🧬│ai-chat

[dreamy heath]

haiii

[gentle canopy]

hi

[chilly lake]

@covert lake

there you go

small correction - inference audio was 12 min, the training set was 9:41m, after split it is a bit longer in Applio

so extra 75 seconds to work with and still faster

[tacit oar]

where can i find some english valorant voices (pls ping with answer)

[chilly lake]

first infer may take a bit of time as the model may take a few extra seconds to load

[warm brook]

hi

[chilly lake]

but in general there should be no noticeable difference between rvc and applio for inference using zluda

direct ml though in my expecience is much slower

[wild holly]

anyone know a good free voice changer?

[cerulean pike]

dubbing AI

[cerulean pike]

It has a lot of voice changers and it has a good soundboard

[hollow tangle]

lmao

[short owl]

h

[calm sandal]

hi

[icy pendant]

original: https://rentry.co/voicechangerguide

fork (optimized): https://rentry.co/forkvoicechangerguide

[snow sedge]

directml is problematic in both rvc mainline and applio (and in general) so no wonder. But well, for me it takes 58s to convert 12m audio with directml. Given that rx 6600m is something like 40% slower than 6700 XT, it sounds reasonable

[tacit oar]

where can i find some english valorant voices (pls ping with answer)

[versed island]

hi

[agile wren]

Hi

[wide lintel]

hIh

[soft portal]

hello

[tacit falcon]

h

[covert lake]

yea inference seems the same

[soft portal]

I can't access the channels..

[covert lake]

training seems like it took some sligthly seconds more in mainline ?

if i read it correctly

[covert lake]

what channel?

[soft portal]

voice models

[covert lake]

also forgot to reply, i was updating uvr5 ui kaggle 😭

[covert lake]

cant u click in #1175430844685484042

[chilly lake]

mainline training 69s/epoch, applio - 42s/epoch, so about 25-30% faster with torch 2.3.1

[soft portal]

yes but I don't have permission

[covert lake]

You don't have permission to post a model

but you can view the channel

[covert lake]

Yea, weird that the inference was slower on termux tho

[soft portal]

yes

[covert lake]

ig it depends by accelerator too

[covert lake]

Yea thats normal

U need to be a model maker to upload models

[soft portal]

ahhh

[feral halo]

Ayo? @soft portal level 1 !!!

[covert lake]

docs are down rn, but here is an archive of the guide: https://web.archive.org/web/20240915173810/https://docs.aihub.wtf/extra/model-maker-role/

this is how u become model maker to upload models

[soft portal]

ok thanks 😉

[jovial totem]

Good Afternoon

this is @jovial totem

today

uhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh ok

b

[elder willow]

hi

[tidal marsh]

koi hai ?

[elder willow]

im genuinely sped😭

[rocky kestrel]

skibidi

[pale patrol]

no

[wary olive]

yo

[feral halo]

Ayo? @wary olive level 1 !!!

[wary olive]

what program should i use for AI voices

i was used RVC BAT but it doesn't work anymore for me

help

[wicked moon]

where is epochs?

[icy pendant]

what

[soft girder]

weights

[wicked moon]

thanks

[dense ginkgo]

Hello

[soft girder]

ага, давай

[odd pumice]

Does anyone know of a good discord server to get help with ComfyUI?

[meager skiff]

Sup

[stable hazel]

@orchid slate can i use real time voice changer in amd cards like rx 6600 ?

[cosmic gull]

Sup

[chilly lake]

You can use directml build, or you can use a cuda build with Zluda emulator... both methods I've found are lacking

[late light]

how do i make it so i dont hear myself from the client

[old condor]

upscale image

[elder willow]

4o canvas is a death sentence for entry programmers

[feral halo]

Ayo? @elder willow level 2 !!!

[earnest dragon]

why and o1 preview canvas once that comes out will be even better

[covert lake]

https://discord.com/channels/1159260121998827560/1221468455723733094

[elder willow]

Blud I can't even imagine o1 preview with canvas
4o canvas obviously has better coding training than 4o
I've used it and it did a better job at coding than o1 preview, astonishingly faster also

[earnest dragon]

o1 preview has the best coding

[elder willow]

Doubt
Also I explained it in #🔊│ai-development

[odd pumice]

Anyone?

[elder willow]

anyone play coc?

[polar flax]

fork wokada works with less cpu usage than og one https://rentry.co/forkvoicechangerguide

[polar flax]

#🔍│help-ai-art
what specific purpose? because u can use it for SD/Flux, LLM prompting, and even RVC

[feral halo]

Ayo? @queen jay level 1 !!!

[bronze wharf]

Hello?

[fair viper]

@scenic token What happened to your group that banned everyone?

[feral halo]

Ayo? @fair viper level 2 !!!

[fair viper]

[whole badge]

chat

happy Halloween

[tacit falcon]

its not the skeleton day just yet

[light bone]

skill issue

[tacit falcon]

i see now, it is 3 am all around the world

[tacit falcon]

okay dude

[sweet harbor]

Hi, is there a website that will ask me a load of questions then give me a character sheet I can give to a ai model? I want to replicate some fictional characters but I'm terrible at writing character sheets

[cloud sentinel]

that answer ur question?

[sweet harbor]

No

[feral halo]

Ayo? @cloud sentinel level 1 !!!

[little nebula]

Ugh, fuck

You're mentally ill!

[sweet harbor]

Can someone tell me what's going on, am I having a stroke? I don't smell toast

[little nebula]

You don't need any help anymore. The dead don't need help

[little nebula]

c.ai?

spicychat?

You can try to find a character that you like or create your own

[steel isle]

Civitai

[cold zinc]

h

[stray badge]

Hi, i want to make a voice model from a character

I want to learn how to do it

Is there a guide that i can follow?

I got a lot of audio files of the voice alone

[haughty ibex]

My birthday is on 22rd this month

[tacit falcon]

too bad its not on halloween

[nova timber]

hi

[sweet harbor]

They let you create characters by adding personality traits and descriptions. I am bad at thinking of personality traits and I was hoping there was a site that would give me those traits based on questions

[feral halo]

Ayo? @sweet harbor level 1 !!!

[upper stag]

hello there

[grim osprey]

no entiendo

[feral halo]

Ayo? @grim osprey level 1 !!!

[night moon]

hello guys

[snow crag]

yo who wanna duo catfish with me

[earnest dragon]

you're weird if you do that and also you're gay if you catfish also

[snow crag]

hehe

[tacit falcon]

you either have self discovery to do or you are gay

[earnest dragon]

he is probably doing it for both

[snow crag]

anyw hmu i been making mad bread from this sh

[tacit falcon]

[earnest dragon]

No you haven't and if you did then you are gay and you can't say otherwise

[tacit falcon]

lolll

[snow crag]

the latter

check out my steam inv lmfao

[tacit falcon]

lollllllllllllllll

[earnest dragon]

so you're gay

[snow crag]

yea lets duo?

[earnest dragon]

Nah I'm not gay and I can make money other ways that are not weird.

[snow crag]

then wtf u talking to me for then?

goofy ah probs looking for gay ppl lmfao

[tacit falcon]

self report

[feral halo]

Ayo? @tacit falcon level 8 !!!

[earnest dragon]

you're the gay one here I'm just calling you weird for catfishing.

[snow crag]

bro got flustered took 5 minutes to reply and had to edit L

ayt buddy u typing too much, saw you had to rewrite too lmfao ayt bruh take the L im out ezpz

[tacit falcon]

self report

[earnest dragon]

At least I can make money not being a James Charles copycat hitting on straight men like you do

[snow crag]

lmao lemme guess u make money by straight up having gay sex? got it

learn to type faster gay boi ur wasting my time, closeted lmao L

[earnest dragon]

No I'm not gay at all and the money I make is from doing sfw non controversial deepfakes.

[snow crag]

oh hell nah thats even weirder holy shit, blocked

[earnest dragon]

that's not weird at all what you do is weird what I do is make a deepfake model of someone that's popular or have consent from.

[elder willow]

I need a code to use a models files the model has metadata.json model.pth and something else

[merry laurel]

hey, is so vits old news now that rvc has become so popular?

What are people using to make AI covers nowadays?

[crimson seal]

Hello, what voicechanger should I install? Please help

[stable furnace]

Va faire kaka

[feral halo]

Ayo? @stable furnace level 2 !!!

[elder willow]

@covert lake sorry but i needa ask one more time

#🔍│find-models message

how do i use this model after downloading?

[vernal abyss]

Has anyone tried swapping out the optimizer or really doing anything else to tune the voice models being trained other than just changing the epochs/batch size?

I don't have many datasets to test with but had some p interesting results swapping out the set-and-forget default adam implementation with prodigy

unfortunately not having a quality control on hand makes real evaluation kind of difficult since the default is p tainted too

[ionic pumice]

@polar flax not sure if this is a virus or not i don't wanna risk it lmfao

[covert lake]

with the rvc cli?

[dreamy basin]

Hello everyone
I’m jah am I welcome here

[covert lake]

whats u rpc gpu?

[covert lake]

yes, rvc is way better

Whats ur pc gpu?

its been since a year people use rvc

[covert lake]

whats ur pc gpu?

[robust vale]

how to train model?

[covert lake]

whats ur pc gpu?

?

[covert lake]

if u dont know ur pc gpu:
You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU

[merry laurel]

I have a 4090

[covert lake]

alr its good enough to do it locally (on ur pc)

docs are temporary down but i will give u the temporary one in a sec

[robust vale]

ty

[merry laurel]

thanks

[covert lake]

As you got a good PC, you can use RVC locally, you can choose between:

• Applio: A fork of RVC with some extra features like Applio TTS, same quality tho
• Mainline: The original RVC

@merry laurel forgot to reply lol

[covert lake]

yw, pls tell me the name of the gpu so i can help you

[merry laurel]

oh lol nw

[feral halo]

Ayo? @merry laurel level 1 !!!

[merry laurel]

So you are saying Applio should be good with my pc?

[robust vale]

it say GPU 0 Intel(R) UHD Graphics
GPU 1 NVIDIA GeForce RTX 3050 Laptop GPU

[feral halo]

Ayo? @robust vale level 1 !!!

[covert lake]

seems like you got an rtx 3050 (kinda bad) and integrated graphics which is worse
Could you also tell me the gpu memory of the rtx 3050?

[robust vale]

ok

it say 0.0

[covert lake]

i mean the number next to that

should be like: number of usage / total memory

[robust vale]

7.9 gb

[covert lake]

ye

[merry laurel]

cool thanks

[nocturne charm]

dang im surprised no ones made a scream voice yet for Halloween

[ionic pumice]

https://youtu.be/oKyqHbk_WlU?si=cBL5EaxRNDX2hav8

man

[covert lake]

not really sure if u would be able to do it locally as i heard rtx 3050 is bad

If you wanna try RVC locally, you can choose between:

• Applio: A fork of RVC with some extra features like Applio TTS, same quality tho
• Mainline: The original RVC

[robust vale]

ok

[covert lake]

Or you can use cloud (remote good pc):

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
  • Applio (ui)
  • Mainline (UI)
  • RVCDISCONNECTED (no ui)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
  • Mainline (UI)
  • Applio by Vidal (UI)
  • Applio by Shirou (UI, no guide as of right now)
• Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
  • Mainline (UI, No guide as of right now)
  • Applio (UI, No guide as of right now)

the best cloud one is kaggle

i would suggest to use cloud as i heard rtx 3050 is bad

[robust vale]

ok

[covert lake]

im not sure if that one is good enough for local

for any issues u can always ask in help channels!

[robust vale]

ok

[ionic pumice]

ultimate dry texter

[chilly lake]

I've tried a few different optimizers, it is not much different if at all

[chilly lake]

3050 with 8GB can run training and inference

gonna be about the same speed as colab

not too bad, not too great

[covert lake]

so its only bad for the rtx 3050 who have less vram?

i was told rtx 3050 is like a gtx 1650

#🧬│ai-chat message

[chilly lake]

it has tensor cores, so can to AI stuff, 8GB VRAM is sericeable

as long as it is not a laptop 3050 that is

desktop 3050 has 8x PCIe so it does not do well on old motherboards that only have PCIe Gen 4

[covert lake]

its a laptop gpu

so ye its better he does cloud

[chilly lake]

yeah, kaggle will be a better choice

[tulip nest]

can anyone explain what divergence is?

[chilly lake]

as for local, maybe inference can be done locally

[covert lake]

for training usually its suggested 8 or more gb

[covert lake]

@atomic vector explain “divergence”

[atomic vector]

so nick asked you to explain divergence, right rocococococo

ok, so divergence is a concept that can be applied in different fields like math, physics, and even finance

in simple terms, divergence refers to the process of something moving away or spreading out from a central point it can also be the point where two or more things separate or move in different directions

for example, in math, divergence is used to describe how a function or a vector field changes as you move away from a certain point in physics, it can refer to the way a fluid or a gas spreads out from a central point

in finance, divergence can refer to the way different economic indicators or market trends move away from each other

does that make sense
-# AI-generated responses may be inaccurate; please verify important information.

[covert lake]

LLMs are 🔛 🔝

[chilly lake]

with 4GB vram and nvidia some features may need to be disabled, like CUDNN to prevent pytorch from running out of memory

[covert lake]

btw, u work for applio & mainline?

i remember seeing u in both of their servers

[tulip nest]

ok the bot is not that useful here 😂 divergence/convergence has something to do with the graphs?

[chilly lake]

not the mainline, I went to check RVC devs server, but it is mostly dead lol

[covert lake]

ye rvc hype is kinda ded

[vernal abyss]

seems pretty surprising, at the very least something lr-free like prodigy should be substantially different than the untuned AdamW that everyone is using as a baseline

[chilly lake]

had some questions but nobody could answer there, so I assume original devs have moved on to better things

[covert lake]

yup

rvc boss now works on gpt so vita

Wok works on a wokada-like tts app

[covert lake]

u mean tensorboard? I thought u meant the word lol
Might need to ask in #1192011222023950368 as i dont really know much what to reply

rvc v3 is as much probable to be released as gta 7

[earnest dragon]

yeah I'm more into seedvc and svc fusion the rvc nowadays

[covert lake]

never heard about those

i mean, ai covers hype is kinda ded in general

[vernal abyss]

it would be a bit unrealistic to expect continued support for rvc at this point anyway, it's actually kinda surprising how long this architecture has lasted without more significant change

[gilded imp]

Hello,
I'm Christian work as a software developer and project manager. Have a Youtube channel on Project Management.
Extremely happy to be here.

[earnest dragon]

seedvc is zeroshot sts and svc fusion is a not that known version of svc

[vernal abyss]

image gen has swapped substantial portions of its core multiple times over the last 2 years

[covert lake]

rvc is better than svc tho

last svc version i heard was 4 or 5 tho

[covert lake]

yea, rvc kinda stayed like this

[earnest dragon]

there really isn't any svc fusion and rvc comparison

[covert lake]

seedvc seems cool tho

didnt see a sts 0shot before

[covert lake]

most people swapped from svc to rvc, i personally have never tried svc but from tea i heard svc was way harder to use

if i remember correctly

not sure if it was tea or someone else who said that

[earnest dragon]

there is many version of svc there is even ddsp

[tulip nest]

gotcha, thanks

[covert lake]

dunno about that

but since like a year everyone kinda uses rvc

like most voice models i ever seen are rvc

almost never seen a svc model

[earnest dragon]

yeah and they don't like trying other sts stuff that's not rvc really

[elder willow]

any i js wanna be able to use the model

but locally even if slow

I'll have an online version yeah but I wanna make it work on my computer by itself yk

[covert lake]

If you wanna try RVC locally, you can choose between:

• Applio: A fork of RVC with some extra features like Applio TTS, same quality tho
• Mainline: The original RVC

[covert lake]

would u mind sending the projects u said tho?

Also is svc fusion few shots or

im mostly interested into the 0shot one tbh

[elder willow]

didn't you send me a archived link of the mainline

[covert lake]

as the docs are still down, i deployed temporarely the docs, hyperlinks work correctly, the guide doesnt change tho

[elder willow]

okay

I downloaded whatever was on the link

[chilly lake]

if you dont want to smash your head against a wall, get a "compiled" rvc version, dont try building from source

[elder willow]

atleast tried but uts like 5gb and keeps failing

[earnest dragon]

svc fusion is few shots

[chilly lake]

latest applio can be build from source or download a prepackaged build too

[elder willow]

I'm trying again from the link you sent

[covert lake]

ye the guides on our ai hub docs are pre compiled

[covert lake]

Would rvc be considersted few shots or, i forgot tbh

[earnest dragon]

yeah few shots or many shots

[covert lake]

nvm searched and the right term should be multi shot

bc it needs way more training than a few shot ai like for example gpt so vits

[covert lake]

btw could u send a link of svc fusion? Was able to find only seedvc on github

it seems also pretty updated

[elder willow]

installed the mainline file that was like 5gb and extracted to desktop what now?

[chilly lake]

unzip, preferable to a folder that is not under OneDrive / has no spaces

run whichever .bat file

[elder willow]

well on my desktop theres no spaces in the file path

but i can put it on the drive if thats needed

"C:\Users\natha\OneDrive\Desktop\SnippetEditor\RVC1006Nvidia"

[chilly lake]

there's go-realtime-gui.bat for realtime voice changer, there's go-web.bat for inference and training

[elder willow]

okay

[elder willow]

yeah but no space in it

[chilly lake]

" to a folder that is not under OneDrive"

just move it to C:\

[elder willow]

is there a reason for that

alright

[chilly lake]

there reason for that is you dont want to sync all the files it creates to MS

[elder willow]

"C:\SnippetEditor\RVC1006Nvidia"

[chilly lake]

and wait until it saves stuff there

[elder willow]

ngl

ion fw onedrive

i uninstalled it after it ran outa space and everytime i ever reset or get new computer immediately uninstall

thats funny

nevermind

[chilly lake]

so, do you have a gpu? 🙂

[elder willow]

its workin

um

[chilly lake]

well...

RVC1006Nvidia why then

anyway, with CPU the best you can do is a very slow local inference

and training that sets your laptop on fire

[elder willow]

quite literally?

[chilly lake]

it is very demanding, so yeah...

[elder willow]

where do i put the takeoff model to use it 😭

[chilly lake]

index goes into logs, pth goes into assets\weights

[elder willow]

my computer is dying

😭

i think u were right

[elder willow]

my computer is finna catch fire not even training

[feral halo]

Ayo? @elder willow level 6 !!!

[chilly lake]

just use https://huggingface.co/spaces/TheStinger/Ilaria_RVC for inference

[earnest dragon]

yeah I'm downloading it right now to upload to Google drive and what's better for you pkg or windows version

[elder willow]

this is beautiful

[chilly lake]

at least it is not your laptop on fire

[elder willow]

idk why i did a whole song 😭

no way it errored

[polar flax]

only if you have melting connector on a 4090

[covert lake]

im on windows, bUt integrated graphics so i will use cloud (so using linux), its not on igthub ?

[earnest dragon]

No they have a website and stuff just all there downloads are chinese download links which I can download from

[earnest dragon]

okay then will get the pkg for you

[covert lake]

if u could send the website in dms or here would be cool

[earnest dragon]

https://sf.dysjs.com/

[covert lake]

will check it out when im home

ty

[elder willow]

why cant it infer anything at all

it all errors

i did a normal song then a snippet that had the instrumental removed

[spiral portal]

Hi

[covert lake]

try without the “”

[elder willow]

this is the cmd line

[elder willow]

i got this to work

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

so everyone that goes on there has access to the model now?

[covert lake]

Sorry but i dunno about local, seems like a low memory issue

[covert lake]

ye cloud is easier

#1159290752195633273

[dreamy basin]

Hello
Who’s here??

[covert lake]

if u need help ask in help channels and be specific

[elder willow]

alright

[ocean badger]

hi, i cant see voice models

[elder willow]

would you happen to know of any python vocal splitters

that are local

[polar flax]

there's no way it can run out of memory while processing not too long voice audio in cpu mode and 16 GB ram

[covert lake]

#1175430844685484042 , you have access to view them but not upload them

[covert lake]

🤷‍♂️ i dont do things locally

[elder willow]

does anything exist to take beats and replicate them?

but better

[covert lake]

this is what made me guess that

i also tried searching for that error but seen there was no solution so

also u sure he got 16gb of ram

[elder willow]

pray for me

can i even ai split on this hunk of junk

[covert lake]

@polar flax yea thats a not enough ram problem, barely 4gb usuable

[covert lake]

so yea i dont think u can do anything on local

[elder willow]

is that to the ai split thing

[covert lake]

ai split? U mean uvr?

[elder willow]

if thats seperating vocals

[covert lake]

if u mean uvr, i really dont think so or it would take hours

[elder willow]

im not sure the terms exactly

[covert lake]

uvr = ultimate vocal remover

to split vocals

i dont think u can do literally anything local

too weak pc

[elder willow]

rip

life is ruined 💔

[covert lake]

u can use cloud like i do

Theres uvr cloud

[elder willow]

idk its not the same

[covert lake]

its the same program

[elder willow]

i wanted to do my project and have it simplified now i gotta goto multiple random sites for each process 😭

[covert lake]

Cloud is your only choice

[polar flax]

lmao 4 GB ram in 2024 💀

[covert lake]

Well not everyone is rich

[elder willow]

im poor what u expect

[covert lake]

idk why u thought he had 16gb of ram tho

[elder willow]

computer was 300 😭

[polar flax]

it's bare minimum even in work laptops

[covert lake]

Even my laptop got only 8gb of ram

i3 🔥

[polar flax]

yea there are still macbooks with 8 GB ram

[elder willow]

twins?

[covert lake]

bare minimum for things now day is atleast 8-16gb but most people dont got money

also macbooks with only 8gb of ram is a scam

[covert lake]

real

bare minimum for anything now days is 8gb of ram

much more suggested to have more ram

[covert lake]

i dont have my pc rn but should be 11th gen for me too

[hidden ice]

hello

[polar flax]

mine is also 11th gen

[elder willow]

can i use weights gg to train model

[covert lake]

For training on weights.gg, you can:

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

[elder willow]

u pay to train them?

[covert lake]

actually i got 10th gen 🔥

[covert lake]

nope, weights.gg uses rvc, so i just use the other cloud options of rvc

like google colabs or kaggles

lemme send rq

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

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
  • Applio (ui)
  • Mainline (UI)
  • RVCDISCONNECTED (no ui)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
  • Mainline (UI)
  • Applio by Vidal (UI)
  • Applio by Shirou (UI, no guide as of right now)
• Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
  • Mainline (UI, No guide as of right now)
  • Applio (UI, No guide as of right now)

google colab = easy but risk of losing work for getting disconnected
kaggle = harder but no risk

[elder willow]

but weights gg is free?

[covert lake]

its free only for inference (use model), or for training only if u do streaks or referall code

but nit directly free training

[elder willow]

oh

[covert lake]

I would personally suggest using kaggle mainline

[elder willow]

okay holdon

nvm idk this confusing lemme think

so i can train 30 hours worth?

[covert lake]

yep, 30 hours of free gpu

weekly

[ocean badger]

i cant see anything

[elder willow]

damn i think one week would be enough to do takeoffs entire discogrophy not gonna lie

[covert lake]

huh kinda weird, may need to refresh discord, could be a bad internet issue

[ocean badger]

i dont think so but thanks

[elder willow]

nick

[covert lake]

yea 30 hours are good asf

[elder willow]

can you add me

[covert lake]

btw, u just need 10 mins of audio for training

[elder willow]

oh

i was gonna make one for every year of his career

[covert lake]

You dont need anything over 1 hour, more than that makes no difference

[elder willow]

like run the ai split on every song from that year and cut them together

[covert lake]

yea u could make each different one, but its useless to make a single program with all of them

[elder willow]

wdym

[covert lake]

might be better if u have any issues to ask in #✨│ai-help , im at school ipad rn so could disappear every second

[covert lake]

like, no need to take hours and hours of his voice, even 10 mins are enough, and could make different models for each era

[elder willow]

im honestly thinking now like

one for each year and branched off that different ways he may've sounded

i started from the beginning so its one album and from 2011

im really disappointed tho i wanted to do atleast of this locally

[covert lake]

rip

[covert lake]

yea

[elder willow]

u didnt add me 💔

[feral halo]

Ayo? @elder willow level 7 !!!

[pale girder]

Guys, is the program definitely safe?

[covert lake]

yes, dont watch yt videos tho, the program is Open Source

[covert lake]

#🧬│ai-chat message

[covert lake]

whats ur pc gpu?

[elder willow]

i js mean like add

[chilly lake]

with 4GB vram? way

had they had 32RAM, they could've used shared memory, not ideal, pretty slow, but they only have 4GB ram as well

[polar flax]

he said 4 GB system ram and he attempted to run in cpu mode

[chilly lake]

yeah, they are trying to do rich boy stuff with poor boy equipment

[polar flax]

I thought 16 GB is in average modern laptops/PCs and 8 GB is the lowest one

[quasi moat]

yo guys can someone humanize a document made using chatgpt for me or suggest some tools to do it?

i need it real bad

[covert lake]

there are still low ram stuff unfortunately

[covert lake]

ai detections stuff arent trustable

[untold saddle]

Hi. Is there a guide that explains what each setting in the RVC GUI do?

[covert lake]

RVC GUI is outdated

dont watch yt tuts

whats ur pc gpu?

[untold saddle]

I have a ARM MacBook.

[chilly lake]

MacBook with MPS?

[untold saddle]

It’s RVC Mangio I’m using.

[covert lake]

mangio is outdated too

i mean usuable but would suggest another one

[untold saddle]

An M3 16gb MacBook Air.

[feral halo]

Ayo? @untold saddle level 1 !!!

[covert lake]

Like applio or mainoime

[chilly lake]

Applio has no realtime (yet)

[untold saddle]

Well, I couldn’t get Mangio to work in MacOS, so I have to run it through windows in a virtual OS.

[chilly lake]

Mangio has outdated requirements and some modules do not have them, so it installs a mish-mash of incompatible mix

[untold saddle]

I’m made my own model .pth file online using the Google tool.

[chilly lake]

Applio is up to date, should have no issues installing on Mac

performance-wise MPS it is not particularly good

but somewhat faster than CPU

at least last time I checked

[untold saddle]

Okay.

I can’t train on my Mac can I?

[chilly lake]

you can attempt it

[untold saddle]

Like I said, I did it on the Google Labs(?).

[chilly lake]

Colab

[untold saddle]

Which is better?

[chilly lake]

Kaggle should be faster

[untold saddle]

That’s it.

I did it with 400 epochs.

[polar flax]

-gui

[rare sorrel]

https://cdn.discordapp.com/attachments/1122285248844144733/1203460490475343953/caption.gif?ex=65d12cec&is=65beb7ec&hm=bd2fb8d010006dd7c6e3c1c67d3ae846fd1478e1a3124c544c31b43086fe54aa&

[chilly lake]

as far as I know, Mainline with torch 2.0.0 is slower than Applio with torch 2.3.1

Kaggle > Colab

[untold saddle]

So what should I do for the best results? It’s a bit robotic atm.

[chilly lake]

again if you try a local training on MPS, newer torch is obviously a better choice

both have the same back-end, so your results will be the same in both

quality-wise it depends on the size of your training set and avoiding overtraining

5 minute audio trained for 400 epoch gonna have bad output

[untold saddle]

It’s around 15 to 20 mins of clean audio.

[feral halo]

Ayo? @untold saddle level 2 !!!

[chilly lake]

usually you get good results even before 200 epochs

depends on the variety of the set and other stuff

[untold saddle]

I had no idea I could do local training on a MacBook.

[covert lake]

Wokada fork is better for realtime anyways

[untold saddle]

What should I use for training and then the conversion on a Mac? I’m completely new to this.

[chilly lake]

as I said, try installing Applio

it is up to date

[untold saddle]

Okay. Thanks. And that does both?

[chilly lake]

yes, training, inference, a bit more

[untold saddle]

Okay. Just downloading it now.

[crude willow]

heh

[untold saddle]

Why does it say this: “rm: *.bat: No such file or directory
curl: (7) Failed to connect to raw.githubusercontent.com port 443 after 15 ms: Couldn't connect to server
Creating venv...
Checking if python exists
./run-install.sh: line 40: python: command not found
Please install Python3 or 3.10 manually.”?

When I try and run the install command in Terminal.

[ionic pumice]

skibidi

[chilly lake]

when you run linux install it depetes windows .bat files

and you dont seem to have python 3.10 installed

[untold saddle]

Okay. I have installed python now..

Now I just get this “rm: *.bat: No such file or directory.”

[chilly lake]

ignore it, the previous run did delete them

[untold saddle]

Now I get “Traceback (most recent call last):
File "/Users/****/Downloads/Applio-3.2.6/app.py", line 1, in <module>
import gradio as gr
ModuleNotFoundError: No module named 'gradio'”

[analog charm]

@marsh ferry yo

[wispy widget]

esex

[elder willow]

Hi guys, how can I beat educate myself on private AI companion potential?

[tranquil dew]

I have a question

[chilly lake]

can you just activate the virtual environment and install requirements manually?

[ionic pumice]

No

[untold abyss]

Does anyone know how to remove the robotic sound at the end of words?

[rare sorrel]

💻 (Local) Realtime Voice Changer

• 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

[raven flower]

Hello

[prisma star]

fien

fien

[elder willow]

anyone know of a good TTS or STTS program for Streaming on twitch?

[abstract estuary]

hi

[novel nymph]

No

[ionic pumice]

https://cdn.discordapp.com/attachments/1166674156188864584/1274069638237388851/image0.gif

[abstract estuary]

https://tenor.com/view/cry-cute-gif-10462372808073906726

no

ı am said

[polar flax]

https://tenor.com/view/embed-fail-dissapointed-sad-teary-seal-gif-3391051642056876120

[abstract estuary]

umm

ı-

yea

[feral halo]

Ayo? @abstract estuary level 1 !!!

[abstract estuary]

yea

my first friend

[mental snow]

How to watch voice model? I can't see

[covert lake]

u mean view voice models? try #1175430844685484042

u have the permission to view them but not upload

u needa be a model maker to upload

[mental snow]

I still can't see voice model..

I can see pretrain models but I can't watch voice models

[feral halo]

Ayo? @mental snow level 1 !!!

[mental snow]

...?

[sage citrus]

d

[covert lake]

huh, very weird, could u show me a screenshot maybe in #✨│ai-help if u cant sitll see it?

[mental snow]

I uploaded screenshot

[lost matrix]

☑️

[gloomy gull]

hello

[feral halo]

Ayo? @gloomy gull level 1 !!!

[primal grotto]

-colab

[rare sorrel]

☁️ Google Colabs

• 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, -overtraining, -realtime, -rvc, -help

[primal grotto]

-rvc

[rare sorrel]

Documentation

AI HUB Docs

https://docs.aihub.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

[regal lynx]

ку

[distant dune]

Hello everyone!

[gentle canopy]

hello

[placid sluice]

Hi

[wise sand]

hi

[crisp ingot]

Hi, can you please help me, when I speak I have system sounds coming out of my microphone, how can I solve this problem? I don't have sound monitoring or listening.

[elder willow]

-rvc

[rare sorrel]

Documentation

AI HUB Docs

https://docs.aihub.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

[glacial fable]

e

[wispy widget]

is this still relevant?

[ionic pumice]

probably

[wispy widget]

[plucky nacelle]

Hello dose anyone know a free image to video

That I can use my picture

Hello dose anyone know a free image to video that I can use my picture

[sweet storm]

.

@paper blaze

i wanted the Turkish version

[wispy widget]

assalamu'alaikum

[zinc mesa]

Hello, I am a junior for learning AI.

[cursive forge]

RVC isnt gonna make my pc blow up is it?

[tacit falcon]

it probably will, so dont test your luck

🥶

[cursive forge]

No cus my pc isnt the best and apparently ai requires alot of processing power

[tacit falcon]

shits literally going to set it aflame

[cursive forge]

I cant tell if your serious

[tacit falcon]

no lol

im just joshin

[dreamy heath]

haiii

[covert lake]

whats ur pc gpu?

it wont blow up

just asking if it can run it

[cursive forge]

nvidea gtx 1060

[feral halo]

Ayo? @cursive forge level 1 !!!

[covert lake]

You could be able to only inference (use models) local (on ur pc), not train (make models)

Its just better to use cloud (remote good pc) in ur case

Are you looking for inference or train

[cursive forge]

Both but for now just inference

[covert lake]

for inference ilaria rvc zero or applio colab

[tacit falcon]

can u even use the realtime apps on that?

[covert lake]

For rvc training cloud you can choose between:

• Google Colabs (4 hours of daily gpu for free, not much hours, but easy to use):
  • Applio (ui)
  • Mainline (UI)
  • RVCDISCONNECTED (no ui)
• Kaggles (a bit harder to use and needs phone number but gives 30 hours weekly of better gpus):
  • Mainline (UI)
  • Applio by Vidal (UI)
  • Applio by Shirou (UI, no guide as of right now)
• Lightning.ai (Kinda hard, needs login, no issue with web uis or anything, but only free 15 credits monthly):
  • Mainline (UI, No guide as of right now)
  • Applio (UI, No guide as of right now)

nuh

[covert lake]

idek tbh

[austere eagle]

is rx 7700 xt good for rvc?

[cursive forge]

How do I use models from the voice models chat on this?

[covert lake]

both of the blue texts i sent are hyperlinks, text that u click and it redirects u to the site, which are guides

[cursive forge]

Ight thanks

[covert lake]

what u lookin for? realtime for calls ? ai covers ? training?

[austere eagle]

realtime for calls

[covert lake]

yea

-rt

[rare sorrel]

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

[covert lake]

1st link, wokada fork, which is optimized

[austere eagle]

ah okay

[feral halo]

Ayo? @austere eagle level 2 !!!

[austere eagle]

How about igpu

like amd 780m

[tepid basin]

It'll be very slow probably

Yeah slow

Realtime won't work but you can convert audio files at a reasonable speed

[austere eagle]

[past smelt]

hey everyone,
I've been looking for a completely SFW image generation models,
But most of the "SFW" models are not actually SFW,
Do you suggest any models that are Actually SFW?

[chilly lake]

AMD has its own totally SFW generator

[cursive forge]

do you know where I can find a tutorial for doing it in google colabs?

[chilly lake]

Blackforest FLUX.1 is pretty SWF

[covert lake]

The blue texts are hyperlink, click them and it redirects u to the site

Also I'd suggest HIGHLY to use kaggle

Way better GPU time but harder

[snow sedge]

amd 780m should be fast enough for realtime

[signal cairn]

i dont recommend any integrated gpu because of vram

and AMD cards are worse for realtime because they dont have CUDA

[grim loom]

Hey there

[fallow sparrow]

how do i use the voice model in #1175430844685484042 after i download it

is there some type of software i use to run the voice, and input files into it

[covert lake]

whats ur pc gpu?

[elder willow]

does anyone have a voice model called better mommy i cant seem to find ot anymore

[pale patrol]

wot

[elder willow]

i mean just like i said

[feral halo]

Ayo? @elder willow level 1 !!!

[elder willow]

there was a voice model called bettermommy but its gone now, so im curious of someone still has the files for that one

[past smelt]

a model?
and is it supported by ComfyUI and so?

[feral halo]

Ayo? @past smelt level 1 !!!

[chilly lake]

it is a standalone thing for RDNA3

Amuse 2.0 or something like that

[past smelt]

it's from AMD,
so i dont expect it to support CUDA, right?

[chilly lake]

nope

[elder willow]

i recently had a voice model from #1175430844685484042 called better mommy but my pc shit the bucket and lost the pth file for it but i cant seem to find it anymore in that channel

[chilly lake]

for comfyui you can download flux

[past smelt]

IK, But what model can i get,
i nead it to be REALLY sfw

like REALLY sfw

[grim loom]

Nice

[past smelt]

most of models are posted with tag "sfw",
but they produce the most nsfw images on first prompt with word "girl" in it

[chilly lake]

Flux girl is a pretty girl

[past smelt]

all/most flux models,
or base model?

[chilly lake]

base

[past smelt]

oh, ok

ty

[chilly lake]

of course if you try to force it to show boobs or other parts, it will try to avoid that

like showing an object in front of the most interesting parts

but as long as you dont force it, there wont be any naughty stuff

[past smelt]

that's what i am after tbh

[woeful nimbus]

does anyone know a good tex to speech app for ai? i have the voice i js need it to read a text...

[chilly lake]

the most basic vanilla tts is the free MS Speech engine from Edge browser

[past smelt]

something like Xtts?

[chilly lake]

good reads, at leat in english, no mistakes, no hallucinations like xtts

so you can dump a book into it and it will read it most properly

[past smelt]

as soon as i thought of xtts lol

[chilly lake]

xtts - run same text 100 times, you get 100 different results

some good, some bad, some awful, so it is not a hand-off tool

but when it is good, it is really good

[past smelt]

iam not really in-depth with tts, but it's the latest i saw on "Aitrepreneur" the yt channel

and i thought it was a good one

[woeful nimbus]

could u send me the link in dms i cant find it and i have to help a friend w one of her school project

[chilly lake]

pip install edge_tts

[woeful nimbus]

idk how to ushe that shit

[past smelt]

in cmd

[feral halo]

Ayo? @past smelt level 2 !!!

[past smelt]

or terminal

and the long text in a .py file, and you can run it in cmd

search on youtube of how to run python

is there any good open source Prompt generation webui?

[chilly lake]

install python 3.10, run the command in cmd prompt to get the module, edit tts.txt to add your text, edit the script to use a speaker for your specific language

[untold abyss]

how many epochs should i train my IA model?

[woeful nimbus]

....... bruh lemme cry

[chilly lake]

take python 101 lesson

[covert lake]

you could just use an HF space btw https://huggingface.co/spaces/Nick088/Edge-TTS

[chilly lake]

it is the very basic stuff, but I guess the iPad generation may struggle

[covert lake]

has a ui so its easier to use

[chilly lake]

or that, didn't know there was one

[woeful nimbus]

i dont want to take no classes i js want to help my friend w her project..

[covert lake]

edge tts is an api how will it struggle 😭

[woeful nimbus]

ty babe

but i have my own voice

[chilly lake]

run the output thru RVC

[woeful nimbus]

HUH

yall usoing all those weird terms like bro speak human

[chilly lake]

this u?

[woeful nimbus]

go take a rope

[chilly lake]

😂

[woeful nimbus]

bruh js teach me

[chilly lake]

Nick gave you a link to text-audio

make an audio with a text, then use illaria rvc to change the voice

[woeful nimbus]

bruh

[chilly lake]

a school project is supposed to teach you something, is it not?

well, here's an opportunity to learn something new and magical

you can give this a try: https://huggingface.co/spaces/coqui/xtts just use a text and 30 sec of your own voice recording

how hard is that?

no need to train models or other crap

[woeful nimbus]

but i dont want to learn anything i js want a easy way to make text to speech w an ai voice reading it

[woeful nimbus]

i have a trained model

i want to use it

[chilly lake]

use the link Nick gave

then use https://huggingface.co/spaces/TheStinger/Ilaria_RVC to get that audio converted using your model

[cursive forge]

What does the "model architecutures" mean?

[chilly lake]

v1 or v2?

[cursive forge]

idk

[chilly lake]

v1 is the original model, v2 has added a few more layers

you should be using v2

[tepid basin]

V2 goated

[mild wagon]

hi

[woeful nimbus]

but the accents suck

[chilly lake]

it is your accent my dude

[woeful nimbus]

??

the voice model accents are tripping what r u on ab

[chilly lake]

text -> edge tts -> speech in a chosen language, usually neutral no accent

speech from edge tts -> rvc with your model -> speech with your voice and accent

dont select french speaker for english voice in tts, lol

[woeful nimbus]

its not english

and i dont have the accent choice so

[chilly lake]

what did you use then?

[woeful nimbus]

lmfao an estonian accent

cus i cant find the one i need

[cursive forge]

When trying to combine models, does the epoch matter?

[woeful nimbus]

no

[chilly lake]

where's the accent then?

[woeful nimbus]

as in i had to pick an estonian accent

[chilly lake]

you pick a speaker

[woeful nimbus]

cus i cant find any better ones that fit what i need

[chilly lake]

for the language of the text

[woeful nimbus]

suntem aici în plattoul de filmări platanos alături de regizorul filmului doina rusti ce va inspirat să scrieți această poveste și să o faceți întrun film de cinematograf?

try this without an accent and a fem voice

[chilly lake]

is it estonian language?

[woeful nimbus]

its romanian

[chilly lake]

so why you're talking about estonian?

[woeful nimbus]

BECAUSE THE BEST I COULD DO

that sounded close to it was estonian

i cant find any neutral voice model

who wont sound weird but all of them dont sound right

[chilly lake]

you need to start with an obvious - there's no romanian speaker in the list

why are you picking a random speaker instead of saying 'there is no romanian, what do I do?"

[woeful nimbus]

girllll bye its not my fault there arent neutral voice models who can just mimic the models accent

[chilly lake]

the accents come from the voice model

[woeful nimbus]

i tried with other ones and it sounded completely different like another language so the accents also matter

[chilly lake]

english audio + french speaking voice model = english audio with french accent

so if you want estonian accent, you need to train the model on estonian speaking audio

[woeful nimbus]

GIRL

are u playing dumb rn

[chilly lake]

no u

[tacit falcon]

[chilly lake]

start from the start and explain what you want in details

[languid summit]

https://www.youtube.com/watch?v=oReykFI_yCM look this AI

[woeful nimbus]

no ty

i literally js did

[chilly lake]

explain like i'm 5 years old

[woeful nimbus]

okay go to therapy

[chilly lake]

go back to school 😛

[errant cliff]

how i can download the voice?

[little robin]

What to use to make ai covers? I have nvidia gpu

[feral halo]

Ayo? @little robin level 1 !!!

[vernal abyss]

Does anyone know if there are any reasonably high quality voice models that were distributed along with their dataset? I want to test some training settings but I have no good baselines

[chilly lake]

there's not much magic is making a model that sounds crisp

[vernal abyss]

I know, this is less about searching for "magic" and more just needing a solid baseline to compare against

[chilly lake]

get a reasonably clean audio (not too clean), reasonably long (mine's 45 min)

my source was a podcast downloaded from youtube

good mic, no boomy room echos, no reverb

[vestal crag]

Where do we post Ai-Art?

[tacit falcon]

#🏙│ai-images

[vestal crag]

Well, mine is not stable diffusion

[tacit falcon]

then what

as long as its an image it should probably be in there methinks, idk

[vestal crag]

Alright then

[feral halo]

Ayo? @vestal crag level 1 !!!

[dusky latch]

I remember being here to make an ai cover for a song but I can’t find the ai cover generators

[bold void]

what tools would I use to make a shitty song sung and made by a friend sound halfway decent as a joke

[earnest dragon]

why not perfectly clean

[earnest dragon]

autotune then rvc of his voice ontop

[chilly lake]

seems to be that the perfect is the enemy of good

[earnest dragon]

but the better the audio the better it will work

[chilly lake]

well, it has something to do with what the model is learning and how it is easier to do nothing than something

whatever the explanation for the silence was, I dont recall right now

[earnest dragon]

but what if it's perfect with no silence

[chilly lake]

i have a model trained using the default noisy pretrain, the source audio has a faint fan noise in the beckground, yet it sounds crisp af

so magic, I guess

here's an example of audio quality that is just not good at all: https://www.youtube.com/watch?v=1fA8JUf_swU

terrible microphone

[earnest dragon]

but what if no faint background sound and also no subsonic or ultrasonic frequencies.

[chilly lake]

in my understanding the model would learn to produce pure silence instead of something as complex as speech

with a faint noise the model is more adaptable to be used in real world scenarios where a perfect silence is rarely an option

it the output of such inference may be weird

so if the audios you infer have perfect silence, then sure, use the model trained with perfect silence

[livid bear]

There is anyone really good at IA making here ?
I need one Image to my RPG, its a male character, I want to mescle 2 imagens to have this character

[floral pulsar]

hii i have a cuestion i can use this voices to talk with my friends on discord?

[earnest dragon]

Yes

[earnest dragon]

-rt

[rare sorrel]

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

[elder willow]

@tepid basin scammer there

[tepid basin]

Ty

[polar flax]

I have also got rid of his messages in other chans

[tepid basin]

[blissful turret]

'HParams' object has no attribute 'data'

[chilly lake]

HParams contains settings from config .json

so I guess it is not loaded/missing ?

[scenic raven]

hi everyone glads to be here

[golden blade]

SIGMA

@tepid basin 👍

[strange mesa]

What happened to the sticker version of this emoji?

[untold violet]

how safe are these models

[vernal abyss]

horribly dangerous. viruses everywhere. watch out!

[earnest dragon]

perfectly fine and anyone that says otherwise doesn't know what they're talking about

[frosty bluff]

Be careful of the one with too many likes, could be riddled with viruses.

[proven python]

-rt

[rare sorrel]

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

[dapper drum]

Cxsmo you going to get the 5090 when it’s released

[young hound]

do you guys know how to fix a sensitivity issue, where the voice changer is picking up sounds like my chair squeaking?

[covert lake]

they are safe

[covert lake]

they are all safe

[supple dew]

this is for up coming projects https://docs.google.com/forms/d/e/1FAIpQLSdGzt2lun5dr2NyUkEW4Q7ApS_ZJBQb-7vuRotUQMIRC0qUuQ/viewform?usp=sf_link

[earnest dragon]

probably not but next gpu for me will be a 5090

[rose flume]

hi

[dapper drum]

I wonder how much faster the training process will be with a 5090

[rose flume]

training?

[dapper drum]

Yeah, for training a model

[rose flume]

#1159949278270193734

[dapper drum]

Oh, I think you misunderstand me. I was saying I wonder how faster it would be on the 5090 with training a AI model

[rose flume]

😄 i just joined i dont understand anything here. apologies

an AI model* sorry forgive me

can anybody online give me a walkthrought whats goin on here

[feral halo]

Ayo? @rose flume level 1 !!!

[earnest dragon]

if it's twice as fast everything then it would be twice as fast but if not then if will be whatever percentage faster.

[earnest dragon]

Ai rvc voice models

[rose flume]

rvc? remote voice control?

can i make William Dafoe say : You'd better run.... BATMAN!!!!!!!!!!!

[earnest dragon]

no stands for retrieval based voice cloning

[rose flume]

so can take a sample of a voice?

what kind of things do you do with it?

[earnest dragon]

you take a trained voice model and use it on clean audio that you want the person to say or sing and then it will sound like them

[rose flume]

do you have anything ready to share?

[earnest dragon]

-hf second link

[rare sorrel]

Suggestions for @rose flume

<:huggingface:1179800228946268270> Hugginface Spaces

• UVR5 UI, by Eddy and Ilaria Huggingface Spaces
• Ilaria RVC Zero, by thestingerx Huggingface Spaces
• RVC⚡ZERO, by r3gm Huggingface Spaces
• Applio, by IA Hispano Huggingface Spaces
• 🆕 FaceFusion UI, by Nick088 Huggingface Spaces

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

[rose flume]

if it's trouble then do not, but i'd like to listen to whatever your working on

nobody has anything to share/ show-off?

[chilly lake]

@earnest dragon btw, speaking of noisy audio, inference of a noisy audio can actually clean it up and remove the noise

I ran the original audio thru UVR just to see what kind of noise there was

none of that made into the output

[mental vapor]

обезьяна

[visual hornet]

anyone want to play asia server dm me

[feral halo]

Ayo? @visual hornet level 1 !!!

[paper yoke]

Hello, we need your help guys, please help us to make our voices heard from your social media accounts, Turkish women need to live, we are trying to make our voices heard in all applications such as Twitter instagram with the hashtag #TurkishWomenNeedHelp , please be one of us, put yourself in our place and help us 🙏🏻

[dusty glacier]

You mean 𝕏??

[feral halo]

Ayo? @paper yoke level 1 !!!

[calm sequoia]

Is anyone using root Stable diffusion around here ?

[wise quiver]

what does bro think this is

[ionic pumice]

lmao

[patent sigil]

no

[elder willow]

How do you make holy water?

What keeps pinging me ☹️

[mystic solar]

guys are there available Laswell's RVC model from call of duty modern warfare 2022-23?

[pallid token]

hi

[scenic folio]

Does anyone know the new URL for https://aihub.wtf?

?

[covert lake]

Check #📰│dev-updates message

[covert lake]

You can search rvc ai voice model at:

• #1175430844685484042
• #🔍│find-models
• https://weights.gg/ (login required)
• https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice models)
• https://applio.org/models
• https://voice-models.com/
• https://thevoicemodels.com/ (for Turkish Models, login required with discord and level 2 on their server)

if there isnt one, you can:

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

[scenic folio]

thanks!

[high eagle]

NAME
main.exe

SYNOPSIS
main.exe COMMAND

COMMANDS
COMMAND is one of the following:

 module_download

 cui

 rest_client

 convert_file

D:\chenger\dist\main>

[clever verge]

Anyone have a good website to download youtube videos / youtube audio?

[feral halo]

Ayo? @clever verge level 3 !!!

[alpine fulcrum]

I'm doing homework and listening this masterpiece rn https://www.youtube.com/watch?v=II47GsODBSo

[alpine fulcrum]

https://www.tubeninja.net/welcome

[steady jungle]

hiiii

[steady jungle]

y2meta?

[covert lake]

#1159290752195633273

[wispy kite]

Hi guys

How can i dowloand applio? (I don't know very much)
I'm glad if you're helping me.
I'm sorry if I make you busy

[covert lake]

What's ur pc gpu?

You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU

[wispy kite]

I dont have a pc, I didn't know this was required to download applio.
But ty for helping

[feral halo]

Ayo? @wispy kite level 1 !!!

[ionic pumice]

there's cloud services you could use if you're planning to make ai covers on mobile

-hf

[rare sorrel]

<:huggingface:1179800228946268270> Hugginface Spaces

• UVR5 UI, by Eddy and Ilaria Huggingface Spaces
• Ilaria RVC Zero, by thestingerx Huggingface Spaces
• RVC⚡ZERO, by r3gm Huggingface Spaces
• Applio, by IA Hispano Huggingface Spaces
• 🆕 FaceFusion UI, by Nick088 Huggingface Spaces

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

[wispy kite]

Tysm i try it after👍

Wow very nice

[hexed elk]

дло

[pale stratus]

yo is there any way of using rvc with an amd gpu

[chilly lake]

directml build, or nvidia build with zluda patch

[pale stratus]

are you able to link me the directml build

[dreamy heath]

hiiiiiiiiiii

[wispy kite]

You're welcome💫

Thank you for ur suggesstion

[wispy kite]

Hiiii sonunda bi Türk buldum

[delicate frigate]

I just arrived here, and I'm a little scared of downloading my first AI voice model.

[vocal monolith]

o7

[polar flax]

Applio with Zluda https://docs.applio.org/getting-started/installation#amd-gpu-support-windows

[delicate frigate]

I don't know if it's safe to download, because I'm a bit paranoid about downloading things without knowing if it's safe.

[tulip ravine]

Is there any AI website or program that is like a TTS but you can upload audios so it try to follow the rhytm? (like old uberduck ai tts used to do)

[covert lake]

the programs are safe

Be sure to be following our tuts and not some random youtube tutorial

First of all, whats ur pc gpu?

You can check your pc gpu via:
ctrl+shift+esc (task manager) -> Performance tab -> GPU

[delicate frigate]

My GPU is NVIDIA GeForce 940MX. Is it any good?

[covert lake]

Nope, not good for local (on ur pc)

You can use cloud tho (remote good pc)

What are you looking for?

• make models
• use models for pre-recorded audios
• use models realtime for calls/games

[delicate frigate]

The 3rd one

[feral halo]

Ayo? @delicate frigate level 1 !!!

[clever verge]

Anyone have good ways to remove intrumentals from vocals thats free?

[covert lake]

For Realtime Voice Changing for Calls on Cloud (remote good pc for those who don't have a good one, YOU CANT DO THIS ON MOBILE):

• Google Colabs (4 hours daily of free T4 gpu, easy to use, require only a google account) :
• How to use Original W-Okada's Voice Changer Google Colab
• Modified W-Okada's Voice Changer Google Colab
• Kaggles (30 hours weekly of better GPUs, T4x2 & P100, harder to use, requires an account and a phone number)
• Original W-Okada's Voice Changer Kaggle
• Modified W-Okada's Voice Changer Kaggle

Id suggest kaggle for way more hours

[delicate frigate]

I'll try Google Colabs. Thanks!

[covert lake]

I remember u said u got an rtx 3060 12gb, so you can do it locally, use UVR (Ultimate Vocal Remover): https://docs.ai-hub.wtf/rvc/resources/vocal-isolation/#local-uvr

[covert lake]

id suggest kaggle, but yw

[clever verge]

Thanks so much!

[covert lake]

yw

[clever verge]

Is there for the AI Training Folder, a certain amount of training i should use? Or is it one of those things where the more i have the better it sounds?

I have a data set of around maybe 3 hours but i think it may be to little so im getting more is that bad?

[chilly lake]

3 hours may be a bit too much, quality and variety of content > quantity

[clever verge]

Gotcha so i should narrow it down to bits that are clear then?

[chilly lake]

good clean-ish audio (may have some background noise), preferably good quality mic, no reverb, no echo, no boomy room

[clever verge]

noted

[chilly lake]

the model inherits the general feel of the audio, if someone records with a shitty mic and bad quality, all results gonna have the same shitty quality

[clever verge]

Im currently spliting vocals and background noise of everything then ima manually edit out bad parts and stick with the soilds in Audicity.

Im training for a "Documentary Voice" for a personal project so i downloaded a few documentarys that have good voices and going through them.

[chilly lake]

but if you infer a shitty audio using a good model, the result is generally good

[tulip ravine]

?

[covert lake]

even 10 mins are good for an rvc model training, over than an hour is not needed

[clever verge]

Noted

[chilly lake]

it will be worse than say 30 minutes, and 30 min will be lil worse than 60, but an extra quality from more than an hour probably not worth the time spent training

lets jus say, even 5 min will be generally recognizable as a specific person speaking, but very poorly

similarly with the duration of the training - the model captures most features during initial 10-20 epochs, after that it is just small improvements in quality

[glad nebula]

10 minutes sound unnatural vs 30 minutes and above

[covert lake]

what matters the most is quality than quantity

[vernal abyss]

though it is significantly, significantly easier to gather a large volume of good enough data than a small volume of nearly perfect data

[polar flax]

you should remove reverb and other post-processing effects/distortions/artifacts, if you can't remove it like e.g. poor mp3 compression (don't even expect Apollo perfectly restore it) and spectrogram cutoff below 16khz, exclude those parts and find other good sources

[vernal abyss]

i am kind of curious about this, though. i gathered about 15 mins of voice data and trained it for 100 epochs and it's definitely not done. some of the fundamental aspects are definitely there, but they're not mapped well enough to accurately recreate the trained voice

even if the features are initially picked up quickly, it still seems that you have to train a fairly significant amount of time to actually map them out properly

[gentle canopy]

hi

[chilly lake]

see the tensorboard's charts

the usual - fm, mel, kl, total d and g

if it is a language vastly different from pretrain, then you may need more data

[feral halo]

Ayo? @chilly lake level 16 !!!

[polar flax]

other than not enough epochs, it might be due to inconsistency, i.e. different sources, different EQ profile, roleplaying other voices, etc.

[patent arrow]

Hallo! im a new user with no GPU and no idea what im doing does anybody know of any non GPU voice changers that i can use through call?

[vernal abyss]

the roleplaying other voices thing is probably a part of it. i'm sure it's just not enough epochs

though the loss was starting to rise toward the end, which could be on account of the lr schedule or just actual divergence

tensorboard splits everything into like d_g/1, d_g/2, d_g/3 etc and i'm not quite sure what that means

[polar flax]

the voice would blend, so you shouldn't

[vernal abyss]

i'll cut that section out of the audio to be safe

[feral halo]

Ayo? @vernal abyss level 4 !!!

[vernal abyss]

how about excited utterances / gasps/ screams etc? remove them too?

i'm using a rip of some game character voice lines, so i've got everything rn including non-dialogue

[tiny pewter]

what is the most reliable realtime voice changer at the moment?

w-okada is a virus according to my pc

pls help

[feral halo]

Ayo? @tiny pewter level 1 !!!

[ripe iron]

Ilaria rvc doesn't work anymore ?

[tiny pewter]

can somebody help me in dms?

[clever verge]

What is a epoch and how important is it?

[chilly lake]

epoch is one cycle of training

a model has to go thru multiple cycles in order to reliably reproduce the desired result

[clever verge]

Whats a good amout of EPoch?

[chilly lake]

it depends on the data, batch size, etc

go for as long as there's an improvement

[austere eagle]

What's the best budget card (450$) for realtime voice-changing

[earnest dragon]

depends used or new