🧬│ai-chat

[covert lake]

Learn to behave in the server, don't insult randomly people or actions will be taken.

It's English only, you read that in one of the First server guide rules

@elfin prism

[elfin prism]

ok

[solar torrent]

Don't try to fight someone because it's your habit. Try to be respect here, otherwise the issue is on you.

[half blaze]

In support channel

😎

[lament locust]

i want learn making agents

i want make whats app agent for my school

any one help me

[supple stream]

AI agents?

[lament locust]

yes

[lament locust]

can you help me?

[supple stream]

AI agents are no little thing

[lament locust]

how?

explain me

[lament locust]

what do you do?

[lament locust]

ok can you help me lerning ai engineer

[half blaze]

How hard is it to learn Ai

You basically just throwing info at it

[kind trench]

)

hello all

[covert lake]

Hello!

[kind trench]

)

Coli send SMS....

infinity

2000 years

ahahaha

[kind trench]

wow Admin....

[covert lake]

Huh?

[covert lake]

Junior admin yeah

[kind trench]

active!

nice bro)

[covert lake]

[kind trench]

Hint Admin, is there a role for a musical and performer, or composer here? If there is, how do I get it? My tracks are on the site in the description.

[supple stream]

An Ai model is made like these steps:

• ML
  in ML Deep Learning; Where Engineers design complex neutral networks inspired from the human brain.
  Then this system is trained
  with vast data.

[supple stream]

what's up buddy

[kind trench]

nothing)

[covert lake]

There was once a while back, but not anymore since barely 2 people had that role, wasn't really useful

[kind trench]

sorry

[supple stream]

nick the Goated admin

🔥🔥🔥

[kind trench]

ок

[supple stream]

I’m so lazy

[kind trench]

What do you think of this idea, colors through a bot? And each role has its own additional name? #🏙│ai-images

[covert lake]

it's better to use #1159516963014451302 pls

[grand notch]

for anyone that uses mvsep

how the fuck do i extract harmonies

off the vocals

[half blaze]

Is it as complex as it sounds ?

[supple stream]

more than it sounds

[grand notch]

or the new UVR colab

either work

i just gotta extract harmonies

💔

[half blaze]

Is not plug and play like the movies 😔

[supple stream]

if you’re facing issues you can check support channels #✨│ai-help #1192011222023950368

to seek help

[kind trench]

[chilly lake]

asking a bunch of 15-year olds about AI financial advisors... not very intelligent @normal flame

[covert lake]

promos aren't allowed

[normal flame]

I thought we got working professionals here lol my bad.

[chilly lake]

out of 500k users, there may be 5

4 of them would tell you the smartest thing is own is a dumbphone

[craggy ice]

incredibly hard

[half blaze]

But it’s learnable

[craggy ice]

well yeah?

oh

fell for the ragebait

lame

[supple stream]

fr

[chilly lake]

what moron would trust a financial AI advisor even if its makers have no idea how it makes decisions?

[naive quest]

Yo can someone tell me how to use voice.ai? Every time i open it a stupid pop up says buy ultimate even though the website said it was free

[chilly lake]

you dont

[naive quest]

So its not free?

[chilly lake]

it is trash

[naive quest]

Im just trying to use it to use the ai models i downloaded

[chilly lake]

most of these paid services use the same rvc behind the scenes

[naive quest]

Do you know any other free voice apps that let me use the ai models i downloaded?

[chilly lake]

depends on the models

from this discord?

[naive quest]

Yes

[chilly lake]

applio, the old rvc mainline, bunch of other clones

[naive quest]

Aight tysm

[chilly lake]

even if you dont have a good GPU, you can still run inference on CPU

that's for speech to speech, not realtime voice changing

for realtime there are other options

[naive quest]

What would you use for realtime?

[paper flare]

um

what is rot 47 doing in applio non ui colab

[chilly lake]

https://rentry.co/forkvoicechangerguide

[chilly lake]

masking the repo so google does not ban it

[paper flare]

how does it work

👀

i doubt it skips the ipynb if its content contains rot_47

[chilly lake]

technically in the most up to date colab there should be none of that

must be some leftovers that needs to be removed

[paper flare]

https://colab.research.google.com/github/IAHispano/Applio/blob/main/assets/Applio_NoUI.ipynb

[chilly lake]

yeah, just old leftovers

[paper flare]

its never used

is there new version?

[chilly lake]

it was more elaborate before, but now I guess we said fuck it

[paper flare]

hmm

!uv

lot changed

[chilly lake]

UI colab still has some

"uggcf://tvguho.pbz/VNUvfcnab/Nccyvb/"

[paper flare]

oh its faster pypi

[paper flare]

👀

[covert lake]

noobies sent u the wokada deiteris fork program, btw be sure to check ur pc gpu first and to use support channels

[olive otter]

just recently got the idea for making the ai sleep
so, the idea would be that it functions to real sleep
it would summarize the chats of the day and store them in a long term memory db
and clear cache, rebuild db, clean-up, fix bugs, write tests
reboot
I'm alsso thinking about having it self-evolve
with evolutionary algurithms
introduce some randomness and mutations
and let it update itself during sleep
and then, I would like it to dream
imagine it like the ai's hallucincations let lose
it would be interesting if you could somehow visualize what the ai "dreams"
google's deep dream comes to mind.
https://www.youtube.com/watch?v=SCE-QeDfXtA

[sterile stream]

Should i move out from Chat Gpt plus to an agent?

[indigo ridge]

Is it just me or does the ai voice for the TF2 engineer have the ability to sing anything almost perfectly

I got him to sing Josiah Queen and it sounds insanely good

[dapper drum]

was the best way to remove reverb

[indigo ridge]

I didn't tinker too much with the advanced settings yet

So idk

I'm just starting messing around with this

[covert lake]

https://docs.aihub.gg

[naive quest]

Dang this is really confusing

[chilly lake]

what's your GPU?

[naive quest]

Nvidia Geforce GTX 1070

[naive quest]

My cpu is a intel(R) Core i7-8700

[chilly lake]

you can download nvidia build

[naive quest]

Is that all i do?

[naive quest]

?

[covert lake]

you gotta install vac lite and the wokada deiteris fork nvidia windows version, carefully read the guide and use #1192011222023950368 for issues

and yeah, video tutorials are outdated

[naive quest]

Mk

[deep isle]

Hi I'm trying to use the backup of a training I was doing but I do not know how to do it, or directly if i can, i was so close but it just cuted 😭, anyone knows how to do it?

if something is misspelled, forgive me, I am using a translator.

I'm talking about applio and stuff

[hearty lily]

I have a 4 minute dataset of pure clear audio.
What's the recommended epoch with the original pre trained models?

[craggy ice]

https://tenor.com/view/dexter-dexter-morgan-morgan-morgan-dexter-dexter-ewww-gif-10083604162228652049

[naive quest]

when i get to the point where it says mmvc do i download all three of them?

[covert lake]

btw pls suggest https://docs.aihub.gg/w-okada/local/w-okada instead

[covert lake]

replied to u in #✨│ai-help

[covert lake]

use support channels

[covert lake]

use support channels

[covert lake]

https://tenor.com/view/yoshaverse-gif-13020786377069990334

[vagrant storm]

hello

[silver bronze]

anyone know of any good deep male voice models other than corpse?

[desert lion]

Dream's voice maybe?

But it's not that deep

[silver bronze]

its not a bad idea, maybe there are some deep voice v tuber dudes?

[desert lion]

Hmm I don't know about that

[polar flax]

havent ever seen old man vtubers

but if I do, it'd be like duckus

[silver bronze]

duckus is a Vtuber??

i thought he just did ai voice trolling

[polar flax]

he did, but the ones using their real voice are just regular youtubers

[magic moon]

@uncut vigil get to work n check dms or ur fired

[uncut vigil]

.3.

https://tenor.com/view/laugh-leave-bloop-gtfo-car-gif-5678489

[worthy coyote]

NEW BEST! 99%

[astral frost]

yo guys

anyone there?

[solar torrent]

What are you looking for? You seem concerned.

[remote grotto]

yes

[supple stream]

hello

[paper flare]

when does overtraining occur?

[gray rover]

https://docs.aihub.gg/rvc/resources/training/#epochs--overtraining
https://docs.aihub.gg/rvc/resources/training/#tensorboard
@paper flare

also this thing I've made ages ago ( should be mostly fine. - just ignore any mentions of " collapses " or such, it is no more valid. )

[red estuary]

can someone me a ai voice changer of a sexy grandma

[solar torrent]

No. For W-Okada the realtime voice changer, go to #✨│ai-help or #1192011222023950368.

[paper flare]

weird, import_google_drive_backup seems to occur only once as well

[wispy frigate]

Is there anyone who can recommend a male voice that really has a reasonable tone? We are women who want a male voice. We are from Thailand.

[magic juniper]

Are you sure you are not one gay man from Thailand? P.S. Marigold

[desert lion]

Cab someone suggest me

A voice changer to girl

In mobile phone tho

[wispy frigate]

No, it's a woman. I'll take a man's voice.

[desert lion]

No I want woman's voice

[covert lake]

not possible on phones

[desert lion]

Wait really 💔 💔

[queen kernel]

If you can root your phone, you can install xiaomi game space or any other game space which contains voice changer, but it's not that good as w-okada

[desert lion]

I will try

As long as it works its fine

[queen kernel]

Rooting your phone is risky

I suggest you to first check the outputs before doing anything

Or you can buy membership of any voice changer app.

[primal nexus]

Hi could you tell me what w-okada fork this is?

[polar flax]

imagine you can't play blue archive on a rooted system

[queen kernel]

Actually you can

[polar flax]

perhaps in the past, when there were negative reviews saying they couldn't run it

[queen kernel]

Fork is the modified version. And w-okada is a real-time voice changer

[queen kernel]

Maybe you can hide root by shamiko and fix play integrity. Also you can add it to denylist.. It should work

[desert lion]

So w-okada is voice changer

?

Only for pc ig

[solar torrent]

[drowsy condor]

Hello everyone! How can I make a song generation with an AI model voice without using Google Collab or elevenlabs? Are there any programs based on Voice Changer Client? I'm just a beginner!

[covert lake]

Use support channels and elaborate ur pc gpu and what u want to do

[solar torrent]

When you say you're a beginner, but you ask in a wrong channel. Also, RVC and realtime voice changer are two different programs of different purposes. For RVC, go to #✨│ai-help.

[jaunty heart]

Yo Zazu

[queen kernel]

Yes. It's a real-time voice changer

For pc Only

[desert lion]

Oh thank you

Last question pls

Where can I download it

[covert lake]

if u want a realtime voice changer, tell ur pc gpu in #✨│ai-help or #1192011222023950368

[desert lion]

I will do it tomorrow I guess

I am sleepy now

[kindred kelp]

I'm working on a JavaScript operating system that dynamically loads drivers and generates them using openai API so my OS will work on all hardware without issues

[arctic flame]

hi guys

can i ask a help bc i want to get a voice changer but it doesnt work

[supple stream]

seek help from support channels #1192011222023950368 #✨│ai-help

[kindred kelp]

Thz

[queen kernel]

Ask in #✨│ai-help

[arctic flame]

ok ty

[gusty ingot]

1

[undone sentinel]

So i’m new to Discord and have no idea what i’m doing here lol I do AI music. Is this the right place for that or should I be in another chat? I’m just trying to connect with other AI music creators.

[vagrant viper]

I'm an AI Music creator. I use Riffusion AI.

I just signed up here but I have been doing AI music for a bit.

Basically, I use my own singing voice and my own lyrics and feed it into the AI

Its pretty cool actually 🙂

[undone sentinel]

Same! I started out using Donna but now I use Suno. Both are really good! I’m not very familiar with Riffusion AI but i’ll check it out for sure!

[vagrant viper]

I'll DM you and show you what I do if you want 🙂

I used to use Suno but I moved over to Riffusion AI. I personally like it better. You can do a lot more with acapellas and different sounds.

[undone sentinel]

Sweet! I’ll share what I have after you send me your work. I mostly do emotional, numetal, post hardcore music but I like to try new things as well.

[vagrant viper]

That's awesome its always great to meet new music artists and share music 😄

[river verge]

DIlly ding, dilly dong! Two new RegalHyperus drum models just released!
AFC Champions League Elite Anthem & Shut the Fack Up (Drum models no. 600 & 601)

[undone sentinel]

Absolutely! Im currently working on my next project now. I’m about to send you a link to the music video I just released.

[vagrant viper]

Nice I can't wait to see it. My friend does videos in addition to the music so that is pretty cool. I'm not great with videos.

That is REALLY good man nicely done! 😄

[elder willow]

does someone has free

guy model

[undone sentinel]

Thanks! That means a lot.

[vagrant viper]

Of course and thank you for listening to mine as well 😄

[undone sentinel]

Sorry, it didn’t notify me that I had a message😅 I just listened to it and sounds pretty good dude! How many tracks do you have?

[half blaze]

I wanna learn how to make music like that but too busy studying right now

[vagrant viper]

46 songs

[vagrant viper]

Well if you ever need help I'm more than willing to help you. I know pretty much the ins and outs of Riffusion.

[spice pivot]

Hello peeps

[vagrant viper]

Heyy bud welcome 😄

I'm just talking to Punisher and Central about AI music

You know @half blaze and @undone sentinel you two should really check out Riffusion. It has a really nice free end to it where you get unlimited generations in relax mode.

[spice pivot]

Thats cool. Lots of fun and a great creative outlet

[half blaze]

I added you as friend 😎

[half blaze]

Am going to give it a go tmrw 😄

[spice pivot]

Yeah riffusion is awesome

Im doing more on Riff than Suno atm

but i ahve about 10 there on public most of my stuff is private Riff Iam starting to do more public but havent been there but a few months

[vagrant viper]

Yeah on Riffusion the vibes feature alone is worth its weight in gold.

[undone sentinel]

Hey everyone, i’m new here so it’s nice to meet everyone😁

[spice pivot]

hello. New here myself

[vagrant viper]

Me too

[half blaze]

Welcome 😎

[spice pivot]

Riffusionis awesome. Very happy with it

[vagrant viper]

@undone sentinel I also sent you a DM with the community link for Riffusion. I hope you guys enjoy using it as much as we do. Video and Music goes hand in hand so I think maybe even some video software can be used in conjunction with Riffusion. That is what it seems like they are wanting to do.

[paper flare]

Backup Complete: 3 new, 0 updated, 0 deleted.
 42% 39/93 [00:57<00:57,  1.06s/it]Files are up to date.
 61% 57/93 [01:13<00:32,  1.11it/s]Backup Complete: 45 new, 1 updated, 0 deleted.
Files are up to date.
 80% 74/93 [01:29<00:16,  1.16it/s]Backup Complete: 6 new, 0 updated, 0 deleted.
 81% 75/93 [01:30<00:15,  1.15it/s]Files are up to date.
testModel | epoch=1 | step=93 | time=22:58:07 | training_speed=0:01:49 | Number of epochs remaining for overtraining: g/total: 50 d/total: 100 | smoothed_loss_gen=0.000 | smoothed_loss_disc=0.000
 12% 11/93 [00:11<01:18,  1.04it/s]Backup Complete: 0 new, 1 updated, 0 deleted.
 13% 12/93 [00:12<01:15,  1.07it/s]Files are up to date.
 67% 62/93 [00:57<00:25,  1.20it/s]Backup Complete: 0 new, 1 updated, 0 deleted.
 68% 63/93 [00:58<00:25,  1.20it/s]Files are up to date.
New best epoch 2 with smoothed loss_g 26.584 and loss_d 4.108
testModel | epoch=2 | step=186 | time=22:59:33 | training_speed=0:01:26 | lowest_value=26.584 (epoch 2 and step 141) | Number of epochs remaining for overtraining: g/total: 50 d/total: 100 | smoothed_loss_gen=26.584 | smoothed_loss_disc=4.108
Saved model '/content/Applio/logs/testModel/testModel_2e_186s_best_epoch.pth' (epoch 2 and step 186)
  2% 2/93 [00:02<01:48,  1.19s/it]Backup Complete: 1 new, 1 updated, 0 deleted.
Files are up to date.

should i disable backup? unlike in kitlemonfoot notebook (45s), applio non ui 1 epoch takes above 1 minute

[undone sentinel]

I’m about to try it out now! Super stoked😁 Thanks for the link! I’ll send you a DM of the track I come up with.

[spice pivot]

sweet glad to have you man

[vagrant viper]

Can't wait till you drop in and try out the product 😄 If you need any help feel free to ask.

[spice pivot]

definitely we are happy to help anyone if we are available.

[gray rover]

bruh " number of epochs remaining for overtraining "
I wonder who was sane enough to even consider it

[glad nebula]

peak 2023 rvc

[gray rover]

https://tenor.com/view/one-peak-one-piece-one-piece-cry-its-so-peak-gif-13413926760013023520

[gray rover]

well, that was actually on point

[glad nebula]

rvc-boss bilibili tutorial also does that and his model sounded ok

i never tried that rvc

but i kinda like how it sounds based in boss videos

[gray rover]

huh

[glad nebula]

https://www.bilibili.com/video/BV1pm4y1z7Gm/

minute 7

[gray rover]

oh yeah, I know that one

[glad nebula]

pitch tracking is worse than today pitch tracking but

it def sounds more natural than what we have today

[gray rover]

welllllllll idk

I always thought it sounds too flat to me

[glad nebula]

sounds like a model without guidance

imagine that model but with appropiate pitch estimation

[gray rover]

also, it's 40khz

those were always easier to train and more forgiving than 48s

[glad nebula]

tru

[polar flax]

perhaps I tried that similar thing on some lossy enhanced audio

it can sound less robotic despite the amount of noise

[glad nebula]

i kinda like old rvc

[gray rover]

nostalgia really be kicking in

[glad nebula]

100% lol

[gray rover]

but then, there was something nice about it back in the day

so novel, so " technical " when we all were still learning

[glad nebula]

it felt so magical to me

[gray rover]

at times even to remember the step / processing order lol

man

[glad nebula]

[gray rover]

[glad nebula]

idk how chinese find raw 40k audio on the internet

looking at chinese posts in rvc's github almost everyone has 40k audio

chinese rvc tutorials also mostly use 40k audio

while in this side of the world almost every model is 32k

[gray rover]

tbf.. asian part of the internet is an unknown world to us

I swear they must have so much of practical and usable content, data that we don't

all those huge libraries but gate-kept behind language barrier

[glad nebula]

[gray rover]

https://tenor.com/view/zs-good-morning-gif-4154880375153228977

[glad nebula]

they seem to prefer mainline realtime over w-okada

maybe they found that it's faster

but every tutorial is using voicemeeter

[gray rover]

could be, also it's easier to open up from the inside out

[glad nebula]

you know that boss never uploaded the new compiled rvc version in his huggingface repo?

it's locked behind chinese only dl sites

[gray rover]

the " new failed " one?

one that ' didn't meet the expectations ' or

[glad nebula]

2.2.231006

its a realtime update

prob the same fumiama has

[gray rover]

x.x.231006

and yet

just in a form of code

[glad nebula]

lmao

[gray rover]

but then, ig, he just got bored or lost motivation for rvc

[polar flax]

[gray rover]

oh, well

the phase vocoder and most of the other things

is rt focused

[glad nebula]

yuh

[gray rover]

yea.. so that's that

[glad nebula]

poggers

well at least they did made realtime faster

thanks rvc boss

🙏

[gray rover]

🙏

[glad nebula]

tbh the only thing that i want to be fixed in rvc is the ability to handle vocal fry, because god, rvc handles those very poorly

[polar flax]

I bet it's for Nvidia gpus, so fork wokada is still recommended for AMD users

[gray rover]

well, for that to work " properly " to the point I can say confidently " yes "
there'd have to be 2 things.

1. vocal fry elements in the dataset because it is kind of considered as ' extra '
   ( 1.1. potentially, pretrain exposed to such extras and more )
2. f0 extractor with very high resolution ( or rather, with very small and accurate window

[glad nebula]

its only for nvidia yee

the original pretrain has vocal fry i guess

[gray rover]

it does have a tiny tiny bit of it but not by design

yet due to that awful " vocal fry " influence on in a lot of english speakers' accents

( don't mind me. I despise that personally )

in other words, some female voices in vctk do speak with vocal-fry manner, so I suppose, some slight or tiny exposition to it is already there

but whether f0 confidently picked it up or nah and to what degree? I can't tell for sure

[glad nebula]

points at p225

[gray rover]

👀

lol

[gray rover]

ah ye

that one

[glad nebula]

aka the voice of the og pretrain

time to do a model of her

let me name it...
f032k

[gray rover]

lmao

[glad nebula]

glad there are some good findings lately

because god that refinegan nonsense

lasted for so long

just to reduce an artifact u cannot even hear

🙏

i hope in the future f0 estimation stuff can be improved as well

[gray rover]

yup

[glad nebula]

[night lake]

https://tenor.com/view/cat-cute-cat-cat-dance-cute-cat-dance-ai-cat-gif-15064057413633942575

[glad nebula]

https://tenor.com/view/cat-hand-gif-23447559

[silver bronze]

do you guys know if there was an update for the modded rvc program

like this one.

i was gonna send a pic but dont have perms

its the one that shows the ms graph

its been acting up lately so i was curious to see if there has been an update

[gray rover]

You'd actually want to go to #✨│ai-help and ask there

[glad nebula]

btw hybrid conversion rmvpe+fcpe doesnt work in your fork in the latest update

[spark pond]

[gray rover]

oh, that's weird

I did use it today just fine

at least 2 samples in there are on hybrid

[glad nebula]

tried other audio and i got the same message

[gray rover]

well shape being 2

how long is the audio you use

[glad nebula]

20 seconds

[gray rover]

huh

[glad nebula]

[gray rover]

models from other rvc / fork?

or one made on current one

[glad nebula]

it's from your fork

i trained that using fumi-fcpe

[gray rover]

turn off split

and retry

btw

[glad nebula]

same

[gray rover]

hmmm

[polar flax]

perhaps the input audio being either stereo/mono/multichannel?

[glad nebula]

happens with every audio

[gray rover]

lemme check, wait a sec

[glad nebula]

[polar flax]

same error or different dimension?

[gray rover]

O I think I see the issue

[polar flax]

[lone haven]

lf help

to set up real time voice changer

lowk confusing i need someone thats alr done it

and i cant find a video on youtube

[polar flax]

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

go to #✨│ai-help

[gray rover]

we'll do a lil test @glad nebula gimme a min

[glad nebula]

[polar flax]

[gray rover]

@glad nebula rename the pipeline you have to something else or back it up, then use this one

rvc\infer\ here

[glad nebula]

okei

[gray rover]

cause if this gonna work, it means we'll need an extra f0 hybrid but for fumi

I suppose?

[glad nebula]

it worked

[gray rover]

ah ye

welp, thought they'd be cross-compatible

at least in that department

do you have any normal fcpe models around?

[glad nebula]

yup

[gray rover]

is there any you made on my fork ( newest, current

[glad nebula]

nop my other fcpe models were trained in fumiama

[gray rover]

well, will do

try to use it on hybrid if you can

[glad nebula]

sure

it worked too

[gray rover]

huh

quirky quirky lol

gonna make a separate hybrid for fumi tomorrow

[glad nebula]

[gray rover]

btw

ever seen such graphs?

[glad nebula]

this looks cursed lol

[gray rover]

wait for this

try to guess what I tinker with
( + what I cranked up to extremity to see if it even works

[glad nebula]

what?

[gray rover]

dropout

[glad nebula]

o

[gray rover]

in a short, it randomly drop some neuronal connections ish

encouraging the model to not " over-depend " on certain connections too much

in other words, a regularization method

[night lake]

what did you set it to?

[gray rover]

ofc, now it's cranked up for test purposes

to 0.5

I'll be doing some tests on 0.1, 0.01 and 0.05

[glad nebula]

i used 0.05 to train this fumi fcpe model

[gray rover]

you added dropout in too?

[night lake]

ive also used 0.05 and its alright

[glad nebula]

yuh

sounds... good?

[gray rover]

Gonna try 2d one too

supposedly better for vocoders

yet about resblocks, unsure, gonna check some combos

Think it could be quite beneficial for small sets where generalization and overfitting are touchy subjects

[night lake]

among us man (dr87) uses 0.05 dropout a lot and he says its good and doesnt lose any quality

[gray rover]

and by small I mean below 15-20 mins

[glad nebula]

speaking of overfitting, it's hard for me to tell when a model is overfitted lols

[gray rover]

generally, the idea behind not losing quality is to find equalibrium

[glad nebula]

amogus 🔥

[night lake]

https://tenor.com/view/among-us-dance-dance-among-us-purple-sus-gif-11799252726647481946

[gray rover]

where it doesn't lose too many connections

faster than it builds such

or so my intiution tells me at least

oh wait

I just now noticed dr has amogus on pfp, lmao

[night lake]

lol

[gray rover]

bruh

[night lake]

https://tenor.com/view/among-us-imposter-imposter-among-us-sus-among-us-sus-gif-158167910449982651

[glad nebula]

supposedly if the graph is going down the model is improving but idk
every epoch in the red box sounds robotic asf while the epochs in the green box sounds more "natural"

[gray rover]

because if the drop is too sudden in general, that's quite likely early-stage of overfitting

it should be more gradual and steady

good idea is to monitor disc alongside

[gray rover]

also, take a look at the metrics other than total

sometimes " better loss " has also worse fm + best mel

[gray rover]

see?

6-8k region is better than 10-13k

[glad nebula]

uh interesting

[gray rover]

sadly, mel metric is carrying the total loss the most

but that's that

[glad nebula]

tbh i still feel fumiama model is beating applio in terms of how natural it sounds

[gray rover]

you can also refer to % mel similarity

if it's sus high compared to surrounding regions, it also gives some clues

[gray rover]

I think most of that would be placebo or flux
cause deep down, those are practically the same in terms of layers and such

the only thing I could maybe suspect, was some changes to encoders and such ( but I don't remember if any changes were actual changes and not just ' let's make it nicer ' code-structure wise

[glad nebula]

fumiama

sounds so much better for me

0:04 in applio sounds metallic in every epoch

but ok in fumiama

also in fumiama you can safely change the volume envelope

in applio... well

uh maybe it's just applio inference that is making the inference result sound worse

gonna try mainline infer

[gray rover]

same model as 74?

[glad nebula]

yes

[gray rover]

also, important thing for you to keep in mind is that each inference is unique

[gray rover]

yet it can happen that 2 or 3 consecutive ones are close enough to have hard time telling apart

so it's best to do multiple inferences, 5-10 x 2

[glad nebula]

applio model got slightly better in mainline infer but still more metallic than the fumiama model

0:11

[polar flax]

I wish there could be random seed option for inference

similar to the SD one

[gray rover]

so let's put aside " metalic " aspect

do you notice most of changes in f0 response? f0 dynamics?

or is it primarily the voice's texture and color

[glad nebula]

applio still doesn't sound like the dataset

despite me disabling radam

basically i disabled both options to train using adamw

but yet the voice isnt quite there

fumiama in other hand its almost an exact copy of the set

[gray rover]

Well, I can't exactly give you any verdict because yeah, there's always some bias involved

[glad nebula]

well i was never a mainline fan to begin

actually been an applio user since noobies got "in charge"

i tried fumiama because i was curious if mainline got any better

[polar flax]

try doing inference 3-5 times on each of mainline/fumiama/applio

[gray rover]

^

[glad nebula]

ok

[polar flax]

though it seems rather tiring to do

[gray rover]

oh yea, mainline / fumiama ( too? ) support filter radius

[polar flax]

[gray rover]

and afaik, applio got rid of support for it (?)

[glad nebula]

fumiama fcpe model

fumiama infer

still not as metallic as applio

now gonna do the same but with the applio model

[polar flax]

well then

[glad nebula]

i can tell ya, applio models are more metallic

[polar flax]

the next thing I'd concern on is handling metal vocals/raspy voices

[gray rover]

I still think it can be within a marginal-error

[glad nebula]

metallic in every infer result

[gray rover]

even if you were to train the model again on same settings
it won't be the same, could be worse, could be better
or could be overal better yet have worse sibilants, it is really random in terms of what the ai gonna lock on the most

[glad nebula]

fumi gave me good results in one training

with every default setting

applio in other hand always is metallic

[gray rover]

and is fumi better than mainline in your opinion

[glad nebula]

aren't they the same thing?

[gray rover]

I ask you

[glad nebula]

no idea, i haven't tried non compiled mainline

1006 is quite old

[polar flax]

and also possibly "slurring" on some singing vocals

like between u ⇨ yu ⇨ i

the spin model is what comes to my mind

[gray rover]

I just remembered that uhh

[polar flax]

(I see the progress is closer to finalization)

[gray rover]

I think, generator has a lil bit different workflow for f0
in fumi ( not sure about mainline tho

[glad nebula]

this voice in specific
Every applio training i tried, it sounded extremely metallic and unnatural

but in fumiama it just sounds ok

i tried different lr

i tried your fork

batch sizes

everything

and every single result was metallic

i tried that on fumiama with every default setting

and its no longer metallic

and also sounds more like him

[polar flax]

btw is it 1006 mainline or 2024 version you were having?

[glad nebula]

https://github.com/fumiama/Retrieval-based-Voice-Conversion-WebUI/tree/dev

im just sharing my experience anyway

[polar flax]

I mean the mainline you were comparing against fumiama and applio

[glad nebula]

ive heard every finetune on #🔊│ai-development and they all sound slightly metallic

[gray rover]

or actually, sine generator

[foggy beacon]

who has Mel denoiser download

[glad nebula]

yeah i havent tried mainline

i believe they should be the same

but we know how is boss

so probably he didnt allowed some fumi changes

[gray rover]

I could dig deeper into it, maybe..

but prior to that, will need to try fumi's first

[glad nebula]

believe me, the results are good, you hear them

[gray rover]

I'll believe once I hear on my sets
don't mind me, that's how I do

[foggy beacon]

ight

[glad nebula]

yeah i understand

i actually i have the same mentality haha

[gray rover]

yuh, in that we agree

tho I think, it could be different if the phase generator has a lil different workflow

as in, different tone quality to it

we'll see.. we'll see..

[glad nebula]

https://huggingface.co/jarredou/aufr33_MelBand_Denoise

i want to believe this is probably a skill issue by me
because if it turns out mainline is better than applio then sheeesh

[gray rover]

then I'll just port some of the workflow from there
( sine gen etc

[glad nebula]

🔥 niceee

[gray rover]

also man xd

how funky it sounds at 0.5

( careful, volume

a zombie-model . Half alive, half dead, literally

[glad nebula]

voices from hell

[gray rover]

😌

[polar flax]

[glad nebula]

one thing too, fumiama got better pronunciation, at least for that specific model

[glad nebula]

wait no

it was the index

lmao

now they have the same pronunciation

still at index 0 applio sounds metallic tho

[gray rover]

well, all I can say for now is the sine generator differences

as far as I understand it, it creates a base excitation signal

which then is 'sculped' further, based on features, spec and / or f0 - depending on what repo uses similar structure
or in more proper terms ig:

SineGenerator produces the base excitation signal:
It synthesizes a waveform composed of:

The fundamental frequency (F0) and

Its harmonics (e.g., 2×F0, 3×F0, ..., up to N×F0),

Plus a voiced/unvoiced-aware noise signal.

[glad nebula]

btw the dataset is pure raw audio, no compression nor separation models, pure unedited mic audio

basically what rvc "expects"

so uhm yea

[polar flax]

another thing for me is the volume/dynamics in applio against fumiama/mainline

or perhaps just because of volume envelope

[glad nebula]

maybe i can do more fumiama and applio comparisons later

[gray rover]

oh yeah, that thing personally sucks in applio

for me

the vol / rms envelope
that's for sure

[glad nebula]

applio inffer just sucks

even old mainline do it better

no hate to applio devs btw

again just sharing my experience of using it

mainline is not perfect either

those dependency errors are such a pain
aaand you have innacurate graphs, but tbh you can just hear the epochs and choose the most "natural"

and thats rlly it

fumi ui is faster aaaanndd

u no longer need to manually put the path of the audio you want to infer

[chilly lake]

??

[jaunty heart]

Yo guys

I saw a lot of youtube faceless channel owner paid $1k to almost $10k per month for elevenlabs, I'm wondering they use elevenlabs instead of other options?

[chilly lake]

it seems to be the best quality option to make your own voice

but there are plenty of free options

[viscid shuttle]

is there a dream (the yter) model

anyone

[solar torrent]

Tyler, the Creator?

[tall needle]

How can I start a faceless yt channel and make money from it any reccomendation for like softwares and stuff?

[shadow python]

Ello

[halcyon wigeon]

hi everyone, i've recently ran into an issue where if i speak the voice doubles itself. has anyone experienced this issue/

[paper flare]

wdym? i just checked the option

[supple stream]

seek help in support channels #1192011222023950368 #1192011222023950368

[chilly lake]

he means this option does more bad than good

[paper flare]

oh

[queen kernel]

You can use comfyui for image generation, image2video, text2video and you can also create an Ai avatar using Fooocus, you can generate text to speech by Kokoro or any tts. By animating and using good assets you can create a wonderful faceless video

Framepack Is good for image2video

[paper flare]

uh

since i cant write in pretrains forum channels, why does klm 4.9 give me _pickle.UnpicklingError: invalid load key, '<'.?

[chilly lake]

you failed to download it completely

[normal sable]

Hi! I would like an AI cover with my voice singing a song. Is it possible? if yes how can I do it? Thanks!

[solar torrent]

Ah, as I expected. Yes, it is possible to make a voice model from your voice. For AI cover, go to #✨│ai-help.

[normal sable]

Ty, sry m8 that im so lost

[paper flare]

wdym? i just downloaded 40k D and G pretrains

[chilly lake]

it fails to unzip

the only way it fails to unzip is because it is either downloaded incompletely.. or instead of downloading the model you got 404 or some other html error

so the file can not be opened as .zip

[bold basin]

!colab

[pine acorn]

📒 Google Colab Notebooks

Google Colab is a Cloud (Remote Good PC) Service. While the Free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

• **Music Source Separation Training (Inference)**

by Jarredou & Makidanye
Google Colab

[quartz delta]

Hello, I used some software locally on my MAC OS that I could change a voice from a song, for a model. I accidentaly deleted and can't remember the name.

[paper flare]

this is my config https://files.catbox.moe/0y9fgz.png

[chilly lake]

ah... so you did not even download it, probably getting some http error

[paper flare]

its downloaded with aria2, cell output is
Setting up... Downloading your pretrained model... Pretrain downloaded. Best of luck training!

[chilly lake]

apparently not, since it does not work, right?

[young gazelle]

i managed to download that version from a github user that uploaded it in split 7z files

it didn't came with the runtime folder so i had to install the requirements myself

[paper flare]

it happens in applio noui as well

[queen kernel]

Promos are banned.

Delete it.

[chilly lake]

you probably not using the right download links

[paper flare]

at least look at image url i sent

maybe i should change blob to resolve?

[craggy ice]

step 1: you dont

theres millions of "faceless yt channels" who pump out generic AI slop

[tight hamlet]

Hi

[frail tree]

which voice changer works wioth these voices?

[minor blade]

Do you mean RVC models?

OG W-Okada and deiteris fork indeed work with them

[frail tree]

i mean what app do i use yk?

whwre can i use these models

[minor blade]

I've already said you can use OG W-Okada or deiteris fork

-realtime

[pine acorn]

💻 Local Realtime RVC

Guides for Programs that use RVC Models in Realtime for Calls/Games

• Wokada Deiteris Fork

Most suggested. GUIDE

• Original Wokada

ONLY the latest alpha comes close to the Deiteris Fork performance, older versions in youtube tuts are way worse. GUIDE

• **RVC GUI Mainline Realtime**

Unavailable, the guide is outdated and the program is worse compared to the ones above, and much less updated

[minor blade]

I highly recommend you the first guide.

[frail tree]

wtf is this bro so complicated

[minor blade]

Nope it's not.

Please, i would suggest you to take your time and read any of these guides. (i mostly suggest deiteris)

[frail tree]

d

ok

[frail wedge]

What is the best voice cloner?
I use Applio.

[cobalt badge]

anti ai in a weights servers is insane ngl.

[minor blade]

Yup, of course Applio is the best one

[craggy ice]

idk who weights is but its common sense

[cobalt badge]

the main ai this server is assosiciated with.

[paper flare]

how dangerous is the
FileNotFoundError: [Errno 2] No such file or directory: b'/content/Applio/logs/test_model/eval/events.out.tfevents.1746562531.62918c90cf30.8082.0'
error

[chilly lake]

did you delete the log while it was training?

[paper flare]

hmm possibly

sorry

Starting training...
Loaded pretrained (G) 'rvc/models/pretraineds/hifi-gan/f0G40k.pth'
Loaded pretrained (D) 'rvc/models/pretraineds/hifi-gan/f0D40k.pth'
/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py:558: UserWarning: ...
test_model | epoch=1 | step=48 | time=20:21:05 | training_speed=0:01:27
test_model | epoch=2 | step=96 | time=20:22:04 | training_speed=0:00:59 | lowest_value=31.906 (epoch 1 and step 35)

ok so i was able to launch this with downloader for myself, BUT

its still slower than kitlemonfot's work, based on mangio

[umbral oriole]

Hey there!

[paper flare]

its supposed to be like that?

or training speed will increase with time

also dataset is 15min22s, 3500 steps for 100 epochs is kinda... small

[paper flare]

on that notebook i was able to get 3900

it may be a minor discrepancy but im still worried

i would blame it on mutes amount but idk how much of them are there

[umbral oriole]

[paper flare]

um

test_model | epoch=1 | step=48 | time=20:21:05 | training_speed=0:01:27
test_model | epoch=2 | step=96 | time=20:22:04 | training_speed=0:00:59 | lowest_value=31.906 (epoch 1 and step 35)
test_model | epoch=3 | step=144 | time=20:23:05 | training_speed=0:01:00 | lowest_value=31.906 (epoch 1 and step 35)
test_model | epoch=4 | step=192 | time=20:24:04 | training_speed=0:00:59 | lowest_value=31.906 (epoch 1 and step 35)
test_model | epoch=5 | step=240 | time=20:25:04 | training_speed=0:00:59 | lowest_value=31.906 (epoch 1 and step 35)
test_model | epoch=6 | step=288 | time=20:26:03 | training_speed=0:00:58 | lowest_value=31.906 (epoch 1 and step 35)

its stuck?..

oh its incorrect(?) comment print

oh so it will have 4800 steps

i wonder if advanced users can predict the model quality from graphs with just few epochs

[wheat burrow]

gonna cry lol

5 hours for one epoch

😅

should have expected this

[chilly lake]

i dunno what steps have anything to do

15min set doing 1m/epoch is about expected speed for collab

advanced users use a tensorboard to see where to stop and which model to pick

[cobalt badge]

please teach me how to run ai locally.

[stuck karma]

gday veryone

dead chat

[tepid yew]

👋 hi yup time to revive it

[kindred kelp]

I should be the CEO of openai

[polar flax]

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

[paper flare]

are there any other vocoders than hifi-gan?

[river grove]

any anime voice model ?

[ember plume]

how do i use voice models

[runic shuttle]

if someone can help me in downloading the app i would be thankful if someone can please add me beacuse i am fasing alot of proplems

[supple stream]

#🔍│find-models

[supple stream]

seek help from hell channels #✨│ai-help #1192011222023950368

[runic shuttle]

Ok ty

[worldly grove]

Can I get RVC link.

[empty pivot]

@broken raft hi I have a question for u, did u have to delete the gura model for specific reasons or its somewhat still available?

[supple stream]

use -RVC in #🤖│bots

[covert lake]

RefineGAN but it's experimental

[covert lake]

You can search rvc ai voice models at:

• #1175430844685484042
• In #🔍│find-models , Do /find with @hidden grotto
• https://weights.com/ (login required)
• https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice 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.gg/essentials/how-to-make-voice-models/

[hidden grotto]

:wave: @covert lake, How can I help?

Available Commands:
• @weights find <query> or /find <query> - Search for RVC Voice Models
• /create - Create an AI Cover
• /image - Generate an Image

[covert lake]

Tell your PC GPU and what you want to do in #1192011222023950368

[covert lake]

Tell your PC GPU and what you want to do in #1192011222023950368

[oblique bronze]

hey

[minor blade]

If you wanna request a model use the #1175430844685484042 and search one, #1159289738314919936 or either #1191429836321849435

[minor blade]

He deleted it because Gura left Hololive and retired from streaming with that company. (I guess he just deleted it for sake of respect towards her and also because he simply wanted)

So you'll probably have to look for a Gura model somewhere else or use the #1175430844685484042 channel

[bleak bronze]

how do i download a voice form weights.com

[harsh tendon]

Guys i need help for thumbnail create with ai can somebody help me?

[dapper drum]

I noticed the de reverb by anvuew removes some if not most of the backing vocals is that normal

[bleak bronze]

i have one question after i donwloadet a voice form weights.com it it dident have the index file it only hat a json file

[queen kernel]

No worries. Index just contains accent

It will work without index too

[queen kernel]

Ask in #✨│ai-help

[queen kernel]

Promos are banned

[covert lake]

promos aren't allowed

[covert lake]

do you need help? if so, elaborate in an help channel

[sterile salmon]

are there any free tools i can use to upscale my photos

to look more realistic

that i cld rn locally also

[queen kernel]

Ask in #✨│ai-help

[covert lake]

ask that in support channels, and elaborate more along with ur pc gpu

[wintry kiln]

whats the best girl voice model

[tired jasper]

generated with Suno

i can't show my AI generated in an AI server?

hur dur

[ionic pumice]

capiche

[tired jasper]

names removed

good now?

[broken raft]

I deleted my Gawr Gura voice model because she already graduated from Hololive and I don't want anyone using it unethically.

[minor blade]

@elder willow Promo is not allowed on the server.

[thorny jay]

hello chat

[carmine flume]

Hello

Anyone here knows a lot from AI? I need help

[elder willow]

Does anyone know, if I need a specific version of Applio to train with KLM 6.0?

[carmine flume]

U know how to clone voice to say what u want to be said?

[paper python]

Hello guys. I need a TTS system that can run through the command line (not a GUI-based executable). It should support RVC models so i can use those models. Are there any good options available? I would really appreciate any help 🙏

[gray rover]

newest applio or my fork
however Imma recommend the latter as it's more noob-friendly ( in terms of spin

in any case, more info you'll find on klm 6.0's page or on #🔊│ai-development ( use search box )

[elder willow]

Thank you, I’ll explore both recent Applio and your Fork.

Last time I trained was back in December, so it’s been while.

And this whole Ai thing gets new every few weeks.

[gray rover]

well, then I'll just say that I wouldn't personally recommend using the klm 6.0 yet
likelihood of artifacts or timbre-related issues is quite high ( at least in regards to what's available atm

klm 4.9 is still the best option, on avg

[broken raft]

No, I do not, but you can follow the guide provided at https://docs.aihub.gg

[elder willow]

I’m specifically training Ai female singers.

Thanks, I’ll try out KLM 4.9.

[gray rover]

doesn't matter who you train, it's unrelated
nevertheless ye, 4.9 is the safest bet

[paper flare]

also 1 more questions regarding applio, how can i turn automatic backups to output one .pth model instead of D_epoch.pth and G_epoch.pth?

[silver bronze]

yo whats the most convincing female model you know of? my and my boy are tryna catch some predators type shi

[gray rover]

you mean the classic behavior where you only ever have 1 G and 1 D models at each save / epoch?

if that's what you mean the " save_only_latest " is what you should be using
( Unless you refer to some colab or whatever backup mechanisms then no idea. Haven't touched those in ages so I wouldn't know how it's handled now )

[paper flare]

i will try

also in what location and structure should i put models in order to be displayed there? https://files.catbox.moe/k8hepc.png

[gray rover]

logs

or
logs/<model_name>

use ^ for the sake of having it all organized

[snow portal]

Are there any free AI create song sites where I can create my own model?

[paper flare]

thanks

i cant use logs/weights/... right

[gray rover]

🤔 I mean, why not:

logs/weights/model.pth, index.index
logs/model.pth, index.index
logs/weights/model/model.pth, index.index

^you see, anything within the log folder, whether nested in folders or not, as long as it is either index file or .pth models, will be detected in the dropboxes

[paper flare]

F:.
├───test model
│       D_1880.pth
│       G_1880.pth
│       test model.index
│
├───mute
...
├───reference
...

this is not detected

works as intended?

[gray rover]

🤔 this is not a model

these are Generator and Discriminator files

those serve you no purpose as long you are done with training

[paper flare]

[gray rover]

no, you cannot reuse them for training without all the required folders/files

however you can extract small model out of generator ( sadly, you won't have any control on which epoch you want

[paper flare]

👀

whats the process called

[gray rover]

aka, you'll only get the latest one from the training

[gray rover]

welp

it used to be a thing in mainline rvc

it's not present in applio
but I guess i could maybe add it to my fork, sometime

[snow portal]

please reply

[paper flare]

its something more complex than def weight_from_g right

[glad nebula]

rvc already does this... the .pth files are converted G files

[paper flare]

i mean manually

in case user made mistake

[gray rover]

small weight / model extractor isn't present in applio sadly ( for whatever damn reason lol blaise and his genius once more

[glad nebula]

... why??? is such a pretty cool feature

[paper flare]

so its present or not

[glad nebula]

mainline only

[gray rover]

^
like I said, only in rvc

[paper flare]

okok

[gray rover]

However, not sure if it'll work

because I am not entirely sure if applio doesn't save dictionary differently

Noobies perhaps would know so, you can ask him ( I can't help - busy

[chilly lake]

there's extractor, you just need to pass the config

[still sigil]

@feral pelican stop telling people to dm you

[rare breach]

Hi, everyone! I'm new to this AI stuff, and I recently got a realtime voice changer. Could anyone help me get ther right settings setup?

[paper flare]

is it possible that my models sound bad after just 10 epochs because i use vocoder?

[covert lake]

elaborate more in #1192011222023950368 also telling your pc gpu, and be sure to not use video tutorials

[covert lake]

10 epochs are very little

[rare breach]

Thanks, will do.

[paper flare]

so are 100..

[covert lake]

be sure to use tensorboard and high quality dataset

also pls use #✨│ai-help

[gray rover]

vocoder?

Vocoder is always used, my dude

hifigan, bigvgan etc, all of those are neural vocoders

[paper flare]

according to model info tab in applio webui, my models from mangio have None vocoder

[gray rover]

Because it was never in mangio or mainline to begin with, that variable / key handling

tl;dr, yes you indeed use a vocoder. but your model has no info inside on what it is exactly that is used

[chilly lake]

it is HiFi-GAN, the default

@paper flare

because that's all we had before

[mossy garden]

anyone knows if there are newer versions of rvc gui thing?

[night lake]

This one: https://discord.com/channels/1159260121998827560/1341216399372062823

[white schooner]

shi

[gray rover]

shi

[gray rover]

Afaik, no.

[last notch]

hello

[elder willow]

where can one download this ai voice changer?

[mint elk]

yo can someone help me out with this okada voice changer

[sinful oar]

any tips to make the voice changer more realistic

the pitch is good but it'll tweak sometimes

[supple stream]

seek help from #✨│ai-help #1192011222023950368

[mint elk]

i need help in terms of editing the batch size, data set and etc. i dont know where to find them

[supple stream]

ask help there

[cerulean torrent]

Hey! I’m new to AI and really want to dive deep, but I’m a bit lost on where to start. How did you learn it, and what resources helped you most in the beginning?
I also wanted to ask – if I want to build real stuff, what should I learn first? Python, LangChain, no-code (like Make), or something else? I’d love any tips or directions!

[covert lake]

rvc gui outdated

tell ur pc gpu and what u want to do in #1192011222023950368

[covert lake]

tell ur pc gpu and what u want to do in #1192011222023950368

[covert lake]

tell ur pc gpu and share a screenshot of the owkada and issue in #1192011222023950368

[covert lake]

https://docs.aihub.gg/rvc-voice-changer/realism/#good-practices
use #1192011222023950368

[sinful oar]

Goat

[supple relic]

https://tenor.com/view/старый-бог-gif-2116010212877302133

[gusty frigate]

Hi all, can you please tell me how to fix the problem that any rustle is converted by AI as a voice? How to fix it? Is it possible to set a restriction that only from the volume, conditionally N, convert it to voice?

[covert lake]

elaborate in #1192011222023950368

[craggy ice]

still yet to see any AI thats useful for the laymen apart from GPT

[stray palm]

hello everyone! where i can make ai songs?

[stray palm]

please😭🙏🏼

[solar torrent]

Suno and Udio are websites that can generate AI songs from lyrics and prompt. If you mean by AI cover, there's RVC.

[stray palm]

yes yes ai cover

[solar torrent]

Well, if you keep asking like this in chat, I can only give you this way.

[stray palm]

bc some website is not working

[solar torrent]

For AI cover, there's Weights. However, if your PC has a good GPU, go to #✨│ai-help.

[stray palm]

what u think, macbook air 2011 is good?

my english is terrible sorry

[solar torrent]

That Intel Macbook is so old. You can't do RVC locally.

[stray palm]

im trying

[solar torrent]

The what? Well, you can run a RVC program on Intel Mac with only CPU, but it would be real slow to process. Are you sure about it?

Again, for an RVC program for Intel Mac, go to #✨│ai-help so I can send a guide for you to try lol. Chat here nowadays just become another help channel about RVC/W-Okada.

Or you can stay for online options that way, which are better than running on your old Intel Mac.

[tidal hound]

What is the best program for using AI models?

and what do I need to download for it ?

[minor blade]

i had a pretty interesting day

[minor blade]

RVC/Applio if you wanna use models locally with prerecorded audios

And Deiteris' fork if you wanna use them on realtime for discord calls/vc

-rvc

[pine acorn]

📚 RVC Documentations

AI HUB Docs

https://docs.aihub.gg

🍏 Applio Docs

https://docs.applio.org

[minor blade]

Read any of these docs above.

[hearty cloud]

Hello people, can someone help me with RVC please

[minor blade]

go to the #✨│ai-help and elaborate about your issue there.

[solar torrent]

Don't ask to ask. For RVC, go to #✨│ai-help or create a thread in #1192011222023950368 and explain about your issue.

[polar flax]

which AI models?

rvc or stable diffusion?

[covert lake]

elaborate in #1192011222023950368

@viscid solar this server is english only

[viscid solar]

😦

[covert lake]

It's english only as we can't moderate other languages, and english is the most common language in the whole internet

[solar torrent]

If you don't know English, you can use a translator.

[viscid solar]

please tell me if there is a model of the Russian artist "Pyrokinesis"

[covert lake]

You can search rvc ai voice models at:

• #1175430844685484042
• In #🔍│find-models , Do /find with @hidden grotto
• https://weights.com/ (login required)
• https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice 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.gg/essentials/how-to-make-voice-models/

[hidden grotto]

:wave: @covert lake, How can I help?

Available Commands:
• @weights find <query> or /find <query> - Search for RVC Voice Models
• /create - Create an AI Cover
• /image - Generate an Image

[covert lake]

it's better to use #1192011222023950368 or #✨│ai-help next time

[brazen shale]

Hello. Is there a Russian chat here?

I have a problem. I need a Pretrained for the Russian language so that AI doesn't highlight my accent!

[solar torrent]

https://discord.com/channels/1159260121998827560/1159346439424573440

[crimson zodiac]

guys hi, can someone suggest a website where I can search some ComfyUI workflows to download?

[astral island]

hey, guys

[solar torrent]

Hi.

[bleak hound]

guys whats the name of the app that turns songs into a model that we downloaded like minos prime

[covert lake]

tell your PC GPU and elaborate more in #1192011222023950368

[stiff zinc]

Ktos z PL i ogarnia te cudo?

[covert lake]

@stiff zinc speak only English here

[stiff zinc]

Okey

I have a problem regarding the voice configuration for the Fivem Game.
Is there an option or someone could help me configure on priv message or ticket?

[jovial sleet]

whats metadata? i downloaded a model and theres pth and metadata

[chilly lake]

a model description, not used for anything other than weigts.gg

[jovial sleet]

alr thanks

[raw kelp]

Give me the settings

[modern heath]

whats the basics in making and learning AI?

[empty pivot]

Oh may I ask what kind of things people could do unethically? Cause personally I just love trolling people with gura voice and play games with it lol, but yeah it's the old models

[broken raft]

What I meant "unethically" was I don't want people using my voice model, or particularly, Gura's voice, for shitposts and A.I. song covers made irresponsibly. Trolling, I wouldn't mind since that's a different situation.

[empty pivot]

Oh I see yeah, the part I hate about trolling is when I meet the people that.. don't know gura and think Im a child but at the same time they try to.. do things that are hella inappropriate, idk what's wrong with people

[minor blade]

Actually i don't really mind any AI song covers made with Gura or anyone's voice as long as it doesn't contain any disturbing, defamatory or illegal stuff

If you make AI covers, i would only suggest uploading them anywhere except youtube because you'll almost always run risk of getting copyright-striked by a record labels.

You can make AI covers of anything, just avoid stuff i've mentioned.

Altho the AI cover trend is already dead anyway

[left cedar]

is anyone here into creative prompt bot writing? I've been getting really into experimenting with creative writing and woul dlove o t chat to anyone with similar interests!!

[minor edge]

@serene star

[serene star]

yo

[minor edge]

sorry for the ping, thought it would be cool if I pinged you

[still sigil]

[ionic pumice]

https://cdn.discordapp.com/emojis/1356202556467974303.webp?size=48&name=skullwide

[gray rover]

Yeah?

Can help ( mogę pomóc )

Hit my dm ( W każdym razie, jeśli coś będzie trzeba odnośnie rvc / applio albo w-okady, daj znać. )

[dapper drum]

what is Light Host?

[night lake]

https://github.com/rolandoislas/LightHost

[dapper drum]

what is deverb mono model by anvuew

[glass junco]

my em model good?

[hexed agate]

ñ

ññ

[thorn mountain]

I think they were right about ai taking over devs jobs 🙏😭

[supple stream]

bro literally translated for him

[gray rover]

Nah, that bro wrote also in english so no kiddie goes sad cause they have to A) use translators but they lazy and B) deal with fomo 🙂
duh

[dapper drum]

Which one do I use to put the og backing vocals
the one that I get with anvuew v2

[olive horizon]

https://tenor.com/view/hatsune-miku-miku-sushi-miku-sushi-belt-dude-what-is-she-doing-why-is-miku-doing-that-gif-3385050863280224331

[nocturne valley]

Ai covers are dead 🥺?

[minor blade]

yep

[nocturne valley]

Well, that is not going to discourage me from making Dr Reflex kill his throat singing Oingo Boingo))))

By the way, could anyone help me with something? How do I make voice model while on mobile, exactly on Android. There aren't any voice models for Mrs. Pomp from Baldis basics and I would to make one myself.

[minor blade]

No one is stopping you of anything, so go ahead.

Just be careful because you don't know if you'll get hit with copyright strikes

[nocturne valley]

Except YouTube copyright ahahahha

Does changing pitch of the music helps, because those slowed/reverb channels stay

[minor blade]

Welp, making a model on phone would be the same as using cloud for training models, with the difference that you can't install RVC on phone.

https://docs.aihub.gg/

There you have a guide just in case.

[nocturne valley]

Thank you 🙏

[minor blade]

You're welcome buddy.

[rugged oriole]

-colab

[pine acorn]

📒 Google Colab Notebooks

Google Colab is a Cloud (Remote Good PC) Service. While the Free plan provides up to 12 hours of daily usage, the GPU is typically available for only about 4 hours each day on average.

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

• **Music Source Separation Training (Inference)**

by Jarredou & Makidanye
Google Colab

[supple stream]

nice approach

[covert lake]

Promos ain't allowed

[stoic radish]

fym promo

that is a funny clip

[covert lake]

This is still advertising, it's your own clip you published 7 minutes ago on a social media

[stoic radish]

so if i make it a dowload it can send it

[covert lake]

@stoic radish don't send anything related to bad jokes like 9/11. If you do it again, actions will be taken.

[stoic radish]

its not 9 11

what makes u think that

[covert lake]

The literal clip. And this is an ai channel

[stoic radish]

first off dirty mind i was just driving into a building secound it was ai

[covert lake]

Don't joke here about those things, it's seen in the clips. You have got warned

[stoic radish]

what makes you think im joking

[cursive prairie]

пожалуйста, включите звук

me unmute pls \

me unmute pls

me unmute pls

@peak dome

me unmute pls

[solar torrent]

Well, don't be an asshole here. Admit your wrongdoings, move on and you will be fine.

[solar torrent]

Please speak English in chat or go to https://discord.com/channels/1159260121998827560/1159346439424573440

[thorn hornet]

I'm tryna make clean versions of songs what is the best separator that extracts lead and back

and all the others

[remote fractal]

Yall know how to upload models to huggingface?

Been trying to figure out how

[covert lake]

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

[covert lake]

https://docs.aihub.gg/rvc/resources/dataset-isolation/

[covert lake]

Speak English only and don't beg others don't be unmuted, along with useless pings

There's tickets for any issues

[remote fractal]

Thanks

[nocturne valley]

I'm new to AI covers but I'm curious how long do I have to wait because it's been so long. I took the model from Weights, then I extracted the zip to get index and other stuff, I uploaded the model and then used it. Before I did the cover of a song that was 4 min with vocal model from discord but now with the weights one I'm waiting almost twice or thrice the time I've waited before.

It's been 8390 seconds

Also I cant understand is this number of seconds, so this means that I've waited for two hours, jeez, did I choose some fancy model or just why?

[fickle brook]

yo whats the name of the app again i forgot

[covert lake]

use #1192011222023950368 and elaborate

we don't help with only a single program

so you gotta elaborate

there's thousands of ai programs