🧬│ai-chat

[rich path]

thank mt dude are they good for using ai assistant as it willl be text to speech not speech to speech?

[covert lake]

Since both SoVitsSVC and RVC are STS

[rich path]

old is gold my dude just uh have to give time but i say its not worth it 100 to 150 good sample crazy man

[covert lake]

Well, nope unless you use another TTS to make first the audio then use it as an input for RVC

[covert lake]

Bro it's not good 😭

We used to have so vits SVC models

Everyone transitioned to RVC

It's way better

[rich path]

didn't under can uh elobrate a bit

[rich path]

yeah i guess uh r righ it gives better voice in lesss time

[covert lake]

TTS = text to speech
STS = speech to speech

So-VITS-SVC and RVC are both STS, not TTS

[rich path]

yeahh we have to use another tts i knw that brother

[covert lake]

The only way to use them as "TTS" is to use first another TTS program to make a TRS audio, then change the voice of that audio by using it as an input in RVC

If you want, there's other TTS options, ofcourse there will be less models and less popular than RVC though

[rich path]

thats the reason i was asking are they good ? the ones that are in voice models

[covert lake]

It got better quality and it's still maintained ye

[covert lake]

That depends on how good the person trained the model, you can listen to the audio samples or try it out yourself

[rich path]

are there any anime girl model and are they good in expressing emotions?

[rich path]

bro i need pre-trained i canot train coz i don't have good samples

[covert lake]

There aren't much TTS models so I'm not sure on your anime girl question

About emotions, can depend, F5, Fish Speech and GPT-SoVITS are good with emotions

[covert lake]

And I'm not telling you to train, I'm telling you to test the models yourself, RVC is good but it depends on how good other users also trained the model

[rich path]

well i can use rvc it will work just have to work little more and thats not a problem

[covert lake]

Those ones in the channel are already trained by other users

[covert lake]

Ofcourse the best TTS is 11labs, but it's paid and not open source

[rich path]

ahh uh mean those who are hosting the model ?

[rich path]

yh i knw but right now i am flat broke

[covert lake]

The models posted by users in #1175430844685484042 are trained by model makers in this server

[rich path]

can uh suggest me few with good voice ?

[covert lake]

Some models can sound more realistic, some others a lil bit less, it's better you test the model you want to use or use it yourself

[rich path]

that's why i am asking for recommendation

[covert lake]

I can just tell you the ones in our guides, but I haven't tested 10k+ models

[rich path]

i knw but anyways thank i will chk myself too

[covert lake]

Alright, goodluck

I gave you those 2, u could check them out

[rich path]

thanks dude

[covert lake]

Yw

[rich path]

yeah trying to download

[covert lake]

In case you are having issues downloading, To Download a Model from Weights.gg:

1. Login
2. Click the 3 dots at the right of the image of the model
3. Click download
4. Download Anyways
5. Unzip the zip, and you might wanna rename the pth and index since all models on weights are renamed as 'model'

[rich path]

uh read my mind lol

[covert lake]

Weights.gg is our Partner, it's a site where models get hosted too

It's ofc still RVC models

[rich path]

Thanks i appreciate the help

[covert lake]

You're welcome

[tranquil lantern]

by LUSBERT!?!???

[rich path]

well can't be helped free always comes with hardships

[covert lake]

Yep

[tranquil lantern]

Zamn

[covert lake]

Huh?

[rich path]

i have to write extra code tts and thn sts

[west burrow]

this is teas voice model

iirc it was a harrasment campaign or some bullshit

[west burrow]

oh dear heavens!

5 lines of ass code!

how difficult!

[covert lake]

you could use Edge TTS for the TTS, it's by Microsoft but free API to use, good with quality and multilingual, but it only has premade voices and not that good with emotion

[rich path]

bro i am frustrated i was stuck in a f*cking error for 2 whole days can't write a single line of code right now 4 lines are hell for me

[rich path]

i used google tts but nahh voice was shit

[west burrow]

why the fuck r u censoring ur own shi

literally use any tts model

even kokoro

rvc will just en-de-shittify it

[covert lake]

Google tts is another tts unrelated to edge tts

[rich path]

not talking abt model

[rich path]

yeah

anyways guyz see ya, will see uh soon

[west burrow]

monkey butt headass boy

i did this shit back in 23' before shit blew up

[rich path]

keep ur shit to urself please

[rich path]

trying to get attention how miserable a persona can be

[west burrow]

ok monkey butt headass boy

[covert lake]

@rich path @west burrow let's not do useless drama here, and Ofcourse respect each other.

[west burrow]

i ain even doin shit Lol

im just talkin about Bullshiot

https://tenor.com/view/cabbit-gorbsniffle96-sad-cabbit-cat-rabbit-gif-13272481107368375313

Fork

🍴

[rich path]

i was but they say give respect take respect

[tranquil lantern]

Lmfao gg

[west burrow]

https://youtu.be/5Z2Fzb-WIF4

[covert lake]

Insulting someone is not accepted here, let's be civil

[west burrow]

firefox is fucked

[west burrow]

ok monkey butt headass boy

this makes no sense LMfao

its just random ass words stringed together to sound like an insult

but it just means that ur a monkeybutt

[tranquil lantern]

that's barely insult that's like 4th grade shir

[west burrow]

@tranquil lantern monkeybutt headass boy

[tranquil lantern]

https://tenor.com/view/unsportsmanlike-wow-that-was-unsportsmanlike-wow-that-was-sports-ronaldo-gif-24314255

[solar torrent]

That still ain't enough to make someone cry. But if someone call you with a hard slur, you'd getting cooked haha.

https://tenor.com/view/crush-kana-arima-kana-plush-kana-x-aqua-gif-13787608196735728961

[west burrow]

lowkey id be the guy spewing slurs

[river adder]

THATS SO OLD LMAO

[west burrow]

Ok but idk who that is

Thats not me

i only go by Mikus bro

❤️

[tired summit]

Hello

[mental goblet]

Guys help pls. What the app and voice model would be best for hide voice in Investigated video

I think L's voice would be good

[lethal pike]

Sorry, I didn't really understand how covers are created. I was in a telegram group of an Italian developer who had created a site where covers could be made, but now it's closed and the site that he had provided where to retrieve the models it was closed. he said he is now operating as admin on this server. could you explain to me how it works now?

I would need the model of an Italian rapper "Guè"

[vernal marlin]

I be looking at the voice models section and I am just thinkin

What are some of these models even being used for?

What use is somebody getting out of "Jeremy finkleforter from highschool 5 years ago"

[midnight wraith]

how you create a custom voice cover?

[covert lake]

what's your pc gpu

[covert lake]

ig ai covers or realtime lol

[covert lake]

all that sites use is RVC, it's better you use that, what's ur pc gpu

[midnight wraith]

rtx 3070 why

[covert lake]

why
Because computer power is needed for doing heavy tasks, just like how games and blender rendering does too for example

anyways, ur pc gpu is good

[midnight wraith]

I just dont even know how to make stuff

I wanted like to choose a song, create a clone voice with AI using clips of videos of me talking, and then put my clown voice singing that music

how could I do that

[covert lake]

Everything you need to know is in our docs

I would suggest you to use locally applio

which is a fork of rvc, fork means a modified version

[midnight wraith]

But I need to donwload something?

[covert lake]

you need to first train yourself your own voice model, then separate the vocals and instrumentals of the song, then inference (use model) on the vocals, then mix the converted vocals with the instrumentals

[covert lake]

Yeah, it's local, which means it runs on your pc

I suggest you that since your GPU is good

[midnight wraith]

can't be online?

[covert lake]

it can be cloud, which uses remote good pc, but you will have to deal with limited GPU time on the free tier

It's used for people who got a bad pc, which isn't really your case tbh

you still want me to give you the links?

[midnight wraith]

hmm

well I don't know how to to all that stuff ahah

[covert lake]

for both cloud and local there are written guides dw

all you gotta do is read them

if you're wondering about youtube videos, they are pretty outdated

[half heron]

Hey guys

I'm a junior js developer (mainly frontend) and currently using this platform - https://sparkengine.ai and building some AI tools cause I like building them without coding. Where can I find more no-code tools?

[rich path]

which tools uh want i am also building a ai assistant but my project is a big one

[west burrow]

monkey butt headass

[mental goblet]

Which chunk I can use in rtx 3060 mobile 6gb (notebook) in Ai voice changer?

[covert lake]

@late nest sorry but promos aren't allowed here

[covert lake]

tell the guide link u followed in #🔍│help-w-okada

[dim escarp]

is mmvc down? I cannot get it to start for the life of me. I would reinstall it but I can't find the link

[solar torrent]

Stupid ass boy.

You insult like what a bully from an elemental school would do. I insult with something harder than yours.

[solar torrent]

For W-Okada, go to #🔍│help-w-okada. The chat isn't where you asking where to download W-Okada.

[keen grail]

Server mad dead now

[polar flax]

bro left

[ionic pumice]

mikus been here for

[tulip cedar]

Guys does someone knows when the day resets on Weights.com? I tried to look around the site but im dumb

[ionic pumice]

a bit actually

og aihub

[dreamy wigeon]

heloooooooo

anyone here?

i need some help

[solar torrent]

You meant Weights.gg?

[solar torrent]

Please be specific on which problem you encountered.

[polar flax]

!howtoask

[pine acorn]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[tulip cedar]

yeah

[solar torrent]

I'm not sure what you mean by the day resetting. Do you mean something like this on Weights?

[tulip cedar]

like, i try to create images and it says theres a limit of generations per day on the F12 console

[solar torrent]

I don't know how many features are limited for non premium users. I got premium, and I found no problem using image generator. It's more likely there are too many queue at the time trying to generate.

[dry hornet]

meant to send in here

can somebody help me w smthn

[polar flax]

https://tenor.com/view/idc-i-dont-care-no-one-cares-when-someone-be-like-spongebob-gif-16322556

!howtoask

[pine acorn]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[proven wyvern]

Elon Musk's Grok 3 AI Is A GAMECHANGER

https://youtube.com/shorts/RhRoX04NN2w?feature=share

[solar torrent]

Game deez nutter.

[ionic pumice]

fix yo shi and maybe people'll like it

[elder willow]

oof

any ais here?

[ionic pumice]

@elder willow in #🤖│bots

[undone cipher]

do you know tts i want to make a model that can speak and when chatting will speak and will have to move when speaking

I want to do it on reactjs only but I don't know if it's possible

[solar torrent]

No, thanks. I can work on my own.

[minor blade]

Promo is NOT allowed here @leaden halo

[leaden halo]

Any careers channel?

[minor blade]

Nope, sorry.

[ionic pumice]

hi guys my name is abhiraj shah and i am NOT the ceo of a new company

[minor blade]

@cyan onyx Promo is not allowed.

[leaden halo]

No worries!

[safe pier]

any content creators on here?

[ionic pumice]

roomie was here

[covert lake]

@tepid basin did you ever play humanornot.ai ?

[tepid basin]

No

[covert lake]

great bc it has a javascript vulnerability 😭

[tepid basin]

Thinking about investing in the mcchicken

[covert lake]

huh

[tepid basin]

It’s a classic for a reason. Savor the satisfying crunch of our juicy chicken patty, topped with shredded lettuce and just the right amount of creamy mayonnaise, all served on a perfectly toasted bun. The McChicken has 390 calories.

[covert lake]

smh

clearly AI

[jaunty osprey]

Hi guys, I have a question. Can we monetize an AI influencer? And if so, how? I've seen a lot of videos about this, but I'm not sure, and I don't want to waste my time. I'm too old to waste time without a payback

[tepid basin]

With ads

[jaunty osprey]

no with onlyfans and fanvues

[ionic pumice]

https://media.discordapp.net/stickers/1277491531195875421.png?size=160&name=thats+kinda+crazy+gang

[covert lake]

@jaunty osprey asking for NSFW help isn't accepted here.

[dusk urchin]

Question: How good is Grok3 compared to ChatGPT and Deepseek?

I've been feeding ChatGPT with sources to learn an obscure language and it generate texts pretty well (I ask it to make an essay about genetical engineering and how this is good to our future and it's pretty decent)

So I want to know if Grok can do better

[covert lake]

I haven't tested grok

[river adder]

https://tenor.com/view/nope-sad-sad-face-stare-look-up-gif-16490915

[gray rover]

Hmmm.. I've been wondering for a while, is there any known and more or less reliable workflow for " recreating " character ai locally?
any particular model / backend / frontend etc?
Looking for something as advanced ( and functional ) as it can get

either way, any info / ideas appreciated ~

[polar flax]

ew

[gray rover]

^ 100% disgusting

[whole merlin]

guys im trying to upload my model and i put the hugging face link and it still isnt accepting it, it says (must be a hugging face link or weights.gg)

and im putting the hugging face link and it still wont allow

[quartz roost]

https://youtu.be/u4-dqYub2uo

[glass junco]

holy chat this neew tupac data sety go crazyyy at 116 this me rapping

[undone cipher]

guys how to add model to tts

[river kestrel]

how do i train a model? can someone point me towards a direction

[river kestrel]

this mixed with adobes audio enhancer would sound so real

you could probably put some audio of tupac before it too so it really gets the job done

[glass junco]

bro this is at 160

[glass junco]

bro 320 sooo crazy just updated

[river kestrel]

the breathing parts sound weird tho

[glass junco]

yeah thats because its me rapping and im white lol

[glass junco]

i might have the realest one

[vale wigeon]

hi

[pallid socket]

hi

I heard pretrain models are good

[rich path]

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

[minor blade]

Nope, that has nothing to do if a model sounds robotic.

If a model's breaths sound robotic that's because you didn't clean the dataset properly or you under/or overtrained it.

[covert lake]

Will reply in #✨│ai-help

[viscid bluff]

anyone know how to earn those points used in voice.ai the fastest without paying?

[solar torrent]

Please never use Voice.ai again. This site is a scam, and eats more performance that any of W-Okada.

[viscid bluff]

I try W-Okada but the changed voice sucks and doesn't even sound good, I have a really good pc aswell😭

[minor blade]

Actually the results will depend on the model you use and your own voice. (also your settings)

[viscid bluff]

ik

[solar torrent]

I think you might have used the original version of W-Okada, you didn't set anything right, and you gave up that way instead of getting help from AI Hub by Weights.

[minor blade]

Also.. I would suggest you to use Deiteris' fork instead of OG W-Okada, including reading the docs.

What's your GPU?

https://rentry.co/ForkVoiceChangerGuide

[solar torrent]

Refrigerator full of game cartridges. 🔥

[viscid bluff]

well my processor is a AMD Ryzen 7 7800X3D 8-core Processor, I also have a NVIDIA GeForce RTX 4070

[solar torrent]

Cool set up you have.

[minor blade]

Hi Unity.

[minor blade]

Nice setup, you should be able to use Deiteris' fork with no problems.

[viscid bluff]

yeah, I actually already tried that lol

no performance issues

[tepid basin]

Might need a couple more 4070s for the voice changer tbh

[viscid bluff]

yeah

[minor blade]

That's good. Altho maybe your issue is getting a proper model that works with you.

[tepid basin]

I'm joking btw

[viscid bluff]

ik ik

[tepid basin]

Runs good on my 4070

[covert lake]

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

[viscid bluff]

lol, I already uninstalled it xd

[minor blade]

If i had an actual GPU i would test realtime with my BM800 mic that will arrive on next week hehe.

[covert lake]

use the wokada deiteris fork 🔥

also the quality can depend on which model you use

[viscid bluff]

yeah

[minor blade]

Nick i think i got motivated and ordered a mic.

[tepid basin]

It won't load but based on the colors of the preview I bet its that spongebob gif isnt it

[viscid bluff]

a good one

spongebob throws voice.ai in the trash

[covert lake]

nice

[polar flax]

there's no way you have any performance issues unless when running a demanding game

[viscid bluff]

yeah ik, it's probably because of my low quality goofy ah mic🙏

[viscid bluff]

btw which F0 detector would you recommend?

[night lake]

Rmvpe

[viscid bluff]

oh okay, I was using fcpe this whole time😅

[covert lake]

fcpe is lower quality

it uses lower resources but has worse quality

[viscid bluff]

yeah, is Rmvpe the highest quality one?

[night lake]

Very technically crepe with a low hop length is but you need very clean audio

[viscid bluff]

I have crepe_full and crepe_tiny lol

I imagine full is better?

[night lake]

It would be crepe full but use rmvpe

[viscid bluff]

okay thanks!👍

[glad nebula]

crepe full has a hop length of 160 iirc

so it perfoms exactly the same as rmvpe

[viscid bluff]

oh lol

[glad nebula]

crepe beats rmvpe when has a hop length of 64

[viscid bluff]

so which one is ultimately the best for quality if you have a really good pc?

[glad nebula]

a good model

rmvpe f0 (both for training the model and for realtime inference)

[weak hemlock]

What happened to the realtime voice changer

Not been here for a while

[covert lake]

Wokada has 2 main versions:

• Original made by Wok
• Deiteris fork (modified version) made by Deiteris

each version has it's own updates

the latest deiteris fork has way better performance and quality than the latest original

you probably downloaded an old original version many months ago

and want to update it

pls tell me ur pc gpu in #🔍│help-w-okada

[pliant laurel]

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

[chilly lake]

they asked about hifigan pretrains

not newer

[covert lake]

oh I read it as refinegan somehow 💀

I gotta buy new eyes

[pliant laurel]

my bad I did mean the new ones, I wrote the wrong one

[covert lake]

lmfao

well lemme write it again

[covert lake]

this is because the latest stable release 3.2.8-bugfix don't contain the latest experimental code

refinegan is the new ones

[pliant laurel]

does that only work with the no ui applio colab? or can I do that with the ui colab

[covert lake]

both

just remove that part I told you

[pliant laurel]

Alright, thank you

[chilly lake]

it would only work in UI

noUI needs to pass a vocoder parameter, that's not there yet

[covert lake]

oh it wasn't added there yet?

@pliant laurel then nvm, just UI

[pliant laurel]

Thank you, that's ok. I perfer the UI version anyway

[desert star]

guys

someone online?

@wide owl

L bozo

[polar flax]

https://tenor.com/view/pookie-offline-sad-cat-gif-4209851872672405309

[ionic pumice]

https://tenor.com/view/byuntear-fainting-desmaio-mad-crazy-gif-6971696022815179065

[hasty marsh]

Hey guys. Maybe someone knows a good and cheap API to create AI avatars from photos/videos and stream them?

[tall vapor]

why does the lightning ai applio have more features than the local one am i tweaking

[covert lake]

what?

[tall vapor]

there's more training options in the lightning ai version of applio

than the local version I have

and i can't find a more recent version anywhere

[covert lake]

which lightning.ai link are u using?

and also what local version

[tall vapor]

https://lightning.ai/guilhermecardoso1/studios/applio-latest?section=all&query=applio
https://huggingface.co/IAHispano/Applio/tree/main/Compiled
Just both of the ones linked on the guide page

[covert lake]

that's because the latest precompiled stable applio version is 3.2.8-bugfix, while the lightning.ai notebook uses the main branch, which has newer experimental updates that aren't in a stable release yet

[glass junco]

my tupac ai new model with suno is really good

[wheat burrow]

damn bruh this is good

[mental scarab]

do i need python/any other things installed for okada

[glass junco]

thanks bro

[mental scarab]

yo

[mental scarab]

do you know abt this?

[covert lake]

#✦│chat message

[steel hemlock]

Help

How to use voices zip to convert into the voice audio

[coarse aurora]

hello, are there people here who make complete music videos like me?

[vague iron]

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

[plush barn]

i noticed my rvc models and my role got removed what happen?

and i wasn't notified

[night lake]

to many of your models were bad so they were removed along with your role

[glass junco]

how do people do this i wanna do this for a kick user https://www.youtube.com/watch?v=EC9Fyz4exII

[noble sand]

is there any more beatrice v2 voice models?

[polar flax]

@dusky grove any account requesting or sharing is strictly prohibited

[dusky grove]

sorry

[orchid oriole]

So u guys have any good ai voice models for girl voice trolling

[solar torrent]

For W-Okada, go to #🔍│help-w-okada

[shut crater]

any Developer here

[ionic pumice]

girl idk is there

[fringe prairie]

https://cdn.discordapp.com/attachments/1302224816648093737/1343094492827484161/sxadffffffffffffffffff.gif?ex=67bc05b8&is=67bab438&hm=473735832223444e15f240355b0b2a4a4207d8c030e4a835d1fb1d0b99ebf64e&

[frigid cloak]

Hello, what's the best TTS you guys could recommend me? Especially for Japanese

[shut sage]

-colab

[red frost]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• Hina's Mod 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
• Hina's Modified Original W-Okada's Realtime Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab
• 🆕 Music Source Separation Training (Inference), by Jarredou & Makidanye Google Colab

ℹ️ Note

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

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

[shadow whale]

I'm really struggling to use colabs, they all seem to be broken

I used ultimate and hina usually but both are broken

[covert lake]

elaborate:

• ur pc gpu
• what google colab link you used
• what you want to do
• the issue/error

[elder willow]

Are the voice models open source?

[covert lake]

wdym

[elder willow]

Can it be freely used without restrictions on social media?

The Voice. ai rvc models

[shadow whale]

I use ultimate google colab and it just doesn't show any mp3 after ive generated it just says 0 seconds

i just need a simple one which works

I tried applio but i couldn't figure it out and also rvc disconnected didnt work

[elder willow]

I just need a realistic female voice translator that these catfish vids use on YouTube.

[polar flax]

the rvc technology is, but don't expect anyone disclose the model dataset used

[polar flax]

ditch that paywalled garbage

[elder willow]

Any recommendation for free sources that is realistic?

Clownfish isn't good

[red frost]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

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

[covert lake]

voice.ai sucks

it uses RVC on the background, which is free and open source

but makes u pay for them and does distributed training with ur pc

I'm guessing you want realtime voice changing right? tell your pc gpu in #🔍│help-w-okada

[shadow whale]

None of them are working 😦

[covert lake]

!howtoask

[pine acorn]

How To Troubleshoot

__**GIVE CONTEXT.**__ 📝

• Don't simply mention your issue, like "my rvc is not working".
• Describe the step you are on, what you're trying to do, the RVC you're using, a screenshot, etc.
• The more context, the better.

__**BE POLITE.**__ <:matsuripray:1159685390156967936>

• Don't be desperate. You can ping a Helper, but if they ignore, they aren't available/don't know the answer.
• It's okay if you're frustrated, but don't take it into this server.
• Don't DM without prior consent.

__**BE PRODUCTIVE.**__ 🤝

• Don't ask for every little instruction. Put your own effort & test things by yourself.
• Don't ask to ask.
• Check if your answer is a Google search away/on our guides website.

If you've changed your mind about your message, feel free to delete/edit it.

[covert lake]

you need to explain the things I said to make so people are able to help you

[cedar nymph]

Does anyone know how to make your own song parody with changed lyrics?

[minor blade]

For that you gotta record yourself singing the changed lyrics and then pass your vocals thru a model.

[cedar nymph]

okay'

[round basin]

hi y'all, quick question

i used to use the voice models with clownfish vc but are they useable with voicemod ?

[covert lake]

don't use those

you want a realtime voice changer right?

tell ur pc gpu in #🔍│help-w-okada

[shadow whale]

I'm not very tech savvy so i'm not sure on some of these but my pc gpu is fine, i have used multiple diff google colab links, I just want to use singers voicers and apply them to other songs and there is a few diff eras

errors*

[covert lake]

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

tell me the name of your pc gpu

also send the links of the colabs you used

[shadow whale]

My pc gpu is fine because it was working yday it just randomly stopped today

okay lemme send 1 sec

https://colab.research.google.com/github/JackismyShephard/ultimate-rvc/blob/main/notebooks/ultimate_rvc_colab.ipynb#scrollTo=lVGNygIa0F_2

https://colab.research.google.com/github/ardha27/AI-Song-Cover-RVC/blob/main/Hina_Mod_AICoverGen_colab.ipynb?hl=en_GB#scrollTo=vC4gLMHI9xb3

and also applio, rv disconnected

[covert lake]

I'm asking you to tell me the name

there's many different gpus

[covert lake]

rvc disconnected got its end of life update, it won't ever be updated again and it's old

is a specific model not working on those colabs? if so, send the model download link

[shadow whale]

im not sure i just clciked the links in this discord

and gpu im using default one

[covert lake]

Do the things I told you on your pc

[covert lake]

.

[covert lake]

Do you remember the model download link

[shadow whale]

im using

huggingface

https://huggingface.co/gleekerperson/Quinn/resolve/main/Quinn.zip?download=true

i used it yday it worked

[dry jetty]

Hi, how do i find these exact models listed in the Best models in UVR beta? https://docs.aihub.gg/rvc/resources/dataset-isolation/

[shadow whale]

can anyone please help

i just want a colab that works

[worthy coyote]

don't worry sir the spider on the bread will get him

[covert lake]

mm this model seems fine

[shadow whale]

ive tried

every colab

nothing works

[covert lake]

did you try what I said? #🧬│ai-chat message

[shadow whale]

which part sorry

[covert lake]

the one to check your pc gpu

[shadow whale]

yes

im using gpu or t4 gpu

[covert lake]

I was talking about your pc one

to check if it's good enough to run it locally

bc google colab is a cloud (remote good pc) service only for people with a bad pc

[covert lake]

also I have just tested this on the Applio colab and it works all fine

[odd moss]

I have an idea for a YouTube channel where AI-generated celebrities teach you different skills. Imagine Danny DeVito showing you how to fish or Arnold Schwarzenegger teaching you how to drive. It’ll be funny and a bit silly. My plan is to use text-to-speech AI to create the videos, then use AI to generate the celebrity voices. Alternatively, we could have someone record the script and then use AI to mimic the celebrity voices. The final videos will be AI-generated. I’m looking to collaborate with someone on this project.

[gray rover]

update: codename fork's up to date with applio changes

[gray rover]

@arctic tinsel Damn man, I just tested ur hifi pretrains

Truly outstanding 🔥

( Just a lil short test on rather low-mid quality -roformer set, [ 32 secs bs4 ] and it already outperformed stock ones by at least 75% )

[toxic cobalt]

Hello, I want to ask you something. I cannot view the Turkish voice models by typing "Turk" in the voice models channel. Could I have made incomplete or incorrect choices while registering? Can you help me with how to fix it?

[shut sage]

There's an app or website for Android or mobile that we can change the lyrics of existing songs?

[elder willow]

nick

[covert lake]

?

[covert lake]

I don't think so

[shut sage]

Okay thanks

[modern gull]

هلا

[toxic cobalt]

@covert lake

[covert lake]

type turkish

Also, You can search rvc ai voice models at:

• #1175430844685484042
• In #🔍│find-models , Do /find with @hidden grotto
• https://weights.gg/ (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.ai-hub.wtf/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

[toxic cobalt]

thanks

[covert lake]

yw

[elfin geyser]

where is the download link

[main vessel]

currently I am in second year of CSE and I was making Deepfake detection project, now I got stucked in training model, means I take dataset of images from Kaggle and while training model it requires more RAM and GPU, I used colab's TPU , but it is free for 4 hours only, and I need more time because my dataset are too large

and if I train the model on 10,000 images then it provides only 0.25 accuracy and if I use 60,000 images then it provides me 0.77 accuracy,

can anyone suggest me with this?

[storm hazel]

serbian artillery is guided by god

[covert lake]

Download link of what? What are you doing/want to do?

[tranquil lantern]

yet still cannot put down Kosovo 😂

[covert lake]

@heavy bone promos ain't allowed

[dapper drum]

Anyone have the AI documenting where it tells you which one you should use for clean vocals

[lusty bridge]

Hello everyone! How to create something simalar to this project on the youtube video (it uses old photos with paralax effects and probably AI Kling, as I guessed). Thanks you for your afvices! https://www.youtube.com/watch?v=T7X-9qkd0iQ&ab_channel=ZAJACFOTO.COM-REMASTERING

By the way, that 4 minutes project took roughly 2 months of creator's time, but quality of work is amazing!

[covert lake]

promos ain't allowed

[storm aspen]

people, what should I do if I have a gtx 560 ti?

[chilly lake]

buy something better

[dapper ginkgo]

Are you mentally insane?

You cant say such shit to a most likely minor

@tepid basin you may want to look at this lol

[tepid basin]

Why ping me 😭

[dapper ginkgo]

You dont see the problem here?

[minor blade]

Bc Sky probably wants you to nag Noobies for what he said to a supposed "minor"

[tepid basin]

I don't do critical thinking but I'll forward your concern to staff

[twin gazelle]

how can i run ai models locally

[minor blade]

RVC models?

Welp, you can install RVC locally by reading a guide i'll give you.

https://docs.aihub.gg/rvc/local/applio/

[twin gazelle]

ty

surely

my rtx 4070 ti would be enough right

[minor blade]

Yup, of course.

[tepid basin]

Wassup

[minor blade]

For further help, use the #✨│ai-help channel 🙂

[twin gazelle]

wassup my clone

[covert lake]

https://www.ign.com/articles/activision-finally-admits-it-uses-generative-ai-for-some-call-of-duty-black-ops-6-assets-after-backlash-following-ai-slop-zombie-santa-loading-screen

[obsidian elm]

anyone has the download because when I follow a tutorial the downloaded file is always different

[ionic pumice]

tf kinda tutorial youre downloading brah

[obsidian elm]

https://www.youtube.com/watch?v=t56mjuv2MK0

[chilly lake]

https://rentry.co/forkvoicechangerguide

[olive otter]

what do you guys use for rt voices? beatrice?

[covert lake]

Beatrice v2 models are Faster and more Lightweight than RVC v2 models, but have Less Quality and there are Less Publicly Available models

RVC V2 is better

[olive otter]

ok, thank you

[covert lake]

Tell your PC GPU and what program download link you used in #🔍│help-w-okada

I will check if u use the newer one

[olive otter]

and what program do you use for real time voice changing?

I'm using vcc

[covert lake]

Wokada deiteris fork, the best quality and most optimized one

[covert lake]

Lemme guess, you seen YouTube tuts?

Please tell your PC GPU and program download link in #🔍│help-w-okada

That is the right channel for this

[olive otter]

where can I download that one?

[covert lake]

Explained you in #🔍│help-w-okada

[kindred kelp]

Anyone wanna help me sell subscriptions to my social media automation tool?

[kindred kelp]

Use ollama

[solar torrent]

I don't share my email account to anyone.

[spring stag]

Do you know of any AI to help me generate an image from another?

[chilly lake]

basically pretty much every AI image generator using image prompt

[dawn temple]

[spark musk]

Hey what is the best voice changer for male to female ?

[elder willow]

is there any deep voice models?

[covert lake]

tell ur pc gpu in #🔍│help-w-okada

[covert lake]

You can search rvc ai voice models at:

• #1175430844685484042
• In #🔍│find-models , Do /find with @hidden grotto
• https://weights.gg/ (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.ai-hub.wtf/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

[tardy belfry]

how tf can i make the ai sound good in spanish

[graceful sparrow]

Hi

/image duck

[green ivy]

Are you all AI

[covert lake]

To generate images for free (text2img), either:

• Use @elder willow in https://discord.com/channels/1159260121998827560/1202754985255764060 (It's powered by DALLE3, from ChatGPT+), pretty easy
• Another easy and good ways with weighs.gg are:
  • Use /image with @hidden grotto in https://discord.com/channels/1159260121998827560/1202754985255764060
  • Create an image on their site https://www.weights.gg/ (which you can also use LoRAs, Low-Rank Adaptations, basically a small trained additional model to adjust your generation)
• Use Open Source Models like stable diffusion & flux that could be a bit **harder **but good, what's ur pc gpu? As you could run them locally (on ur pc) or on cloud (remote good pc)

[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]

be sure to get a good quality model, or dataset if you’re training one

also use help channels

[green ivy]

Thank you

[plain hollow]

[cedar bay]

Dont die

Make a game

Fail with the game

make a new game

succeed with the game

Make lots of money

Make a new game thats literally the same but new story

Make lots and lots and lots and lots and lots of money

Name your studio Rockstar

[rare night]

does anyone know a free alternative to elevenlabs ?

[covert lake]

There are different Text To Speech (TTS) AIs:

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

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

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

With RVC Models:

RVC is natively for Speech To Speech, but forks such as ilaria rvc mainline & applio have built in tts (using Microsoft Edge TTS to make a generated tts audio, which i suggest you to choose a tts model that is the same gender and language of the rvc model you wanna use, and then convert it with rvc)

If you wanna do tts locally with RVC Voice Models (if you got a good pc):

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

If you don't got a good pc you can do tts with RVC Voice Models on cloud:

• Ilaria RVC Zero (Running on A100 GPU, free fasted rvc on cloud) and the guide

• Use Applio UI Colab (with google colab T4 free daily limit gpu)

• if you don't wanna use edge tts, you could try another tts ai from our tts index and use the output as an input in rvc

[kindred breach]

hey

can anyone help me out why my rvc trained models sound like robotic vocoder

idk were i am going wrong

[shadow whale]

applio broken for anyone?

it says gradio cant be found

[chilly lake]

new colab instance?

see the output for the install cell

[shadow whale]

what does that mean sorry - yes for colab

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

[hybrid hill]

How do you find a jet ai model?

[alpine fjord]

#🔍│find-models

I have my own question though

I haven't messed with ai in multiple years, so there is probably new software. So what is the best app I can run locally on my pc to create tts that allows custom models and stuff?

I also need it to be able to emotion to the voice

[covert lake]

There are different Text To Speech (TTS) AIs:

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

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

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

With RVC Models:

RVC is natively for Speech To Speech, but forks such as ilaria rvc mainline & applio have built in tts (using Microsoft Edge TTS to make a generated tts audio, which i suggest you to choose a tts model that is the same gender and language of the rvc model you wanna use, and then convert it with rvc)

If you wanna do tts locally with RVC Voice Models (if you got a good pc):

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

If you don't got a good pc you can do tts with RVC Voice Models on cloud:

• Ilaria RVC Zero (Running on A100 GPU, free fasted rvc on cloud) and the guide

• Use Applio UI Colab (with google colab T4 free daily limit gpu)

• if you don't wanna use edge tts, you could try another tts ai from our tts index and use the output as an input in rvc

[dawn temple]

OMG NICK GOT HACKED, HE IS A BOT NOW

or he always was 🤔

What is that supposed to mean?

[cobalt coyote]

Heyo

[covert lake]

Real

[solar torrent]

Nuh uh. I don't accept friend request from random people, so don't worry about me.

[tiny olive]

When I use a voice model to speak with the voice I have to use a virtual cable. All that is fine and also working. Though once I do that I dont hear myself anymore and I'd like to do that. I there a solution?

[solar torrent]

For W-Okada, let's go to #🔍│help-w-okada.

[tiny olive]

oh wah, missed that channel. Sorry!

[cobalt coyote]

huh'

[cobalt coyote]

This channel has been set up to receive official Discord announcements for admins and moderators of Community servers. We'll let you know about important updates, such as new moderation features or changes to your server's eligibility for Server Discovery, here.

You can change which channel these messages are sent to at any time inside Server Settings. We recommend choosing your staff channel, as some information may be sensitive to your server.

Thanks for choosing Discord as the place to build your community!

[covert lake]

100% rial

[solar torrent]

Did you copy a message from an admin channel of a Discord server you managing?

[cobalt coyote]

ded chat

[cobalt coyote]

hehe

sup @solar torrent

[solar torrent]

There are two chats: #✦│chat and #🧬│ai-chat.

[solar torrent]

Hi.

[cobalt coyote]

yus

[pale patrol]

this is new o- o

[solar torrent]

What is this Gradio UI for?

[ionic pumice]

uvr ui

[thorn plank]

Does anyone have a storyteller type AI voice

?

[crude portal]

Elevenlabs ai is good

[crude portal]

As I know it’s just a UI front end used for easy development of web application or chatbot kind of things

[ruby sun]

YUPPPPI

[lapis onyx]

Hi

Gm

[ruby sun]

HELLO GM

[short elbow]

Hi

Can someone tell me why character consistency is difficult?

Or anyone know how is this possible?
And how the characters can have customised in faces and features as per input image?

Is it possible??

If not why not?

If yes then how?

Do anyone know any tool.

?

[elder willow]

https://www.instagram.com/share/reel/BBq6TazPsG

which ai is this??

https://www.instagram.com/share/reel/BARkum-9UE

Or this

[delicate nova]

anyone using the rvc via colab ?

I have a problem

[chilly lake]

some are broken

applio ui and noui should work

[wind stratus]

a colab update broke some notebooks

of all kinds of AIs

[light girder]

crazy there is still no refinegan model out rn

not possible for me to try out

[foggy totem]

hallo

[delicate nova]

😦

I was using it for real time voice changing any alternatives in games my gpu was too weak for it :/

[balmy tusk]

Hey I installed okada tts, how could I get new models for it?

[foggy totem]

How to I find ai voices from some ppl?

[brittle sky]

I made a cover with scott knuckles and jennifer amy models and it sounded so awful

[tepid basin]

#1175430844685484042 or https://weights.com

[tepid basin]

If youre using RVC, use these for models ^

[balmy tusk]

yeah I use the voice models from that channel but It doesnt seem to work with tts version

or you only find tts ones in that website then?

[cobalt coyote]

Hmm

Hello

[balmy tusk]

yep pretty much dead chat

[cobalt coyote]

Fr

Idk y

[dapper drum]

Question, do you really need to normalize the backing vocals?

[tranquil lantern]

is this a mixing question

I'd say yeah, normalize the backings for consistency, but don't let it be as loud as the lead vocals

because backings supposed to be subtle (depends on the music though)

[idle dagger]

hi, im fairly new to making my own AI but I know how to code,
i want to make an ai for a game i made in python using pytorch. I trained some models and even got claude 3.7 and it made a completely new structure but it still sucks
Could someone help me to train a good model and maybe explain to me a couple of things?
Might take an hour or two

[gray rover]

@ancient swan Hmmm, got lost a lil in the doc but, is the dim_t conversion / necessity to align it, still a thing?

also how about the ui's overlap?

so, to be more exact, say I wanna use the ' dereverb_echo_mbr_fused v2 '
now, if to assume we aren't meant to tweak or align anything related to dim_t, but focus on chunk_size, it's 352800 samples, that'd be / 44100 = 8

So now, most important question:

1. Config has " num_overlap: 4 " should it be aligned with the yaml's chunk ( 8 ) or nah?
2. Does ui's overlap have any kind of effect? ( as in, override the yaml ) Specifically these two:

edit1: ( Hmmm.. seems like advanced tab's overlap does work I think )

edit2: Correct me if I'm wrong; I read up a lil more and it seems it can go both ways for most rofo if not for all recent ones?

• chunk_size = 352800, dim_t = 801, segment = 8
• chunk_size = 485100, dim_t = 1101, segment = 11

and I suppose, a bonus,
3. Is overlap's misalignment a major issue? ( any form of degradation / discontinuity happens? )

[blissful wren]

Hello

How I put on a voice changer

[arctic meteor]

hola

[cobalt coyote]

[covert lake]

Tell ur PC GPU in #🔍│help-w-okada

[inland plover]

How I can optimize voice changer

I have amd ryzen 5700 with radeon graphics

[covert lake]

Show a screenshot of ur wokada in #🔍│help-w-okada

[subtle nova]

where i download the voice changer?

[covert lake]

tell ur pc gpu in #🔍│help-w-okada

[pure ice]

Hell o

[zinc jungle]

How can I download a voice for a voice changer?

nvm i think I got it

[versed jetty]

ai

[cosmic sage]

guys, what's wrong with the weights site?, it stopped working, error Internal Server Error

[thorny frigate]

idk

[viscid quail]

Hello guys
i need some help to find something !!
int his video at 0:55 i love the way the voice is
do you guys think its possible to find an ai that can do the same kind of voice with an accent
https://www.youtube.com/watch?v=4sA3TyjS3k0

[pure ice]

Hey I created a VC for myself can I get approved

[polar flax]

chunk_size = 485100 corresponds to dim_t = 1101
overlap doesnt relate to them, still the higher the better consistency (less noticable cut parts) unless you are using the older patch. for most cases 4-8 is enough and power of two numbers are desirable (otherwise the higher number like 50 could decrease the SDR).

[acoustic patio]

augh

[pine acorn]

Staff Applications Open

---

We're looking for dedicated team members to help grow and manage our community! If you're passionate about AI and want to contribute, apply now!

Click here to apply!

[polar flax]

https://tenor.com/view/yap-yapper-yapping-what-huh-gif-17267607410251793922

[acoustic patio]

https://tenor.com/view/morgan-freeman-true-morgan-freeman-true-nodding-gif-13973817878387504960

[solar torrent]

VC of which?

[polar flax]

get approved of what?

[solar torrent]

Are you developing your own fork of W-Okada or you trained a voice model and wanted to get approved in this server?

[pure ice]

@solar torrent no I'm trying to create a custom voice channel

[solar torrent]

Ehh, if you mean by #jointocreate, you'll have to be in a voice chat for a while and then you'll get approved to create a custom voice channel in this server.

And also please use the reply feature instead of pinging me in a separated message.

[shut sage]

Why the #promo channel has been removed but maybe because of the copyright?

Or maybe it's just changed the name/title?

[ancient swan]

keep overlap at 2, i think no allignements are needed, you can just use whatever values the configs are coming with

[covert lake]

promos ain’t allowed anymore

[shut sage]

Oh thanks ☹️

[hollow dagger]

Ramadan mubarak

[ionic pumice]

ramadhan mubarak to you too, talha

[humble jackal]

where do i download

[covert lake]

elaborate:

• what you want to download
• what's your pc gpu
• what do you want to do

[humble jackal]

just download the voice changer in question

[covert lake]

realtime voice changer for calls?
also, what's your pc gpu still?

[humble jackal]

yes for calls

and ion got a gpu

[covert lake]

are you really sure? you got a pc right?

let's talk in #🔍│help-w-okada

[humble jackal]

yes

[gray rover]

tho, according to yaml values overlap, mathematically would be either 8 or 11, according to models I use. So why 2?

[ancient swan]

people have tested different overlap values and their speed and their both audible and scores quality changes

and the only thing that improves when you put something like 8 overlap instead of 2, sdr rises by 0.1-0.2, but no audible differences

so putting anything higher than 2 is just not worth it

same with batch size, it doesn't affect anything so now everyone just uses batch size 1 for inference

higher batch sizes should increase the speed of inference at the cost of more vram being consumed but that just doesn't happen for some reason

[gray rover]

I mean, I did notice different alignment of low frequency range / respiratory range, becomes less " chaotic "

the higher the overlap

I tested 8, 11 and custom 16

speed however doesn't matter for me, you know me, I favor accuracy over anything else

[polar flax]

it is only slower but still has virtually same vram (4 GB on MSST colab), unless batch_size is changed

[gray rover]

oh yeah, also In higher overlap and settings generally, some clicks are ignored ( background related clicks )

[ancient swan]

if you can hear the difference then go for it what can i say, i can't really spend 20 minutes inferencing an audio that will sound mildly different from audio inferenced in 12 seconds

don't use it that high with apollo models though, it'll ruin them

[polar flax]

use batch_size = 2 in the model config file and there wont be noticable clicks/pops

[ancient swan]

nah, clicks and pops now go from overlap setting only i'm pretty sure, i think jarredou fixed a bug that was causing batch size 1 to cause some weird clicks and pops

[gray rover]

Well then, I'll stick to formula values. I can wait 3 mins no issues, given I extract few mins by few mins, not all at once ( my workflow )

was just curious on proper methodology / mathematically correct settings

[polar flax]

in the zfturbo script there are no clicks for batch_size = 2 and any overlap value, but the higher overlap the less cut parts and more consistency (less noticable on >8)

[gray rover]

I don't use msst btw

doesn't work for me, uvr does, so

[ancient swan]

yeah, but i think batch size 1 now also shouldn't cause any problems like it used to

[gray rover]

btw, does batch have any effect on inference?

[ancient swan]

still couldn't make it work? hmm, i wonder what could be causing it not to work

[gray rover]

tried 2 in models where 1 is in config and tbf, absolutely no difference

[gray rover]

yeah, no budge

[ancient swan]

pretty sure it doesn't

[gray rover]

maybe I'll reinstall all the packages but then, probs no point given uvr works just fine

unless msst does bring audible difference ( which I think it doesn't cause why would it

[ancient swan]

if you match the settings then it should be the same

[gray rover]

ah ye

[polar flax]

ig it should work

[gray rover]

Also, I'm curious.. Anyone already tested gpt-sovits v3?

released yesterday iirc

( I personally haven't tried it at all, incl. v1 and v2 but supposedly, v3 compared to v2 doesn't " average ish " over fine-tune dataset and is more faithful to reference audio )

There's also Zonos that came out and ngl.. that thing is instantly outstanding for zero-shot

You can test it on the site: https://playground.zyphra.com

And repo's in here:
https://github.com/Zyphra/Zonos
( There's also an exp windows fork, just need to read the desc )

[lethal merlin]

@hollow quail

@daring tundra

[hallow lynx]

how can i use the voice model

pls help mee

[night lake]

Your email is on the bottom left btw

[gray rover]

oh, lol

good point, but then, that's the one I barely even use, almost tras/risky content-dedicated one with fake info on it but, good point still 🤔

@night lake Either way, you should give it a try, it's actually op and actually outputs / uses 44.1khz

compared to whatever else

[night lake]

Really?

Will try then

[gray rover]

yeeee

Currently trying to get it locally

[night lake]

But imo tts is overrated

[gray rover]

from what I've read here n there or on their disc, windows exp has some issues with some jp related stuff

[night lake]

Any amd support?

[gray rover]

Haven't seen any amd mentions

unfortunately

[night lake]

Sadge

[gray rover]

Butttt, from all zero-shot I tested

only that one was capable of handling Esdeath's voice

In any case, looking forward to it and potentially fine-tuning, in future
( currently not supported

[night lake]

That's wild is her voice that unique

[gray rover]

No no, that's not my point

my point is, it's super robust, compared to sovits / F5 / E2

[night lake]

You mean accurately replicating?

Oh

[gray rover]

yeyeeee

As for tts being overrated.. well

voice to voice is amazing and all

but if you're like me and want to, in future, play around roleplaying llm + stable diffusion + input voice recog. / tts output ( silly tavern ish workflow pretty much

that's some nice shit

[night lake]

Hmm I see

[ember scaffold]

where is the screaming emoji

[pine acorn]

Staff Applications Open

---

We're looking for dedicated team members to help grow and manage our community! If you're passionate about AI and want to contribute, apply now!

Click here to apply!

[covert lake]

lfg ?

[warped pagoda]

dont mind if i ask , is there any website that generated ai music covers?

[covert lake]

ai covers or u mean just the music?

[warped pagoda]

AI generated music

[covert lake]

you can try weights.com, suno or udio

[warped pagoda]

got it

Thanks

[covert lake]

yw

[warped pagoda]

yw?

[covert lake]

it's an abbreviation for "you're welcome"

[warped pagoda]

so can you extend a song in weights? like upload an audio and extend the song from a specific time stamp?

[covert lake]

you can try using a reference track

[warped pagoda]

like instrumental reference track?

[warped pagoda]

nvm

[amber oriole]

пашлі нахуй 😇 😇

fuck you🫡 😘 😇

[deft surge]

bizarro

[stoic saffron]

@flint zenith never come back. thank you.

[deft surge]

🥶

[exotic arrow]

hello im cris nice to meet eevryone

im need help make a good ai cast for my realities shows of big brother or survivor

[wispy anvil]

Hi everyone, I hope you're all doing well! I'm looking for a specialized website that allowed users to upload recordings of their voice to create a personalized voice model saved online. This platform even let you install different AI models and use your cloned voice to sing over songs. Unfortunately, I lost the link and can’t find it anymore. Has anyone come across this site or know of something similar? Any help would be greatly appreciated. Thank you so much!

[gray rover]

@covert lake Hmmm, you up?

hmmm.. about to test gpt-sovits but 🤔 curious if 3-bert should be empty? Tried to find anything on that but, nope, no info

guess I'll try first lol ( edit, all works ye. Now to figure out optimal settings.. gosh

[shut sage]

Why the promos ain't allow anymore its because of copyright right?

[gray rover]

I believe that's the case yeah. I've seen some people started to get account claims on content they uploaded even months ago n stuff

[shut sage]

Oh okay

[solar torrent]

Likely some people here kept promoting some random job applications in promo chat with no trace of origin.

[gray rover]

oh, it got that bad?

[shut sage]

Oh yeah

[polar flax]

#✦│chat message

[gray rover]

I know such things been here for a while. I meant it more in a " it got that that bad? " manner

[burnt turret]

someone should make a song maker based on the artist

like for example "Lady Gaga Song, Synthpop

and u put the lyrics

and u can upload audio files for ideas

up to 5

[solar torrent]

Really? If it comes to DAW music making software, nah I just suck at replicating other artists styles.

[burnt turret]

i'm making a lady gaga album

lyric snippet for song on album "she's breaking the system, breaking it down"

[solar torrent]

Laga Dyga 🔥

[gray rover]

lady dyga xDDDDDDDD

sounds funny af in polish

[burnt turret]

this one my friend did and they won't tell me what they used

[dry pagoda]

Can anyone guide me how to install the voice model once I have downloaded it from voice-models .com ? I am using tortoise tts btw

[solar torrent]

https://cdn.discordapp.com/attachments/997036654290292776/1292881925794435154/Edge_Yourself_Rizz_Records_Drake_2011_preview.mp4

[covert lake]

Bc of scams and other things that would have kept it hard to moderate

[shut sage]

Oh yeah okay

[covert lake]

Sorry but I don't know much of Gpt so vits

[tranquil lantern]

@chilly lake Yo when next (Kaggle) Applio update so the spilt infer bug would be fixed?

[leaden parcel]

I have a voice model that isn't in #🔍│find-models or #1175430844685484042 Link is here: #1345696074005876808

[chilly lake]

next official release? dunno

you can always clone the latest code

[covert lake]

@chilly lake is it me or this looks like applio with built in music separation

seems like an applio x aicovergen/aicover maker lol

got nice models tho

[gray rover]

well, rvc originally had uvr built in

so no surprise someone decided to bring it back

[glad nebula]

welcome back mainline

[covert lake]

yeah

[covert lake]

mainline with a cool ui

ngl I like applio's ui

it got a cool theme and easy

[chilly lake]

yeah, it does

[covert lake]

oh well the colab is dead bc of using uv

[chilly lake]

applio has uvr plugin as well

[covert lake]

oh cool

I never tested applio's plugins

[bright arch]

@covert lake is mainline back

The colab I mean

[covert lake]

no

when it will be back, the engineer will tell me and #📰│dev-updates will be updated

[bright arch]

Any estimates though

[hidden wasp]

duks

[covert lake]

idk when Hina will update it

I can just suggest u to use Applio for now

[versed hill]

hello I need a model for using it on my male voice to turn it into women voice

[covert lake]

tell ur pc gpu in #🔍│help-w-okada

[versed hill]

ım using colab till my rtx 3070 came from my order

[covert lake]

let's talk in #🔍│help-w-okada

[winged burrow]

This guy got 19k subs with a single video. https://www.youtube.com/watch?v=r2OckF3Gk_8&t=210s
What AI Image generator is he using to create the black and white animated pictures, any idea? Please@me if u reply

[dusk oriole]

Can someone make a big scarr ai voice pls😭🙏🏻

[pale talon]

⊣∷ᒷᒷℸ ̣ ╎リ⊣ᓭ! ╎ ᓵᔑᒲᒷ ╎リ !¡ᒷᔑᓵᒷ, ╎ ᓵᔑᒲᒷ ⎓∷𝙹ᒲ !¡ꖎᔑリᒷℸ ̣ ꖌᒷ!¡ꖎᒷ∷-452ʖ, ᒲ|| リᔑᒲᒷ ╎ᓭ ᒲᔑ∷⍊╎リ ᔑリ↸ ╎'ᒲ ᔑリ ᔑꖎ╎ᒷリ.

[night lake]

purple

[ionic pumice]

minecraft enchanting table

[elder willow]

where do i msg if im trying to find this ai voice i found this one on this video and im trying to find it so i can change it a bit for a use on this edit

[spiral junco]

hey guys, so I am running the AMD fork of A1111 stable diffusion, but years later, I don't think there has been any updates. Is there any upgrades I can give to my Stable diffusion to get better outputs?

I am using an 8GB VRAM GPU (yea, shush), and I am using the DPM++ 2M Karras, but i wana know if there are better stuff

[polar flax]

that should be enough for SDXL and 1.5

flux models could give better results but 12 GB vram is bare minimum for it and Nvidia gpus are more recommended as idk if the rocm version could work well

[spiral junco]

imma be real, I don't use the SDXL nor 1.5 checkpoints.

i use some anime checkpoints. However, Imma need suggestions to expand my negative prompt and some prompts to improve finger issue which I noticed some AIs have now "resolved that" this year.

plus, would hypernetworks help or the styles.csv ?

[polar flax]

it is actually finetuned from the base SDXL/pony/1.5 model

in the civitai page it should tell if it is based on which SD

[chilly lake]

is that DirectML one?

[pine acorn]

Staff Applications Open

---

We're looking for dedicated team members to help grow and manage our community! If you're passionate about AI and want to contribute, apply now!

Click here to apply!

[spiral junco]

Yes

[chilly lake]

SD.Next would be a better option

but it requires some effort to install hip sdk and tensile libraries

what's your gpu?

[spiral junco]

rx 6600

[chilly lake]

okay, there are optimized libraries for that

hip sdk 5.7

https://github.com/vladmandic/sdnext/wiki/ZLUDA

https://github.com/brknsoul/ROCmLibs/blob/main/Optimised_ROCmLibs_gfx1032.7z

[spiral junco]

i mostly use a checkpoint called perfect world which I just checked and its base model is SD 1.5

I use A1111 SD

I also noticed I have some LoRas of certain anime characters, but when using the LoRa, the output image ends up totally destroyed, as if the dataset somehow got contaminated by NIGHTSHADE (fugg nightshade)

[chilly lake]

too much lora weight?

some are just shitty

[solar torrent]

You've just got trolled for using certain LoRA models lmao. Unless you're using some badly trained Stable 1.5 models you downloaded from somewhere.

[spiral junco]

that is true

[solar torrent]

You could've download the good ones from CivitAI instead. Make sure you don't just click the download button before you use it, try look at some generated image posts under the model to see if it generates good or not.

[spiral junco]

which i have been doing even since I started using SD

[elder willow]

Hi

[solar torrent]

For anything about W-Okada, go to #🔍│help-w-okada.

[crystal eagle]

okok thank youu

[marsh fable]

Hiii so im trying to change the lyrics with suno for the song, suno gets the verses right but ruins chorus

[covert lake]

#✦│chat message

[solar torrent]

No, thanks. This is not where you promote your thing.

[gray rover]

Hmmm.. funny enough, gpt-sovits has vram leakage

at least in v3 sovits training department, ain't sure about the rest

[cold hearth]

good morning to all members i have a question how you manage to have cleaner ai cover voice

i tried to converts vocals using weight g models but it fails the vocals have some glitch etc also the result obtain is far from the selected artist

[solar torrent]

This is not where you post an entire codebase.

[river verge]

DIlly ding, dilly dong! A new RegalHyperus drum model just released!
Sweet Dreams (Are Made of This) (PuppetMaster cover) (Drum model no. 581)

[polar flax]

cough https://www.techspot.com/news/106983-new-ai-voice-technology-alters-accents-indian-call.html

[elder willow]

scammers about to get rich

[distant thorn]

Uh oh

[polar flax]

https://tenor.com/view/sh-quiet-india-indian-tiktok-gif-16906342

[solar torrent]

Imagine scammers using "original" W-Okada for phone calls.

[elder willow]

How ai works here

[solar torrent]

Which AI? Each AI works on its own.

[elder willow]

This ai chat

[solar torrent]

#🧬│ai-chat is basically just another normal chat.

[covert lake]

it depends all on how the model was trained

@warped dune we allow only english here

[river adder]

hm thats a little trivial but u could give mel band a shot on https://mvsep.com/

[river verge]

DIlly ding, dilly dong! A new RegalHyperus drum model just released!
Energy (Drum model no. 582)

[pale patrol]

ai chat was verified chat before weights took ai hub :/

and the lounge is general chat

[elder willow]

czy ktoś z polski może zrobić model andrzeja dudy

[stuck ridge]

sup yolo

[solar torrent]

Are you sure you remembered it right? Because #✦│chat was once named the verified chat.

No, thanks. If there's no game demo, I ain't gonna take a survey anyways.

https://tenor.com/view/i-saw-what-you-deleted-cat-gif-25407007

Well, he deleted it real fast. I could see that.

[kindred kelp]

I'm hiring prompt engineers. Anyone interested?

[solar torrent]

Why are you hiring a prompt engineer?

[kindred kelp]

Because I'm a prompt engineer and my workload is too heavy at this point I have about 7 pending projects I can't finish alone. I offer compensation for time, subscriptions to premium services (v0.dev), hourly rate, and training

[solar torrent]

https://cdn.discordapp.com/emojis/1016021511913603202.webp?size=48

[kindred kelp]

DM if interested thx. Sorry if this wasn't the appropriate place

[uncut forge]

Hai

[prisma sage]

Mel loss goes up to 30, I'm going crazy, I can't figure it out

[chilly lake]

mel loss should not go up

[prisma sage]

how

The sources I researched say that if it increases, the naturalness of the sound will deteriorate.

So what can I do to increase it?

[polar flax]

you're more likely to have messed up some training parameters and collapsed the model
normally the default rvc parameters and decent pretrains shouldn't cause that issue

[prisma sage]

I don't know, I will try writing code to fix loss mel

[polar flax]

which code are you working on?

[prisma sage]

I will make changes in train.py where mel loss is written.

[chilly lake]

One more time - mel loss does not go up.. ever

it has to converge down to some value

[polar flax]

you prob meant g/total?

[chilly lake]

g total may go up if fm loss goes up

[prisma sage]

thank you

[magic hazel]

Guys, what's the best way to mix songs with AI for free? I'm trying to combine a music track with a vocals track to make it sound perfect.

I'm trying to find an AI that can perfectly sync vocals to music in the best way possible.

[solar torrent]

There's an AI-assisted track mixing website that was once used by a team of Kanye East, I just don't remember the name. But this is what I can do about mixing.

[mellow spear]

Hi

[polar flax]

https://x.com/haoailab/status/1895557913621795076

Hao AI Lab (@haoailab) on X

Claude-3.7 was tested on Pokémon Red, but what about more real-time games like Super Mario 🍄🌟?

We threw AI gaming agents into LIVE Super Mario games and found Claude-3.7 outperformed other models with simple heuristics. 🤯

Claude-3.5 is also strong, but less capable of

X · Feb 28, 2025, 7:33 PM

[pale patrol]

breh

[prisma sage]

INFO:testoo:loss_disc=4.044, loss_gen=0.000, loss_fm=0.000,loss_mel=0.100, loss_kl=0.100 These are my model values. After working for a day, I coded it myself and achieved a great result 🙂

[obtuse flax]

Damn what happend to Mike Dean Doing the mixing

[pine acorn]

Staff Applications Open

---

We're looking for dedicated team members to help grow and manage our community! If you're passionate about AI and want to contribute, apply now!

Click here to apply!

[solar torrent]

Nitrous also got him.

[obtuse flax]

Oh yea, I forgot about that

[dapper drum]

What’s the curling best model for duets

[prisma garnet]

I have extreme AGI

[reef hound]

Hii guys we are organising a hackathon in our college and looking for monetary sponsors is this the right place??

[kindred kelp]

I'm hiring prompt engineers for $12/hr 30 hours a week if interested

[kindred kelp]

Tell me more

[kindred kelp]

I also tried to make an agi

[winged ridge]

Why does voice challenger work like a bug for me, like it says how...... it's going and at the end it's Korean?
speaks in syllables

[past token]

aye anybody know if its possible to use raltime ai in dc

[trail storm]

GE

[next shard]

guys whats the best tool to make a tts in hebrew with trained model?
like elevenlabs can make it kinda realistic which is great but they don't support hebrew

[polar flax]

bro left to improve his DEX

[magic hazel]

Question, what's the best AI program that can create really good Bas Relief images from a source image?

[urban bridge]

Hi, I love what you do, you don't have to answer if it bothers you, by the way I have a project and I need to do a study: “What are the 3 most repetitive and boring tasks in your business right now?”. Thanks

[polar flax]

https://www.youtube.com/watch?v=PD76HCowEvI

[kindred kelp]

Lol.

Any prompt engineers I can train and hire?

[gentle isle]

hi dude

[grave crystal]

hey you don't bully my friend

[gentle isle]

fine......... ughhhh i hate this......

[grave crystal]

that's what i thought

[solar torrent]

When did I bully your friend?

[deft surge]

Tecnologia bolsotron

Patriotron

Novo projeto da Microsoft

[solar torrent]

https://cdn.discordapp.com/emojis/870264781276348416.webp?size=48

[deft surge]

🥶 🥶 🥶 🥶

American safado

Donald biden faz nada

Crime happens mc Donalds

[solar torrent]

Please speak English.

[deft surge]

American crime hepens joe trump make nothing mc donalds

🔥

[solar torrent]

I still don't understand what you're trying to say, but let me make that readable.

Nah, I don't talk politics here.

[deft surge]

Im eliot scott Obama

[solar torrent]

Sigh.

[deft surge]

Rayman

Fire 🔥

☝️

🖕

✋

🫰

[solar torrent]

Please chill with your emotes.

[deft surge]

Im chilling with my brother emote 🙅 🧏 🤙

[polar flax]

https://tenor.com/view/cat-cute-cat-yap-yapper-yapping-gif-5642199211123099306

[deft surge]

Im canada

[deft surge]

Mexico peperoni

[solar torrent]

No, you aren't reading my messages that right.

[deft surge]

You talking about oxygen

🐴

[solar torrent]

I don't do crack, sorry.

[deft surge]

Metanfetamina

Walter white

🤫

[solar torrent]

😐

[deft surge]

Bloodborne

[solar torrent]

Well, that's pretty much it, kid.

[deft surge]

Hehe

Fortinite

[solar torrent]

Fuck.

[hard panther]

why does ai art still suck ass

[rare shoal]

Is there something similar to Mellotron but newer?

[covert lake]

Explain:

• your PC GPU
• what guide link u followed
• the issue
  In #🔍│help-w-okada

[covert lake]

We don't allow promos here

[covert lake]

Explain:

• your PC GPU
• what guide link u followed
• the issue
  In #🔍│help-w-okada

[covert lake]

There are different Text To Speech (TTS) AIs:

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

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

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

You can check TTS in our tts index

With RVC Models:

RVC is natively for Speech To Speech, but forks such as ilaria rvc mainline & applio have built in tts (using Microsoft Edge TTS to make a generated tts audio, which i suggest you to choose a tts model that is the same gender and language of the rvc model you wanna use, and then convert it with rvc)

If you wanna do tts locally with RVC Voice Models (if you got a good pc):

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

If you don't got a good pc you can do tts with RVC Voice Models on cloud:

• Ilaria RVC Zero (Running on A100 GPU, free fasted rvc on cloud) and the guide

• Use Applio UI Colab (with google colab T4 free daily limit gpu)

• You could try another tts from our tts index and use the output as an input in rvc

Maybe you can find something better in our tts index, but I wouldn't expect that much

[polar flax]

similarly, do you still hate DLSS?

[hard panther]

dunno what that is

[covert lake]

Maybe you can try flux.1-dev since it's the best open source image model, LoRAs might help

[covert lake]

Use better models and prompts

[hard panther]

no

[deft surge]

Balls

[covert lake]

then yeah that it won’t look good if you don’t use good models like flux

[solar torrent]

That's a true real fact. I like real art that made by human better.

DLSS is a feature found in newer NVIDIA GeForce RTX GPUs. This feature uses AI technology to generate some frames for a game you're playing, rather than using raw power to process.

If you ever feel like you wanna compete with AI fans about art drawing, I suggest try to pick a pencil, draw anything you like and tell me what you feel about it. ✅

[covert lake]

We weren’t comparing it with human art btw

[peak grail]

i was afk while my model trained the and the runtime after it finished where can i find the pth file

[covert lake]

#✨│ai-help

[polar flax]

https://ollama.com/library/qwq

[cobalt coyote]

humblo

[covert lake]

check #📰│dev-updates

[next shard]

does anyone know any good realistic ai for tts in hebrew that is realistic and emotional i also can get trained models in hebrew if that helps and referance for emotions of someone speaking to speak like him

[covert lake]

I think I gave you some TTS options #🧬│ai-chat message