phone number but gives 30 hours weekly

But why would u use that when you can use ur own pc lol

[minor pendant]

yeah thats fair

funny it told me to extract the files, i did, now it wants me to extract files from the extracted filees folder

[obtuse spire]

google colab doesnt work anymore?

[forest crypt]

Why do you say so?

[obtuse spire]

i getting errors each time i try to use it

[forest crypt]

you have already tried to discuss your problem in #✨│ai-help ?

[obtuse spire]

and some people under video tutorial saying that

[obtuse spire]

no

[forest crypt]

N_C_like

you can wait for someone to help you, also, you should send some evidence of the error, of how it happens.

[timber quartz]

Do u know some free options ? Im using RVC project from github with my ryzen 5600x and it work perfectly fine now i just need to learn how to creates my own voices

[west belfry]

hi

[small raft]

hello

[lethal flax]

still wondering

[polar flax]

fr many ppl asking for wokada are using igpu, hopefully it could have better support on last gen igpus & NPUs
https://www.tomshardware.com/pc-components/gpus/discrete-gpu-sales-dip-while-nvidia-continues-to-dominate-igpus-increase-while-discrete-gpus-decline

[solar torrent]

I don't have a CashApp account, I don't have my PayPal account verified, and I don't even live in the US either, so nuh uh.

Please find your love instead of me.

[slender estuary]

What is this for? 💀

I don’t know why a name of a fruit is considered offensive in this server, but bro.

Also, dead chat.

[covert lake]

Which? And are you sure you followed the amd guide of it?

[ionic pumice]

rhymes with 🍇

[covert lake]

The one on github https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/en/README.en.md (aka Mainline) has amd support only for linux

are you on linux or windows?

[solar torrent]

This g word contains an awful r word.

[chilly lake]

the Scunthorpe problem

stupid filtering results in stupid results

[alpine granite]

scrаpе grapе drаpе

[chilly lake]

when you use a profanity filter a lil too aggressive

[alpine granite]

scrаpе is used for ai dataset speaking

rip that gets filtered lol

[covert lake]

#1159290752195633273

[elder willow]

whats a good voice changer like okada

with a better UI

[chilly lake]

there's like one outdated af voice changer, there's wokada, there's somewhat optmized wokada

there's no better UI unless you just go and make your own

[covert lake]

none

wokada is the best

[solar torrent]

It's just W-Okada. Other than this either paid or requires you to sign in.

[covert lake]

https://cdn.discordapp.com/attachments/1159290193619189821/1283652791931506759/sponge-bob-night_.gif

[covert lake]

all my homies hate voice.ai

[meager remnant]

anyone here who is decent with prompting

chatgpt

if yes please hop in voice chat

need quick help

[hybrid temple]

what voice should i troll with on fortnite

[alpine granite]

troll with a drum slop model

[analog cosmos]

should use this model https://discord.com/channels/1159260121998827560/1287181931154509825

[high patrol]

Hi

[analog cosmos]

[high patrol]

How are you doing

[analog cosmos]

I'm good, just a hot busy day wbu?

[lapis raptor]

Guys, my page on weights doesn't quite work, that is, I can watch models there, but I can't use them and insert my audio, I couldn't delete my account, how can I fix it?

[high patrol]

Good thank you

[worthy coyote]

a

where did my ai chatterboxes go

now i'm a normal human being

just like you and me

[limpid bone]

hello

[light girder]

waiittttt

didn't one of the dev said they're experimenting v3 ?

[polar flax]

shouldn't call it v3 without approval of RVC-boss

[covert lake]

it's not v3

it's @gray rover 's fork what you're talking about i think

[gray rover]

^ me n noobies.
Key difference now tho, I intend to do it for 48k, Noobies most likely 44.1k

[lavish sage]

Yo everyone, something very simple again

I wanna train a voice to make an AI model, so it can become something like that

https://huggingface.co/imcertibtw/RVC_MODELS/resolve/main/LeonKennedy.zip

How can my audio can become a link to use ?

[ionic pumice]

upload your model into huggingface, after it's done uploading left click the download button and copy the url :p

[lavish sage]

I may be fucking dumb but

Huggingface is a site FULL of AI stuff ; where's the AI training one ?

[ionic pumice]

oh wait you meant training?

uh yea huggingface can't do training unfortunately

but there are options to train

[lavish sage]

Yeah, I may be explaining poorly

[ionic pumice]

if you have a good enough gpu you can train locally

[lavish sage]

Excuse my french, litteraly haha

[ionic pumice]

you're fine dw

[lavish sage]

I'm on a phone :/

[ionic pumice]

ah

[ionic pumice]

-cloud

huh

[ionic pumice]

-colab

[rare sorrel]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

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

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

[ionic pumice]

there's also kaggle but idk if it's been updated yet :p

-docs

[rare sorrel]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[ionic pumice]

have the docs too just incase xD

[lavish sage]

Thank you, but That's not what i'm looking for ;-;

I would like a website, just to train an audio into an AI model

[ionic pumice]

ye the colab will do that

as long as you read the instructions correctly youll have it training in no time

also i apologize if you can't understand what im saying lolz :p

[half basin]

hello bro's

[silk plume]

Yo what software do you all use to clown voices i feel like some software is better to clown voices with than other ones

[gritty rover]

#✦│chat

[covert lake]

@elder willow did u copyright strike me

[next spoke]

Hi I’m dina

[covert lake]

smh how dare u

do u know also @pale patrol

we someway got copyright striked by sony at the same day

sony is doing a genocide ig

what gif

[pale patrol]

ahem

[covert lake]

i think u should have perms for that

[covert lake]

yea fr

the notification was at 5 pm for me

wbu?

if they did it at the same time damn

[pale patrol]

this was yesterday

i just checked it today

[covert lake]

smh ps5 should get the switch treatment

-# people basically speedrunning at making emulators for it

wait dont u literally make ai covers

[pale patrol]

tbh tho its just a song it cant hurt u ;-;

[covert lake]

it actually does tbh

that's 2 more times not 3

[pale patrol]

jokes on u i didnt upload for 6 months

that ai cover was trash anyways cuz i was using adobe podcast to clean my model at that time

and yes i deleted it

[covert lake]

wtf sony owns IT girl??

[pale patrol]

DAMMIT

they have limited swearing and ive been here for like a year now

[covert lake]

deleting every cover

not worth it lol

[pale patrol]

im keeping mine but ill private the others

[covert lake]

they can stilll get a copyright strike

[pale patrol]

bruh

[covert lake]

bc if sony saw it and did a claim when it was public, it can still strike u

and well if they saw one already, they prolly checked others

so ima just delete

[trail pasture]

Anyone got exra money on cashapp im saving up for a bowling ball

[light girder]

what is codename doing as of right now

what's new in that new fork ?

that they are preparing

[alpine geyser]

i keep on seeing like songs with ai travis scott n shit and it actually sounds real

[light girder]

send one

[alpine geyser]

would anyone just so happen to know how to do that?

[light girder]

wanna see

[alpine geyser]

hold on

[obtuse flax]

I wanna see too👀

[alpine geyser]

you got spotify?

[light girder]

uhh yes

yes i do

[alpine geyser]

TWXN on spotify makes music with ts

he uses carti, travis and others

[night lake]

Take a look in https://discord.com/channels/1159260121998827560/1159290193619189821

He's pretty active there

[light girder]

i don't have access there smh

[night lake]

Whaaa

[light girder]

what is that channel ?

[night lake]

AI testing channel

[light girder]

Did they sent example of their development ?

or audio results ?

[night lake]

Him and noobies send development updates, audio samples, etc

[light girder]

care to send some audio samples ?

🙏

[night lake]

They are tests on very small sets with very prototype build

It ain't going to be good

[light girder]

Oh but what's the difference like

what are they implementing

and what is expected in the final ver

[night lake]

Codename and noobies are doing different things so

Codename is adding a new discriminator and some stuff

Noobies is adding 44.1k training with also a new discriminator

[light girder]

what does it mean, better better audio quality

or not that much ?

[night lake]

Hopefully

[chilly lake]

better results from training

updated vocoder with MRF layer

better fills for upscaler gaps

[light girder]

Oooo Noobies comin in clutch

[chilly lake]

new 44100Hz with very little mirroring at the top range

[light girder]

so you're expecting major improvements ?

[chilly lake]

unfortunately both need a new pretrain created

I expect more stable training and better fidelity

[light girder]

I see, and when do you expect it to release

[chilly lake]

the new branch for Applio is available to try

but again, no pretrains are available

[light girder]

and if realtime would be usable ?

[light girder]

Ohhhhhhh

i expected it to be used with uh

realtime

[chilly lake]

it is not realtime

[light girder]

ahhhhhh, thanks for the info!

[chilly lake]

i mean the models may be used for realtime later, once wokada authors add the same generators

[night lake]

Can't you just ctrl c ctrl v the generator or will a lot of stuff have to be changed for realtime

[chilly lake]

different model

different size of weight tensors, they are not gonna fit

[night lake]

I see

[chilly lake]

it is like when you tried all those config changes

most of them did not work with a default pretrains I bet

[night lake]

I actually created a new pretrain for that test

[snow sedge]

Model code should be simple to add once it's ready to use

[unreal canyon]

HI

[night lake]

Hello 👋

[unreal canyon]

CAN i ASK YOU SOMETHING

if I train model does it take a lot of time ?

[night lake]

Depends on your gpu

And batch size and dataset length

[unreal canyon]

yeap I will try might share here too people offer here a lot

[chilly lake]

batch size has only an indirect effect... it the epoch takes about the same time whether you use batch 2 or 22

[manic heron]

Yo guys i got a question

[chilly lake]

what happens when you use batch 2 vs 22 is anoher story and the length of overall training may be affected

[manic heron]

what do i use for the voices

[solar torrent]

Use what?

[manic heron]

1 sec

[night lake]

I see

[solar torrent]

Please specify on which program you would like to use for voice.

[manic heron]

idk

which decent ones is there

idk i found this random video https://www.youtube.com/watch?v=oOBjntI2xK0&t=15s

uh sorry if i sound dumb lol

what progams should i use

[solar torrent]

If it's a tutorial, tutorial videos on YouTube are outdated, just letting you know.

[manic heron]

uh

so i cant make ai voices say stuff?

which program ):

[solar torrent]

Real-time or audio conversion?

[manic heron]

yeah like the audio to the ai voice

like the ai voice saying some audio i put

[solar torrent]

-rvc

[rare sorrel]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[manic heron]

O:

[solar torrent]

If you have a fast and newer GPU in your PC, you can run Applio on your PC locally. Otherwise, you might wanna run it on a cloud service instead.

[manic heron]

hmmmm

if i tried cloud would it be the same speed?

[chilly lake]

if you want text to speech, it is a different thing

[manic heron]

like is there wait times like other people waiting for theirs

[manic heron]

i need audio

audio to uh ai voice saying it

[chilly lake]

text to speech is text to audio

[manic heron]

yee

uh so

which one should i use ):

[chilly lake]

https://huggingface.co/spaces/coqui/xtts

[manic heron]

o

[chilly lake]

you can upload a reference audio (15-30s) for a voice you want

[manic heron]

can i import custom voices?

[chilly lake]

you can import any audio file with someone speaking and you'll get that voice out

[manic heron]

wait so i can get some custom ai voice to say some random audio i give it?

[chilly lake]

step 1) get an audio file of someone talking 2) use the link and upload that audio 3) enter text 4) click generate button 5) enjoy the audio of the person saying thins with the voice you provided

[manic heron]

o

[solar torrent]

Ok.

[manic heron]

I ment if i get a custom ai voice then get a audio file of some random guy singing (idk) and can i make the ai voice say whats in the audio file

but its copying

so it doesnt sound boring

[solar torrent]

-tts

[rare sorrel]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[manic heron]

ooo

[solar torrent]

There's RVC TTS fork under W-Okada command?

[chilly lake]

seems like you replace one voice with another

[manic heron]

ye

[chilly lake]

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

[manic heron]

uhh ill just use it for now

ima test it for now

[solar torrent]

The Ilaria RVC is a fork of RVC, developed and hosted only on a Hugging Face space. There doesn't seem to be a local or Colab version of it yet.

[manic heron]

ok

[solar torrent]

What the fuck? I can only use Colab for demucs now.

I don't have a CashApp account. What do you mean?

Bro got his account hacked bruh. Did he even set up more security on his account?

[subtle garden]

Hello, I'm not good at English. I use Google to help translate and talk to you guys. Please give me some advice. Thank you.

[errant sphinx]

Did ever use 2 factor Authentication??! 🤔

[neat bane]

Hi everyone, Im radiant, Good to be here

[carmine wadi]

hi

[solar torrent]

Who are you asking to?

[copper sun]

Bio suggests he's probably Thai.

[polar flax]

please refrain from posting songs that may be copyrighted as attachment instead of using external hosting

even as a safe bet, you shouldn't also directly attach musics generated using suno/udio (as they might get into legal trouble too)

[solar torrent]

If you ever feel like term and rule of AI Hub were too much for you, post an audio of you singing that song instead.

[covert lake]

I don't think it works like that

[covert lake]

Yea especially now

They are doing genocides

[solar torrent]

Not only you show your own skills to let the world knows, you can also scare the shit out of AI fans that don't have a skill but only use anything AI for their entire lives.

[covert lake]

you can also scare the shit out of AI fans that don't have a skill but only use anything AI for their entire lives

Not like this is going to completely stop any type of AI covers to be pubblished lol

People can just:

• don't pubblish it
• buy a mechanical license
• Pubblish on tiktok where under 1 min video aren't subjected to copyright
• do like piracy

Anyways I don't really care about some old covers, that's not what my channel is about

[polar flax]

though non-public contents shouldnt get copystruck, otherwise it would be privacy invasion

[covert lake]

They do actually get copyrighted

[polar flax]

||the safer way is to use temporary hosting by request, torrents, dark webs, or kind of p2p ways||

[carmine wadi]

im back

but bored

[subtle garden]

Thailand 🇹🇭

[polar flax]

there are some thai fellows here (not me ofc)

[solar torrent]

AI music is a form of AI art, much like real music, but I still have them seperated from real human-made art.

[covert lake]

[slim siren]

Why is laughing sounds weird

Setting rec?

[chilly lake]

for voice changer? it is more or less impossible to make it work

[slim siren]

What's your settings

Cuz mine sounds off at some point

[chilly lake]

it is impossible to make non-speech sound correct

taking = ok, singing = maybe, laughing/breathing/coughting = big nope

[slim siren]

Talking sounds off sometimes, idk why but I assume it's my settings

[slim siren]

Can I get your settings?

[chilly lake]

👉 https://discord.com/channels/1159260121998827560/1159290161683767298

[polar flax]

just try not to laugh

[slim siren]

But talking tho

Sometimes talking just sounds

Off

[polar flax]

seems the model quality issue

[slim siren]

I think so

[polar flax]

or the accent incompatibility

[slim siren]

Gotta find a voice model that suits me

Is that what u tryna tell me

[worthy coyote]

Upper Right Block Diagonal Lower Middle Left To Lower Right

🭓

well i can't see this it's just a square

[tulip bobcat]

👋

[solar torrent]

[frank quarry]

[polar flax]

https://tenor.com/view/hiii-cat-wave-cat-wave-hi-gif-174708314057938083

[tired jasper]

hi, what's the best denoiser for voice?

i'm trying to denoise samples like this

[waxen knoll]

They changed roles color and style???

Wtf where's my Musician role

[polar flax]

[waxen knoll]

Did they deleted that role?

[solar torrent]

There was a server admin exchange going on, likely to make this server to feel better I suppose. Not only they changed the "AI Hub" to "AI Hub by Weights", they removed certain server roles they thought were useless, and changed current roles to something else.

[tired jasper]

welcome to the cult. Sacrifice is tonight, don't miss it

i made some adjust in resemble enhance code, now it is much better

it don't deal well with some specific noises yet, but it's much better

it was trained on common environment bg noises

make sense

static noise for example will be ignored

if someone knows a finetuned resemble enhance for uncommon noises please tell me

[harsh rampart]

Yo taylor

Im really happy for you imma let you finish

But... sicko mode or mo bamba

Why not both

[tired jasper]

🔨

[solar torrent]

What the fuck?

[west stratus]

hi!

[solar torrent]

Hi.

[warm stream]

hi

[thorny drum]

hello guys, how do i use an ai voice in real time?

[covert lake]

what's ur pc gpu?

[solar torrent]

-realtime

[rare sorrel]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[solar torrent]

If you have a GPU that's faster and newer than NVIDIA GeForce GTX 1000 series in your PC, sure you can do it on your PC. The first link for forked version runs best. Otherwise, you might wanna look for a cloud service instead.

[thorny drum]

rtx 4080

[solar torrent]

Great.

[covert lake]

Great, u can use Wokada (program that uses RVC, speech to speech, models in realtime for calls), and the deiteris fork (modified version) has even better performance

-rt

[rare sorrel]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[covert lake]

1st link

[fair geode]

Hi!

[bright stirrup]

Hi!

[scenic flare]

hi

[tired jasper]

someone able to use wokada on linux?

[covert lake]

https://rentry.co/ForkVoiceChangerGuide#linux

-# I didn't do it myself, but it's in the guide

[tired jasper]

thank you

[elder willow]

also as colab uses linux (iirc), you can use colab notebooks of w-okada there if you don't have a good pc

[tired jasper]

the problem i'm facing is to route the audio stream to right devices

[elder willow]

do you have virtual audio cable installed?

[tired jasper]

my distro doesn't have portaudio i guess

[elder willow]

oh, well you should switch to ubuntu ig

[tired jasper]

it's ubuntu baseed

alsa is installed and also pulseaudio

[elder willow]

oh, well i'm sorry but i can't help you too much as i don't know too much about linux

[tired jasper]

wokada don't recognize my audio output

[elder willow]

maybe you dont have speakers or headphones, or you have problems with the portaudio

[tired jasper]

i have headphone

[elder willow]

oh cool then, it may be just the portaudio

i have no idea what's going on with your linux tbh

[tired jasper]

idk too much about audio drivers on linux

[elder willow]

maybe youtube tutorials can help you

[tired jasper]

i just know that my distro uses pulseaudio

other distros don't use it

[tired jasper]

yeah, maybe, i will try again later

thx

[elder willow]

yw

[warm wedge]

hi guys

[elder willow]

wassup

[tired jasper]

hi

[tired jasper]

2024-12-05 15:53:06,850 ERROR [VoiceChangerManager] Voice Change is not loaded. Did you load a correct model?

it start to work for few minutes, then only this ERROR

[woeful remnant]

Oh

[elder willow]

EVERYONE FULL o1 OUT

[keen garnet]

is there any free way of cloning my voice with ai?

[elder willow]

No

[minor blade]

Of course it is.

[minor blade]

Of course, all you need is record at least 20-30-40 mins of yourself speaking (if possible with a good quality mic)

And then read the docs:

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

[elder willow]

Didn’t know

halo

[minor blade]

We got a variety of ways/guides which technically make you able to train models for free since long time ago.

[granite geyser]

wsg

[hot panther]

yo

[tulip coral]

hi guys im new here and im hoping someone could help me?
im trying to make a AI version of the infamous playboi carti and lil uzi vert 16*29 grail "super soaker" ive never made AI music before ever in my life and was wondering if someone could hlep me

[elder willow]

where can i find

the rvc models that arent avaible

[solar torrent]

#1175430844685484042

Please scroll a little bit in channel section and you'll find this channel. If you have a problem viewing model posts there, refresh and wait a little bit.

Me and mods have been saying this for many times, but because some of you are new to this server so.

[tulip coral]

can you perchance direct me to a tutorial (if there is one) on how to make ai music

i downloaded a carti voice model and im tryna make ai of a unreleased grail snippet of his

[tulip coral]

im sorry im new here

[solar torrent]

What? No. Most tutorial videos on YouTube are outdated, so no.

[tulip coral]

and there's none in here?

[tulip coral]

time to go witch hunting in every server im in for someone who knows how to make ai carti music

[solar torrent]

-rvc

[rare sorrel]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

• How to use RVC Mainline Colab by Cauthess
• AICoverGen Colab Guide by Eddy (Spanish Helper)
• Create a model with RVC disconnected (colab) by Angetyde

[alpine granite]

your gonna have to still rap it since its a voice changer

[tulip coral]

what does rvc even stand for

[tulip coral]

woah woah woah slow down

[alpine granite]

retrieval voice conversion

[tulip coral]

i have to genuinely rap out this song??

[alpine granite]

either that or someone else has to

[solar torrent]

I don't know what kind of a voice model you've downloaded, wtf. But RVC is an audio voice changer.

[tulip coral]

are you familiar with trackerhub

[solar torrent]

No.

[tulip coral]

oh

i js got the ai model from there

[alpine granite]

better be one thats trained off leaked protools sessions

[tulip coral]

see the tragedy here is that i found the PERFECT ai carti voice i needed in #1175430844685484042 but whenever i click the link it says 'invalid password'

[solar torrent]

What kind of a voice model you've downloaded there? RVC or something else? Just want to make sure a voice model is compatible with RVC.

[dawn zenith]

training on 48 gb vram card and its taking 1 hour per 300 epochs

crazy

[solar torrent]

Who are you talking to?

[dawn zenith]

some information bud

[alpine granite]

which one. Maybe weights archived it if its posted in #1175430844685484042

[alpine granite]

do you just post that everytime someone pings you

[tulip coral]

no idea.
trackerhub is a google spreadsheet full of various artists with basically every leak, snippet, released song, music video, artwork, tracklist etc from specific artists
https://docs.google.com/spreadsheets/d/1qzeFdpUPr7E0jOFwWSXd8LF30ZLjz1CSVEBiG8gPHTU/edit?gid=1792554832#gid=1792554832 heres the 'ai models' sheet on trackerhub it said at the top to join gg/aihub so idid

[solar torrent]

Some light blue users didn't turn the @ off.

[tulip coral]

https://discord.com/channels/1159260121998827560/1246669071752106025

[dawn zenith]

wont work if you dont click on my message - reply

[solar torrent]

Ok.

https://cdn.discordapp.com/emojis/1113764166260494436.webp?size=48&animated=true

[tulip coral]

whatd you delete my message for

[solar torrent]

Who deleted your message? I'm not a mod, I'm just talking around.

[tulip coral]

is there an auto delete system or smth?

@sage musk why is your carti self titled ai model no longer available to download from huggingface? tryna download it for a AI im doing

[solar torrent]

Yes, there is. To make sure you don't send anything that's suspicious.

[analog oak]

hola chat 🙂

[spark perch]

Damn.

They changed Lora in Weights, now there are no pictures of real people. It's a waste

[spark perch]

Come on Weights, go back to what was good. Generic generation is horrible, boring, the site will lose a lot of visitors. It's a shot in the foot

If you don't want it to be public, make it private.

[solar torrent]

Yo bro, this server is AI Hub by Weights, not Weights.

If you've joined Weights' Discord server, I suggest you can open your issue about using this website there, or contact Bea the admin for more information.

[spark perch]

Its name is WEIGHTS, the Discord is by WEIGHTS and I can't say anything here It's like buying a house, having a car, a bed and choosing to sleep on the street and on the floor.

[solar torrent]

Saying like this ain't gonna fix anything, just letting you know.

[spark perch]

But I just want you guys to know, people need to give feedback here and on Weights.

I am sad

[spark perch]

You are the creator of Weights

We know

🔥

[solar torrent]

NO, no. I didn't make this site, I just did some translate for this site.

[spark perch]

I created a photo of Freddie Mercury and it generated a regular Indian guy

Damn!

[spark perch]

Test yourself

It's been a while since I've been here. Good times with Applio and RVC

[night lake]

Weights doesn't "own" ai hub

[solar torrent]

I don't know if you are being SARCASM or not, but like please understand that I'm just here to give my basic knowledge to people about using AI. I'm not here to fix things like an admin would do, I'm not even a creator of this site either.

[spark perch]

It was sarcasm, I know Weights is someone else's

[spark perch]

It's like 'YouTube by Google'

[night lake]

Ehh

[tepid basin]

More like the Waze situation where Google buys it and lets it rot slowly

[polar flax]

you can download the lora if not wrong, then use another site or do locally

[spark perch]

I only use smartphone

That's the problem

[tepid basin]

We got RVC on android so I wouldn't be shocked if there's SD for android somehow

Prob slow asf

[polar flax]

there's A1111 for android but yea would be slow af

[spark perch]

I like Weights because it's practical and free.

[polar flax]

or you can also find a colab notebook

[spark perch]

But man, why did they mess with something that was good?

Everyone is complaining about it.

[night lake]

Money?

[tepid basin]

So you can't generate "real people" anymore despite the lora? Trying to catch up here.

[night lake]

Idk

[solar torrent]

An average Flux LoRA model size is kinda large to download it to your phone, but it eats way less storage than downloading a Genshin Impact.

[spark perch]

No

It's not money, probably someone was correcting the sharpness of the generated photo and made this mistake.

[solar torrent]

I don't know what you are on, but I can still generate some Ye West images there on this site.

[spark perch]

Try generating pictures of Freddie Mercury

[solar torrent]

No.

[tepid basin]

That lora worked good before right? Maybe its a shit lora?

I wouldn't be shocked if they disallowed generating celebrities, its very common for these online generators to do that. There's lots of legal bullshit they need to keep up on

If you wanna get around it, there's local SD solutions or Google Colab

[faint condor]

35 hours of minecraft

[tepid basin]

I host a server that's why

[solar torrent]

There are many options available, please listen.

[faint condor]

that showing single player

[tepid basin]

LAN world + reverse proxy (e4mc mod)

[solar torrent]

My laptop is unable to play Minecraft because it's too old.

[spark perch]

The problem is that it still generates Lady Gaga, Trump, for example.

[faint condor]

[faint condor]

my pc is too new but I don't have wifi

[tepid basin]

Maybe the lora is shit, or weights celebrity filter isnt very good yet. Idk

I don't know how they even block that stuff

Surely it just grabs keywords from prompts

[spark perch]

They shouldn't ban it, after all it's just a copy, something generic.

[polar flax]

my intel igpu laptop can still run it alone, not with a shader or ofc voice changer

[solar torrent]

Wi-Fi card in your PC or a Wi-Fi router? Buy one, and you'll be able to connect to Wi-Fi.

[spark perch]

I hope they fix this, it was generating normal photos until now. Today it got worse.

[tepid basin]

Sadly due to legal stuff, they need to do what they need to do. Try looking into online notebooks (colab) to run these loras

[faint condor]

I don't have wifi access or connection

[tepid basin]

Discord messages coming through a wormhole?

[polar flax]

mobile data fr

[solar torrent]

I didn't have to worry about generating image of human not looking good. I just train some random images of things I like as Flux LoRA models there.

[faint condor]

it's s coming from mobile data dude

[faint condor]

yeah

[spark perch]

I think the error is general, even private photos of ordinary people are being altered

[faint condor]

dude I don't know anything about ai but I want to learn

[spark perch]

A man commented on the feedback with a photo, he created a model of himself and generated photos of him working out in a fitness tank top. Today he tried to take a photo and a muscular woman was generated. He was very frustrated lol

[granite tartan]

How nice, they sold open source to some American trash

[faint condor]

[granite tartan]

Hey, Weight guys, I'm letting you know that they removed the models more than half a year ago without my permission and they still haven't done it. You already know my discord

[solar torrent]

Much like Stable Diffusion XL, my recommended dataset image to train on Flux is to remove background from every image so it can focus on the thing, while every image starts off with 1024x1024px.

[glad nebula]

removing backgrounds of the images will end up in your character looking copy pasted in the image rather than naturally blending in it

[solar torrent]

Really? I'd put the white background instead. Sometimes, a trained model can be generated on a background and blend with it.

[glad nebula]

i mean it does not removes the possibility of creating backgrounds, so is like if you prompt for your character in the night, the character will look very bright in comparison to the background

thats for 2d characters

idk about real ppl

[solar torrent]

I never train a model of real people, I only train models of cartoon/anime characters.

[faint condor]

[alpine granite]

good thing flux is open source

i think civitai allows real people though

weights gg kinda cringe cause it follows ai ethics

[faint condor]

huh huuh huuuuh

[solar torrent]

I'd try to do both.

[solar torrent]

Training an NSFW model on Weights isn't really worth it.

[faint condor]

[faint condor]

[elder willow]

https://media.discordapp.net/stickers/1257836590265274429.png?size=160&name=HAPPY+MONDAY!

[solar torrent]

[rough roost]

too many pixels i can just barely read the text

[elder willow]

fake nitro sticker

[quaint sinew]

beloo

[tepid basin]

Sounds scary similar to the real never gonna give u up

[red escarp]

hi yall

[solar torrent]

https://cdn.discordapp.com/attachments/1159265300915687534/1296466465867694151/BBL_Drizzy_Metro_multiplayer_edition_short_only_vocals.mp4

[solar torrent]

Row Row Your Boat in the style of Carnival by Ye Dolla $ign, generated on Weights.

AI song generator on Weights is kinda goofy, it only has a few lyrics input to begin with.

[pale marlin]

,

[vale merlin]

Hi everyone, I'm newbie just joined a few minute

[uncut oak]

warwick ?

[covert lake]

Huh

[errant sphinx]

I talking about you sir..!!

[errant sphinx]

Stare

[errant sphinx]

Oh dang, another person had SD Creators...!!

[polar flax]

just caption the background

[solar torrent]

Caption or labelling?

[polar flax]

say "it features white background" or something

[glad nebula]

if the background is a forest, then you caption forest

but dont tag too much elements of the background, i noticed if you do that, the lora starts to believe these are parts of the character

[solar torrent]

https://cdn.discordapp.com/attachments/272814335238340608/1314631633214767176/distort.jpeg

[solar torrent]

Every dataset image I made is just simple white background.

[errant sphinx]

Let me tell you one answer for me...?!

[errant sphinx]

How did you even get that a quest..?!

I just blur content for some purpose!!

For like WTH..!

[tired jasper]

F5 still training, omg, i'm feeling so stupid

hope it worth

[minor blade]

If you want to promote something go to the #1159290752195633273 channel.

[elder willow]

stop spamming that

please, nobody cares about your cashapp or paypal thing

[analog cosmos]

no

[ruby wharf]

this is ai generated?

@minor blade pls remove this

seems like a scam not gonna lie (is a scam)

sugar mommy...

[analog cosmos]

no sugar mommies here

[river verge]

Dilly ding, dilly dong! A new RegalHyperus drum model just released!
Waking the King V2 (Drum model no. 548)

[elder willow]

0.0

[wind stratus]

interesting, this reminds me of those old jukebox ai continuations

[elder willow]

i will use the vocals of that song for my inference inputs

[stark tangle]

Does anyone know how to make an AI beat, based off an other beat?

[tired jasper]

try this https://magenta.tensorflow.org/studio/

if someone knows a project that restores vintage music, please tell me

not exactly a remix

like a feeling that the song was made nowadays

https://www.youtube.com/watch?v=y9ukd6ZEbQY

you know that is vintage without any label

[rotund wren]

hi im kiyo im new to ai covers but i will try to learn

[placid hill]

How did you fix this

E:\RVC1006Nvidia>runtime\python.exe gui_v1.py
Traceback (most recent call last):
File "E:\RVC1006Nvidia\gui_v1.py", line 59, in <module>
import librosa
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\librosa_init_.py", line 208, in <module>
from .cache import cache
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\librosa_cache.py", line 6, in <module>
from joblib import Memory
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib_init.py", line 113, in <module>
from .memory import Memory, MemorizedResult, register_store_backend
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib\memory.py", line 32, in <module>
from ._store_backends import StoreBackendBase, FileSystemStoreBackend
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib_store_backends.py", line 15, in <module>
from .backports import concurrency_safe_rename
File "E:\RVC1006Nvidia\runtime\Lib\site-packages\joblib\backports.py", line 22, in <module>
import distutils # noqa
^^^^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'distutils'

E:\RVC1006Nvidia>pause
Press any key to continue . . .

[solar torrent]

This is not where you send your Python runtime issue. And I see that folder name, the OG RVC fork is outdated. So please go to #✨│ai-help for more information.

[tepid basin]

its not too hard. I suggest trying out Applio, with the models that are posted here in #1175430844685484042

[tepid basin]

i havent made a model in a long time

[astral matrix]

Hi,guys some one knows how to train a pretrain model on colab?

[night lake]

It's exactly the same as training a model just more data

[astral matrix]

I put all the data in a single speaker?

[night lake]

Yeah

[astral matrix]

Or i split them by speakers?

[night lake]

Nope

[astral matrix]

So i just disable the pretrain thing and just save last g and d right?

[night lake]

Save every d and g

[solar torrent]

A pre-trained model is a model that was trained from scratch so any model can be trained on. If you're looking for on how to train a custom pre-trained model, be careful on what you're doing.

[astral matrix]

Are 35 hours of data okay to train a custom pretrain from scratch?

[solar torrent]

At this time session length, you'd run out of time on Colab free version. Are you sure about that?

[astral matrix]

I have 83 units of colab pro soooo

[solar torrent]

-colab

[rare sorrel]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

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

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

[astral matrix]

Thankss

[solar torrent]

I'm not sure which one of these has the option for "training" available, but you might wanna try one by one with CPU mode on to save up your Colab unit points. If you found a notebook that has this option, you can use GPU later on.

[elder willow]

applio, rvc disconnected and rvc mainline

[astral matrix]

Also what are the HuBERT thing?

[rare sorrel]

HuBERT: A transformer-based model that extracts text from raw audio, previously trained on a masked prediction task, which RVC uses to train the voice models. There are several types of Hubert, some of examples are ContentVec, Japanese Hubert-Base and Chinese Hubert-Large. You can learn more about it in the Applio Docs

[mint beacon]

Hi

[astral matrix]

The page is broken lol

[rare sorrel]

📚 Documentación

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ Más guías

• Cómo usar RVC Mainline Colab por Cauthess
• Guía de AICoverGen Colab por Eddy (Spanish Helper)
• Creación de un modelo con RVC desconectado (colab) por Angetyde

[harsh basin]

hi

[forest crypt]

https://docs.applio.org/applio/faq

[faint pond]

Excuse me but is it possible to get rvc on mobile

[elder willow]

hihihi 잘부탁해

[solar torrent]

Hi.

[frank quarry]

[astral matrix]

Using colab yes

[faint pond]

It works

[elder willow]

it should work

[shut sage]

-colab

[rare sorrel]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• AICoverGen-WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Modified W-Okada's Voice Changer, Google Colab
• 🆕 FaceFusion UI, by Nick088 Google Colab
• 🆕 FaceFusion NO UI, by Nick088 Google Colab
• 🆕 EasyGUI, by Rejekts Google Colab

ℹ️ Note

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

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

[shut sage]

Omg wow easyGUI is back? thank God because I still need this although I rarely train models,I'm already learned in RVC v2 Disconnected and the regular free weekly training on Weights is back because this is the easiest Colab ever and because maybe the models that trained from there has better quality than models trained from Weights

[flat meteor]

hi

[barren mauve]

hi

[shut sage]

How to use FaceFushion?

[barren mauve]

ooh your in huggin face also : D

[ionic pumice]

[drifting bison]

what is the most realistic female voice you guys recommend me?

[barren mauve]

not sure lol

[worthy coyote]

real females

have the most realistic female voice

[drifting bison]

uh

like, youtubers or should i just search for females

lol

[worthy coyote]

were you rushing or dragging

[tepid basin]

Dragging I think

[lean gyro]

Y

[polar flax]

you shouldn't post full songs as attachment, no matter if it's generated

[ionic pumice]

[polar flax]

bro doing the opposite of what I said

also the right place to post is #1159290752195633273

[covert lake]

#1159290752195633273

#1159290752195633273

that’s the channel where you promote the content you make

It’s a role, u will be able to pubblish models on #1175430844685484042 and do free requests only

You can read https://docs.ai-hub.wtf/extra/model-maker-role/

Wdym? Show a screenshot

Btw you can send images here, no need to send in dms

And the error explains you that you need to upload the model either on weights.gg or huggingface

what’s ur model download link?

You need to put only either weights.gg or huggingface

not both of them

if neither of them works, maybe @night lake may know why

Try to wait for what razer says, he’s qc not me

Try putting the link u get by clicking the share button

Yw!

@little bobcat Btw it on the bot of #outdated-model-maker-role it should specify that its the share link

[normal stream]

hello lovely people, im to dumb to understand all the stuff in the github. anyone got a video explaining how to set it up?

[covert lake]

There is no updated video

All RVC & Wokada Videos are outdated

Explain me what u tryna do and what's ur pc gpu

I can link u the best written guide

[normal stream]

Im trying to use a voice changer in general. The voice changer i was trying to install is voice changer client. From github. I am on a laptop with a i7 12th gen and a rtx 2050

[covert lake]

I'm guessing you're talking about: https://github.com/w-okada/voice-changer
Which is commonly called Wokada / W-okada, it uses RVC (Speech To Speech AI Type) models in realtime for calls

There's also a fork (modified version) recently used, known as the deiteris fork as it optimizes the performance

-realtime