🧬│ai-chat

[acoustic sorrel]

👋

[noble tulip]

HII

[feral halo]

Ayo? @elder willow level 2 !!!

[velvet timber]

Alguem Br ?

[earnest dragon]

Are you using the webui

[tired jasper]

i'm using vst3

i guess vst3 have better control and speed

[earnest dragon]

Vst3 are not gpu accelerated unless gpu audio plugin and are you using Beatrice v2

[earnest dragon]

The vst3 is only for inference not creating

[tepid vale]

[earnest dragon]

Just use mvsep bs reformer viperx

[tired jasper]

but i'm using vst3 for inference

vst3 is not for training, what you mean?

[earnest dragon]

Then what are you using for training

[tired jasper]

beatrice trainer

[hard trail]

Hello epta

[brave holly]

anybody here know any programs similar to Wav2Lip for easy deepfake lip syncing using the voice of the person in the original video? Wav2Lip has some issues when I get to the last step

[feral halo]

Ayo? @elder willow level 3 !!!

[vestal temple]

Heyy I am f please

[pseudo granite]

just saw the correct channel btw

but ty

[native saffron]

Interesting

[tacit falcon]

what the fuck is even happening in this channel dude

[grand relic]

Idk

[feral halo]

Ayo? @grand relic level 2 !!!

[earnest dragon]

So the webui version

[earnest dragon]

Facefusion is good for free but still not as good compared to the paid ways

[solar torrent]

AI music isn't the real art, it is a form of AI art.

[brave holly]

I see, what would be the best paid versions?

[earnest dragon]

Talkingavatar.ai https://www.talkingavatar.ai/r/HZNCL5RO is the best I have used but over priced but good and for open source free then facefusion

[brave holly]

Awesome thanks

[feral halo]

Ayo? @brave holly level 1 !!!

[remote star]

hiya

[deft merlin]

Hey guys, are we allowed to share prompts, new here. Looking to share some prompt and/or learn some new ones

[orchid portal]

Guys I am gay

Guys I am gay

[left hawk]

hiii

[tired jasper]

No

[earnest dragon]

Then you are not doing it the correct way then

[tired jasper]

🤣

[earnest dragon]

https://github.com/JarodMica/beatrice_trainer_webui

[earnest dragon]

Because the colab way is not as good and if you have a Nvidia gpu then local webui way is the best

[tired jasper]

I will give it a try, but I haven't seen many changes from original beatrice trainer. He says that it runs in 3rd gen nvidia cards, but he set batch size to 32. The maximum i got to run good was 10, in 12 it start to share mem, this is weird to be so high, but i will give it a try and see what happen.

🔥 💳

[buoyant star]

his trainer is based on v2 beta 0 which is old

[sudden anvil]

п

[buoyant star]

He uses an rtx 4090, so he was able to set the batch size to 32

[shut mantle]

hi every one! i'm new in there

[earnest dragon]

Is it

[buoyant star]

yeah, it was updated four days ago

[cold lintel]

nvidia 4090s aint it

[earnest dragon]

Yeah but also colab is older

[earnest dragon]

Why

[earnest dragon]

I thought he uses 4080

[cold lintel]

i just dont think their that good compared to other alternatives

[earnest dragon]

What alternatives

[buoyant star]

the colab version pulls the latest version from hugging face. I tested the colab version yesterday, and it uses the latest beartrice code.

[earnest dragon]

Okay but colab is not good local is the way is there any local webui with latest

[buoyant star]

I don't think so. People don't even bother with Beartice

[earnest dragon]

Yeah because they don't like switching even when something might be better

[buoyant star]

I would do it if they included more languages in the pretrained model

[weary gale]

hola

[nova peak]

Mind if I self promo? ✅ Walker Extreme yo

[earnest dragon]

No only in #1159290752195633273

[brave holly]

Talkingavatar seems to be super new? I can hardly find any info about it online besides from them

[viscid arrow]

.

[earnest dragon]

Yeah it is super new and not that known I have examples I have made

[granite idol]

is it allowed here to post telegram links?

[earnest dragon]

No

[stark plinth]

Hi guys can someone suggest an ai voice fir journalist youtube channel

[brave holly]

Sure would be curious to see them

[covert lake]

#1159290752195633273

[earnest dragon]

Dm'd you all them

[polar flax]

better avoid posting possibly illegal links

[covert lake]

he left 😭

bro joined just for promo

telegram piracy 🙏

[paper hinge]

Anyone else having problems with the Applio Colab (NoUI) and Pytorch?

[covert lake]

What’s ur pc gpu?

[paper hinge]

It’s on Google Colab, my PC GPU shouldn’t matter, I think

It just started happening yesterday

[covert lake]

yes it doesn’t as its cloud, but many people got a good pc and don’t know they can do it locally

so that’s why im asking that first

there are people with an rtx 4070 using colab, and people with a miserable integrated graphics trying to do it locally 😭

[tacit falcon]

couldnt a ryzen igpu probably do it to some degree

lol??

or at least infer

[covert lake]

integrated gpus suck

[tacit falcon]

yeah but im not talking about doing this with an igpu practically

[covert lake]

I think it would be just using the cpu

I got an intel cpu so igpu

[tacit falcon]

intel igpus suck but i dont think ryzen igpus suck as much

[covert lake]

still doesn’t have cuda cores, and im not sure if zluda could work with igpu

@chilly lake would zluda work for ryzen igpu?

[tacit falcon]

im not really thinking of doing it but im just wondering out of curiosity since ryzen igpus are not super super super terrible

[covert lake]

100% u can use cpu

[tacit falcon]

but they are not super super good either

[covert lake]

its just not recommended

as its slow

[dense heart]

whats the new AI hub doc link

[covert lake]

#📰│dev-updates message

[solar torrent]

#1159290752195633273

#1159290752195633273

[pallid harbor]

yo

[ionic pumice]

would any1 like a bsky code

[pallid harbor]

what is that

[ionic pumice]

bluesky

[pallid harbor]

slide

@ionic pumice

[ionic pumice]

meaning

[pallid harbor]

send

[cunning ore]

There is a video that remixes the existing MR with just tags, does anyone know what AI it is?

[elder willow]

@covert lake

[covert lake]

Please don't ping me much

I dunno about that AI unfortunately

[cunning ore]

I have a link to the video on YouTube, so would it be helpful to show it to you?

[elder willow]

sorry

[covert lake]

It's fine dw

[covert lake]

From what I understood it's about remixing music, but I fr don't know any ai that does that

[cunning ore]

I understand, but thank you for your help

[covert lake]

Wish you good luck to find one, only I know is https://replicate.com/sakemin/musicgen-remixer (paid only) and idk if Suno maybe does that

But never tried both

[cunning ore]

Oh, this is a big piece of information, thank you

[feral halo]

Ayo? @cunning ore level 1 !!!

Ayo? @sleek wadi level 6 !!!

[rare sorrel]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

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

[covert lake]

Your welcome

It's not in the docs

Well..not yet since...

Read https://discord.com/channels/1159260121998827560/1233407331405004954

[austere herald]

how to make pictures with ai

[chilly lake]

lol no

nobody was crazy enough to compile libraries for it

ugh... i was wrong.. someone was, but would not recommend

[visual fable]

Uh where is useful-links?

[snow sedge]

infer would be fine, but training very unlikely

[covert lake]

Lmfao

[chilly lake]

i'ma gonna test it

[covert lake]

Let's just talk in the help forum post u made, no need to ping me about the same thing twice in 2 channels

[chilly lake]

"C:\Users\Administrator" - dont, use a normal user

[elder willow]

how do i fix the error?

[grand relic]

1

[chilly lake]

nope, the libary set for igpu does not work

[mossy sage]

1

[covert lake]

Rip then

Are AMD CPUs really better btw?

[elder willow]

Tbh i rather use colabs

[chilly lake]

it is a choice.. paying for max performance at max power consumption vs lil less for much less. Also DDR5 is better on Intel, better memory controller, usually can use 4 sticks of RAM without issues

but again 13k and 14k have the degradation and corruption issue, probably fixed by now

[sturdy elbow]

Damn, I didn't know that

[chilly lake]

if you had those 13900k for some time, they are probably fked

[sturdy elbow]

I see

[ionic pumice]

okay ! killing myself

[loud kiln]

I got the tools, now I just need a better computer lol

Makiing clean audio for the model training gonna be delish

[polar flax]

arrow lake incoming

[chilly lake]

the release is today

AMD's 9xxx x3d in 2 weeks

[tired jasper]

Who? standard trainer is beta.2

[polar flax]

https://www.tomshardware.com/pc-components/cpus/ebay-listing-sees-intels-core-ultra-7-265k-offered-one-day-before-release-but-its-not-exactly-a-bargain

[brisk stream]

hello

do you have any beatrice model?

[shut ivy]

hello

[tired jasper]

I plan to post overwatch full characters today, idk if i will

[chilly lake]

what are you gonna do with it without a motherboard?

[polar flax]

I'm not sure if Z890 mobos should be launched together with the cpu

[brisk stream]

thats nice but i wonder if people will start to train other model on beatrice

[tired jasper]

I can train any model for you for a few bucks

[chilly tapir]

im gonna try out weights.gg model training!!! what's the minimum duration for the regular training?

[chilly tapir]

-docs

[rare sorrel]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

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

[hasty elk]

is there a channel that allows sending voice messages here?

[silent pewter]

[covert lake]

U need to get level 5 for those perms

[tired jasper]

RuntimeError('Repository root is not found.')

[hasty elk]

oh...ok

[hollow tree]

hi! does anyone know if nova app is legit or n

[dim mural]

im gay

[feral halo]

Ayo? @dim mural level 1 !!!

[ionic pumice]

real 👅

[round walrus]

hi

[solar torrent]

Imagine playing a gacha game at 3am, and suddenly one of Kendrick Lamar's songs plays.

[ionic pumice]

i need him

[solar torrent]

Ellen Joe Blahaj.

[white sky]

I have a request for the player of the fire emblem series. I have a request for a word from Lyn's voice lines, does anyone know the words and can do it correctly?

[strong oracle]

Does anyone know someone who wants to hire an editor?

[covert lake]

#1159289738314919936 or #1191429836321849435

[covert lake]

Maybe in #1159290752195633273 would be the right channel

[strong oracle]

Tysm

[pseudo salmon]

hello there general chat!

[mint nymph]

@orchid slate Hi, can you please help with the setup and with the ai voice version?

[queen kernel]

Hello 👋

[queen kernel]

Have you downloaded RVC ? And di you have a GPU ?

[mint nymph]

well, I have a video card, just which version to download 1.5.3.1 or v.2.0.65-beta

+how can you set up your voice well so that it is not robotic and does not lag

[queen kernel]

Are you using real time voice changer

[mint nymph]

yes

[feral halo]

Ayo? @mint nymph level 1 !!!

[queen kernel]

Which GPU do you have.

[mint nymph]

GeForce RTX 2060 super

[stray merlin]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs :nvidiagpu:

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[queen kernel]

Follow the RTX 20-Series

[mint nymph]

can I write to you in the bos? I'll send you a screenshot

[pseudo salmon]

kinda weird how the server is called "AI hub", yet theres no llm channels

[feral halo]

Ayo? @pseudo salmon level 1 !!!

[queen kernel]

Bos ?

[stray merlin]

Voicechanger Settings (Okada)

⠀

Settings for Nvidia GPUs :nvidiagpu:

F0 Det.: rmvpe (suggested for all series)

RTX 40-series: 80-96 chunk | +16384 extra
RTX 30-series: 96-112 chunk | +16384 extra
RTX 20-series: 112-128 chunk | +16384 extra
GTX 16-series: 128-192 chunk | +8192 extra
GTX 10-series: 128-192 chunk | +8192 extra

Advanced Settings

Protocol : Sio or Rest
Crossfade: 4096 start 0.2 end 0.8
Trancate: 300
Silencefront: Off
Protect: 0.5
RVC Quality: Low

⠀

Made with love by Ilaria 💖

[mint nymph]

what? I just have a strange version and there are no such chunks, there are 2400, 3840, etc.

which version should I use?

[queen kernel]

We work for Voice to voice conversion and We have some helpers regarding stable diffusion also we work for facefusion that's why it's Ai hub because it consists many Artificial intelligence channels

[queen kernel]

Use The Fork

Don't know about this GPU.

[mint nymph]

What does this mean, I'm not good at English

[stray merlin]

Voicechanger Download (Okada)

⠀
Download for Nvidia GPUs :nvidiagpu:
Version 18a cuda

Download for AMD GPUs :amdgpu:
Version 18a directml

Download for Intel GPUs :intelgpu:
Version 18a directml

Download for Mac :macgpu:
Version 17b Mac
⠀

Made with love by Ilaria 💖

[mint nymph]

sps, I just used the 2nd generation version

[queen kernel]

Basically the fork is the modified version of a original program to provide best quality at low specifications

[queen kernel]

Have you downloaded it from the following link ?

[mint nymph]

Now yes

[feral halo]

Ayo? @mint nymph level 2 !!!

[queen kernel]

Maybe your GPU is not enough powerful for it. As I don't know much about you GPU

[queen kernel]

-realtime

[rare sorrel]

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

[queen kernel]

Download first one @mint nymph

[feral halo]

Ayo? @unborn jacinth level 1 !!!

[queen kernel]

It's good to ask in #🔍│help-w-okada

[mint nymph]

Well, I downloaded the topmost one

[unborn jacinth]

ok

[queen kernel]

That's not great. Use fork instead

[molten herald]

Would you guys recommend Okada or RVC?

[mint nymph]

And where can I get it? I thought it was a fork

[queen kernel]

Basically RVC is a program used for voice to voice conversion. And Okada is a live version of it. You can do voice to voice conversion in real-time using okada. RVC is the program and okada works on thr principle of RVC

[queen kernel]

I sent you the link

[half gust]

Hey, I have a question for you guys.

[queen kernel]

Yes please ask

[half gust]

can you create an generative AI from If - else statements alone?

[rare sorrel]

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

[molten herald]

noted i apprechiate it

[queen kernel]

@mint nymph the first one (fork)

[mint nymph]

Yes, I've already found it.

[mint nymph]

deiteris-Fork
/
voice-changer-windows-nvidia-b2309.zip I'm rocking this one

[queen kernel]

No. It's a very complex process to creat an Ai model. Not only if else statements you have to use couple of things such as libraries like tensorflow and many frameworks.
Basically if else is a kind of decision making statements. You can use it to make decisions.
For example
If value = 1
On the light
Else
Off the light

[queen kernel]

Yes it's correct. Follow the guide. And if you have any trouble use #🔍│help-w-okada
This is only for general chat

[mint nymph]

❤️

[half gust]

I'm aware it's solely decision-based and rigid, and that's why I'm going lock myself in a room and drink lots of coffee and figure out how to do it. I like making impossibilities into possibilities.

[feral halo]

Ayo? @half gust level 1 !!!

[queen kernel]

Great. I recommend you to learn python to create your own AI stuff. And it's a very long process. Work on it day by day. You can't achieve it in one day. I appreciate you confidence. Good luck buddy

[half gust]

I've already started learning Python for months now, and by the way, I'm doing this to make AI(s) much easier to grasp by beginners and also because I hate calculus (not Maths, just calculus).

[queen kernel]

I'm not a developer but I hope you will became one

[covert lake]

that's not really how it works, if u wanna check the big messy code u can see https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/

And yea RVC code is pretty messy and left to rot, still the best tool for speech to speech models tho

the creator is working on gpt so vits instead of rvc so rip

For learning python u could check https://www.w3schools.com/python/

For learning about AI coding (with python) u could check the HuggingFace (biggest ai platform ever) docs https://huggingface.co/docs, iirc rvc uses Diffusers, just like Stable Diffusion

No one has found any quality improvement in RVC since a year, and I can assure you it's code is spaghetti and hold by duct tape lol

[covert lake]

This is also known as mainline, the original project

there's also forks (modified versions) like Applio which are slightly better & faster https://github.com/IAHispano/Applio (still no quality change and ofc the models are interchangeables)

[queen kernel]

Haha. Enough nick. He got his answer already.

[covert lake]

well, yea it's mostly just RVC & Wokada

has an help channel for ai art too tho

but mostly Speech To Speech conversion

[queen kernel]

Why are you replying same questions which I have answered? ?

[covert lake]

oh lol, just wanted to be specific

[queen kernel]

🤕

[covert lake]

wtf discord changed the notification sound

[covert lake]

my bad i was just scrolling the channel

[queen kernel]

Never listened that 🤣

[covert lake]

were u thinking of being an helper?

[queen kernel]

Can I be ?

[covert lake]

it's more spooky

[covert lake]

i mean u gotta /jointeam

[tough cove]

hii

[covert lake]

i asked as u seemed to wanted to help people

[queen kernel]

And someone told me they will check my activity in server.

[covert lake]

oh cool

wish u goodluck that u get it then

i just answered to be more specific and that there aren't llm channels here unfortunately

[queen kernel]

Ya I love to do that but in my free time. I'm student. And don’t have much time for it. Whenever I'm free I love to help individuals

[queen kernel]

Hello

[covert lake]

Understandable, i'm a student too

[queen kernel]

We should find few helpers for LLMs that will make this server more useful

[covert lake]

Agreed,
I just meant to say that most stuff here is just rvc & wokada as u can see

[queen kernel]

Well I have mentioned about facefusion too. I remember you ate the only who help for facefusion

[covert lake]

yea i appreciate that, unfortunately i can help only for cloud as i got a bad pc

[queen kernel]

Your specs ?

[covert lake]

cpu =i3 10th gen
ram = 8gb
gpu = integrated graphics

Decent for some gaming but shitty for AI

[queen kernel]

See the facefusion I have mentioned.

[covert lake]

plus i got only 256gb of storage

[ebon ravine]

Anyone knows how to install Tensorboard in the Vanilla RVC?

[queen kernel]

Oh no. I recently upgraded to Intel i7 12700k
32 gb ddr 5 RAM
2TB ssd storage and a RTX 4060

[covert lake]

I mean i could do CPU inference locally (tried fastsdcpu sometimes) but i don't wanna wait hours

[chilly lake]

tensorboard can be installed independently of RVC

and then you just run it and points to a log folder from command line

[covert lake]

damn u lucky fr

my pc is good atleast for some gaming tho

i mostly play indies so

Like portal, undertale, some emulation (3ds)

fortunately i can do that atleast

[queen kernel]

I earned. I saved and I make it.

[covert lake]

Ye im glad for u

[queen kernel]

If you need any help or want to use my resources. You are free to do that 😊

Can I send you friend req ?

[ebon ravine]

Got it

[feral halo]

Ayo? @ebon ravine level 4 !!!

[ebon ravine]

oh

Dammit

[covert lake]

Use your resources? wow i honestly thought u were mad at me bc i replied to the same question u did 😭

[covert lake]

sure lol

[queen kernel]

No I respect my seniors and you have more knowledge than me

[feral halo]

Ayo? @queen kernel level 27 !!!

[covert lake]

Dunno about using someone's resources as i kinda feel bad about that but ofc u can send me a friend req

[covert lake]

How old are u btw?

Cus im just 15

i thought u were 16

[queen kernel]

Honestly, I thought you are 23. But I'm shocked

[covert lake]

damn do i sound old 😭

[queen kernel]

Don't want to reveal in server. Lemme DM you

[covert lake]

Understood

[queen kernel]

You are mod..

I think mod is a great position

[chilly lake]

damn, i've joined a kindergarden it seems

[covert lake]

Most people here are minors

[covert lake]

Are u 20 or 30

I mean i would have guessed that tbh

But yes most people here are just 14 or smt

most staff is minor too iirc

[queen kernel]

What about admins. Tea ? I guess ilaria is 27.

[covert lake]

Dunno their age but surely older more than 18

I dunno ilaria's precise age but should be close to what u said

[queen kernel]

Check her profile. You will get it

[covert lake]

Oh 25 yea

[chilly lake]

that's my professional programming experience lol

[covert lake]

DAMN

are u like

60

[chilly lake]

get off my lawn, junior

[covert lake]

I knew u know lot of shit of coding

but didn't think that much

[chilly lake]

not that old

[covert lake]

well, welcome here

[covert lake]

btw W portal pfp

i just finished portal 1

[loud wraith]

how much space i shloud have for ai cover programs on disc ?

[covert lake]

@chilly lake gradio 5.3.0 (1st image) vs 4.31.0 (2nd image)

i think they adjusted 5

[chilly lake]

what's different? does it have an option to save .wav as file in firefox? 🙂

[covert lake]

dunno about that as i'm on chrome, but it uses SSR instead of SPA

meaning way faster loading

and ofc they changed the default theme colors a bit

someone PRed me about it https://huggingface.co/spaces/Nick088/Audio-SR/discussions/3 and tbh doesn't look as bas as u showed me in that screenshot

u might wanna retry on applio

its not a dev version anymore, its officially out so

[chilly lake]

there's gradiop 5 update coming, I think

[loud wraith]

how to make ai covers and how much space i need?

[feral halo]

Ayo? @loud wraith level 1 !!!

[covert lake]

Oh wait i might have confused u with @sturdy elbow #✦│chat message

[covert lake]

my bad, confused u with someone else

[chilly lake]

we are on the same team

[covert lake]

oh ye u work on Applio too right?

[loud wraith]

!aicover

[covert lake]

cool

[chilly lake]

yeah

[stray merlin]

RVC Colabs and Spaces

⠀

Google Colabs :colab:

⠀
AICoverGen-WebUI
Useful for making quick covers, by Hina.

AICoverGen-NoWebUI
Useful for making covers, doesn't include a UI, by Ardha, by Eddy, Hina and Gdr.

RVC Disconnected
To train new voice models, by Kit Lemonfoot.

EasyGUI
The OG interface, by Rejects.
⠀

Made with love by Ilaria 💖

[loud wraith]

damn

[covert lake]

what's your pc GPU first of all?

[loud wraith]

gpu card?

graphic card?

[covert lake]

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

[loud wraith]

NIVIDIA GeForce RTX 3050 Laptop GPU

[covert lake]

What's the GPU memory?

[loud wraith]

where i can check?

[loud wraith]

i cannot post screen

[covert lake]

U can only in help channels bc ur level 1

Do it #✨│ai-help

[loud wraith]

i did it i even pinged you

[cyan bramble]

what does RVC stand for?

[chilly lake]

some think it is realtime voice changer... but it is not, it is actually Retrieval-based Voice Conversion

[sudden hill]

[feral halo]

Ayo? @elder willow level 2 !!!

[grizzled token]

hello

[barren lichen]

Hlo

@barren lichen

[wispy dock]

https://x.com/ednewtonrex/status/1848741876662436078?t=J0fkqEkEpz0bwWwoVAlL4g&s=19

[feral halo]

Ayo? @wispy dock level 1 !!!

[tender stump]

hi

[tired jasper]

@earnest dragon and @polar flax i have to agree with you about RVC and Beatrice, RVC is much better, but Beatrice is not trash, it's good for normal voices, more for female voices. Hope they improve Beatrice soon

[real lion]

hello huzz

[tender stump]

do not call me huzz please

[tired jasper]

Ramattra from Overwatch 2

now i just need a RVC trainer CLI

to put a queue on characters

[feral halo]

Ayo? @tired jasper level 11 !!!

[tired jasper]

i have seen people complaining about multi voice is trash

i guess can be good if merge after solo train

[quaint jewel]

sup

free video ai sugestions ?

[feral halo]

Ayo? @quaint jewel level 1 !!!

[tired jasper]

Cogvideo

Also pyramid flow

I tried to run locally but they are very heavy

Yet good quality

[quaint jewel]

nice, thanks

im actually a noob at ai, anywhere i can learn it ?

or do u think exposure is my best friend ?

[brisk stream]

Can I dm you @tired jasper

[tired jasper]

huggingface.co

[tired jasper]

Yes

[tired jasper]

You need at least Python language basic knowledge to go deep, but you can always follow tutorials about install GUI versions

[quaint jewel]

then i am done, dont even know wtf is python lmao

[iron light]

wazaa

[polar flax]

the beatrice drawback is the 24k sample rate which is less desirable by audio engineers (don't even rely on the imperfect AudioSR, Apollo, or other existing upscalers so far)
also would better discuss it in #🔊│ai-development

[tired jasper]

The problem in my opinion is not totally about sample rate, but the beatrice input is 16k, too low

[polar flax]

there's RVC in comfyui though haven't tried it

[devout chasm]

How do I get rid of background vocals in UVR

[tired jasper]

With Demucs or manually removing noise with audacity

[devout chasm]

Alright thanks

[tired jasper]

Oh, sorry, I didn't see "UVR", but i guess you can get rid of anything that is not the main vocal with Demucs, it is also multi band, can separate multi instruments

[elder willow]

This ain't no self-promotion - someone else uploaded this compilation! https://youtu.be/Yq-huwHv120?si=lbqQwn1-H3NZRTHg

[quaint jewel]

can i find some tutorial for GUI at huggingface ?

[tired jasper]

I don't see a good reason for using it in ComfyUI

[tired jasper]

Maybe on youtube, huggingface is most for just store models

[quaint jewel]

got it, thanks

[devout chasm]

There's a zillion demucs, which should I download on UVR @tired jasper

[tired jasper]

Idk, i use the original by command-line

[devout chasm]

i don't see that

i cant screenshot it for some reason

[nova pike]

Does anyone here know a good local TTS so I can clone voices and make x famous person speak, so he can say what I want?

[devout chasm]

weights.gg

wait wdym by local

[buoyant star]

try GPT-SoVITS and F5 TTS

[nova pike]

And those programs, in addition to having the typical Asian languages ​​and English, how many more languages ​​do they have to be able to make TTs?

[devout chasm]

lmao i reinstalled UVR

how do i download kim vocal 2 again

ayyyy BVE is back

[buoyant star]

GPT-SoVITS v2 supports English, Chinese, Japanese, Korean and Cantonese while F5 supports English and Chinese.

[devout chasm]

wait it's installed but not in my list?

[feral halo]

Ayo? @buoyant star level 3 !!!

[devout chasm]

nvm i see it

[chilly lake]

for some languages that are not supported by common TTS you either have Edge TTS from MS Edge screen reader, or you need to train a TTS from scratch for your language

[nova pike]

Good idea, I'm going to try it

[chilly lake]

there's a limit on speakers in the free Edge TTS though

well, there's of course 11Labs

[nova pike]

No, that doesn't work for me, since that is based on an already created voice that is neutral, and then my voice model when making inference, sounds like the neutral voice, with the same accent, and what I am looking for is something more like my model having its own voice, like what ElebenLabs has, but local on my PC.

[tired jasper]

I need tu eesleep

I uant tu eesleep

Uai eesleep?

[keen jackal]

can someone recommend a realistic female voice

[tired jasper]

Any voice can be yours for few bucks

[keen jackal]

how many bucks we talking

[tired jasper]

20

[keen jackal]

i dmd you

[feral halo]

Ayo? @keen jackal level 1 !!!

[gloomy bay]

hi guys

[void hound]

l. realb video

oops wrong server

[ionic pumice]

I l. realb video too ...

[covert lake]

#1159290752195633273

[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.aihub.wtf/tts/gpt-sovits/

Freemium 11labs: A 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

If you want actually good tts, use a tts program like gpt so vits or fish speech 1.4

RVC is made for Speech To Speech, no tts

[nova pike]

I'm interested in fish-speech, but isn't there an easy way to install it or do I have to put code in Python?

[covert lake]

what’s ur pc gpu btw?

u should be able to do it following https://speech.fish.audio/

[nova pike]

whats?

[covert lake]

i did a typo, but i meant to ask what’s ur pc gpu

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

[nova pike]

RTX 3070TI 8vram

[covert lake]

yea that’s good enough

[covert lake]

try following this guide

[timid musk]

how do i make an ai cover

[nova pike]

It's a shame that all TTS applications require Python code and there is no executable yet.

[covert lake]

1. Make ai cover
2. Upload it to youtube
3. RIAA comes to ur house

[covert lake]

Well actually it’s pretty normal for almost every single AI project, unfortunately not everything can be 1 click

Anyways they already tell you the code to run so you don’t need any type of coding knowledge to run it

[timid musk]

okay but what if i live in Rwanada do i have to follow the US laws

[covert lake]

if you go into github, basically no ai project has an executable file

at most they have .bat or .sh files that just run the same commands that u should

[covert lake]

Rwanada? When i search it shows a french toast 😭

[nova pike]

How annoying to have to program to be able to use fish-speech

[covert lake]

u mean Python?

well, its an Open Source AI program so

its not that bad, you just have to copy paste code

Tbh, never seen an AI Open Source project with a .exe and don’t think you will ever see one either

[earnest dragon]

Python or cython is the way to go exes are stupid for ai applications unless it's a whole program

[covert lake]

Or github projects in general

It’s also because it’s easier also for security to read the code while for a .exe it would be hard, and .exe are technically runnable only on Windows (well you can still run them on linux with Wine, even if it could have some issues, but it’s better to just give the users the code to install it)

[feral halo]

Ayo? @covert lake level 113 !!!

[covert lake]

never seen a .exe on github

[earnest dragon]

I have before but on github there is always the source code usually

[covert lake]

Oh lol

i think people should know more that almost no ai project has a .exe

Python 🙏

[earnest dragon]

Cython is the best but not as much stuff uses it

[covert lake]

Yea never used it either ngl

What programming languages do u know btw?

[earnest dragon]

https://github.com/cython/cython

[earnest dragon]

Really only basic python because most of the time someone else did the coding for me

[covert lake]

ah its basically same python code, but it translates to C so it’s faster

i would be okay with it by just knowing python right?

[covert lake]

oh understandable

[earnest dragon]

Yeah so we need w-okada cython and applio cython

[covert lake]

Im just a student but i know python & html

[earnest dragon]

Probably

[earnest dragon]

Yeah so someone should convert those 2 to cython instead of worrying about messing with rvc models optimizers and stuff

[covert lake]

@chilly lake what do u think about this ? May u know more about cython

The difficult thing is that rvc is spaghetti

[earnest dragon]

Yeah that's why applio and w-okada not default rvc

[snow sedge]

I already mentioned that python automatically compiles code with CPython since quite a while ago. You do not have to write CPython manually unless you are crazy enough to do that
https://peps.python.org/pep-3147/

[covert lake]

huh python uses that by default?

[snow sedge]

At least since 2010

You can notice that when you run python code, it creates pycache folder with pyc files for each file (and for all its imports as well in their own locations)

That's compiled bytecode with CPython

[nova pike]

What just happened, I left for a second and they started a tremendous conversation

[covert lake]

And there is no difference between Cython & CPython?

[covert lake]

its a sign u have to learn python

[nova pike]

I should, but I don't want to, it's too much work.

[north sphinx]

how can i use instant dataset creator

im looking for a dataset creator that can filter out the vocals and cut silence and everything else

[snow sedge]

Oh, wait, cython. No, that's different, it actually translates python code into C code that is then compiled with C compiler. So it is as if you'd write a C/C++ library with bindings for python, but in a more convenient/python-specific way

[loud kiln]

There's alot of work that goes into restoration, adobe is coming out or came out with an updated ai tool for cleaning up audio, it's still far from perfect but it's not too bad

[feral halo]

Ayo? @loud kiln level 1 !!!

[covert lake]

u could see w3school tuts tbh

[covert lake]

i dont think there’s that, u have to use UVR to clean the dataset

[north sphinx]

ok

fair enough

[covert lake]

So it’s better than CPython?

[snow sedge]

In terms of performance, yes, because code produced by Cython is not python, it's a pure C code. But you will tradeoff portability of such code since it has to be compiled separately with C compiler for each python version

[covert lake]

from what i understood CPython compiles it in bytecode

While Cython actually translates it in C and compiles it right?

[covert lake]

oh damn

Did you ever try it out?

it’s interesting

[snow sedge]

Personally didn't, but not sure if it will be an easy or useful rewrite in combination with torch. Even though rvc model code is python, it's compiled and optimized by torch already

And wrappers like realtime/offline inference mostly use numpy arrays/torch tensors which are, again, C code already

[polar flax]

the better way to do is the guide like this https://rentry.co/RVC-dataset-RX11

[little stone]

Which AI Voice is the best?

because I want to try it for the first time

[willow karma]

hi

[pine mural]

.wk xaviersobased

[indigo pike]

xaviersobased in AI HUB

👑  ⁦adiiose⁩ - 5284 plays
** 2.**  ⁦#igrok⁩ - 4388 plays
 3.  ⁦remy⁩ - 3999 plays
 4.  ⁦exan⁩ - 3174 plays
 5.  ⁦luminal⁩ - 2823 plays
 6.  ⁦yeresin⁩ - 2555 plays
 7.  ⁦sqwab⁩ - 1990 plays
 8.  ⁦vice⁩ - 1808 plays
 9.  ⁦wg⁩ - 1346 plays
 10. ⁦osc⁩ - 1191 plays
 11.  ⁦boshia⁩ - 1105 plays
 12.  ⁦Teckmek⁩ - 1032 plays
 13.  ⁦Markus⁩ - 1030 plays
 14.  ⁦buga⁩ - 1026 plays

cloud rap - dark plugg - sigilkore Artist - 273 listeners - 54230 plays - 198 avg

[steel heath]

f

[covert lake]

you mean program to use for speech to speech models?

RVC

what’s ur pc gpu and what are u looking for?

[ashen blade]

Hello

[zealous zephyr]

hi

[ashen blade]

Yo

[uneven lake]

Hi everyone

Someone know to do this ?

[covert lake]

Do what?

[long onyx]

hi

[icy pendant]

Nick do you know how to do this!!!!!

[covert lake]

😭

i dont know why some directly ask how to do this

its not the first time i see smt like that

[icy pendant]

And then they disappear into the void forever

[covert lake]

fr

[tired jasper]

Training all Overwatch2 tanks as single RVC model

[unborn jacinth]

Hello everyone, there is some kind of lightweight version for AMD video cards, in the version MMVCServerSIO_win_onnxdirectML-cuda_v.1.5.3.15 my voice is either normal or broken

[timber egret]

i wanna know what realtime voice changer works best for me
ihave
15 440
16ram
rx 470 4gb
windows11

iwanna it simple like ui and this stuff and good

[shut nimbus]

If you live in the US, check your chicken, it
might need to be thrown out. There's a recall
on a lot of frozen chicken brands relating to
listeria

Frozen waffles and pancakes, and fish in
some places were also recalled.
(I'm sorry for the off topic just wanted to warn everyone)

[unborn jacinth]

maybe someone knows

[covert lake]

its gonna sound like shit

[covert lake]

i think you are talking about wokada, well there is a fork (modified version) which has better performance if that’s what ur looking for

-rt

[rare sorrel]

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

[covert lake]

1st link

[tired jasper]

Why?

Is RVC pretrained (170 voices as single model) like shit?

[unborn jacinth]

can you give me a stable lightweight version? I will be very grateful

[covert lake]

ur gonna mix multiple voices into a single model

[covert lake]

but that’s used a pretrain not model

[covert lake]

read the 1st guide

[covert lake]

-rt

[rare sorrel]

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

[covert lake]

1st link, wokada fork

u could try, but im not sure with such low specs

[unborn jacinth]

thank you

[covert lake]

yw

[tired jasper]

Pretrained model is the same we use, bro

[covert lake]

you’re talking about using multiple voices in one tho, u can’t like use each voice

[tired jasper]

I'm not mixing, they are in each id

[tired jasper]

Yes i can

[covert lake]

iirc speaker id didn’t work, unless im wrong

[tired jasper]

That's 12 speakers in one model

I already tested, bro, they are all correct in the model

Just need to train more now

[covert lake]

huh fr?

@chilly lake speaker id works all fine?

[solar torrent]

You used one of many voices from a mixed voice model? Where did you get this information from?

[covert lake]

last time i heard of someone tryna use it i remember it should have been set to just 0 as it didn’t work

[tired jasper]

Not mixed, just changing speaker_id.

[tired jasper]

In webui? No. They are lazy to implement, that's why i don't like GUI

[solar torrent]

So is it like many voices packed in one model like a ZIP? Because I've never heard about it being used generally.

[pastel stirrup]

prequel: I wanted to run comfyui with FLUX on colab servers and so I installed all required files like unet, clip etc. And when I clicked to generate the image, it took about 2 minutes and then the UI just shutted down. Other models work normal, so I guess Colab's 12 gb of RAM aren't enough

And I decided to run this stuff on kaggle because it has ~32 gb of RAM. And I rewrited the same code for kaggle but when I ran the last cell, it returned error

[covert lake]

Oh i was talking about UI, there is an option but doesnt work

are you using the rvc cli fork by blaise?

[snow sedge]

Rvc kinda supports multiple speakers, but I'm not sure anyone uses it. Wokada allows choosing speaker id and it should actually work

[tired jasper]

Yes, like a zip behavior, but the model size is the same as single voice

[tired jasper]

Yes

[covert lake]

ohh

i wonder if its fixed in mainline or applio ui by now

[pastel stirrup]

any ideas?

[tired jasper]

Last time i tried didn't work, but works in RVC webui, idk why yet

[chilly lake]

for Applio the model needs record of speaker IDs, then you can use them

[tired jasper]

Even blaise is broken, i had to fix many parts

[snow sedge]

Fork or original? I don't remember if I ported speaker choice properly from mainline but I never tested

[solar torrent]

Something doesn't exist in Torchvision? Have you tried reinstalling Torchvision?

[tired jasper]

Fork, changing speaker id don't change the voice

[snow sedge]

Then it might be that I either removed part of rvc code responsible for speaker choice or forgot to pass it to model

[tired jasper]

I can test again later, could be false positive since i didn't tried all voices

But VC kinda stuck with my model

[chilly lake]

changing speaker does change the voice, I tested on OG pretrain

[covert lake]

So I'd just need to make a dataset with a folder for each speaker, train it and id be able to use that feature?

[solar torrent]

Maybe make sure you compare between fork and original to make sure which code was added and missing.

[snow sedge]

If you have a working model, I can try fix this

[chilly lake]

yes

i really need to take a look and chage the way the speakers are enabled on UI

[tired jasper]

What version you tested?

[chilly lake]

any, if you pass speaker id it works, default is 0

[solar torrent]

I'm wondering. Who will train anime twin voices into one single model?

[feral halo]

Ayo? @solar torrent level 13 !!!

[tired jasper]

Yes, you need custom feature extractor and custom dataset preprocess as well

[chilly lake]

no

you just need to preprocess

with speakers placed in different folders

[tired jasper]

Where to send it to you?

[chilly lake]

OG pretrain with all the speakers

take a look

[tired jasper]

Idk if i trust 😂

[chilly lake]

extract your own then 🙂

[spice wasp]

can u share? would like to try

[tired jasper]

Yeah

[spice wasp]

awsm

[pastel stirrup]

da pomogite naxyu

[tired jasper]

It's not finished, also don't know what to do with it yet, i'm considering to sell or something

[chilly lake]

emb_g.weight torch.Size([109, 256]) torch.float16

this is a tensor with speakers - 109

[spice wasp]

id rather pay for a tutorial on how youve don preprocessing lol

[covert lake]

How would others know which Id is which voice vtw

[tired jasper]

If you know Python you can just take a look at rvc-cli from blaise

[chilly lake]

install a compatible set - torch, torchvision, torchaudio

[spice wasp]

ill have a look thx man

[covert lake]

I mean it should be fixed in Applio UI too

And ye Ik python, it's just I do things in cloud

[chilly lake]

you dont even need rvc-cli, just use env\python core.py infer

[tired jasper]

This is the failure, people need to attach a text file to model path. In beatrice they solved it using toml file

[snow sedge]

You can send a link in dm to google drive for example

[tired jasper]

I only heard about Applio once, idk what exactly it is

[spice wasp]

would this also mean that it requires less computing training several models at once?

[snow sedge]

This can potentially be solved by including id-name mapping in model metadata, but no one does that currently

[tired jasper]

I will send multipart zip ok?

[snow sedge]

Sure

[covert lake]

RVC fork

Made by Blaise and the Applio team

Like Vidal too

[covert lake]

Including a text file would make so it doesn't work when someone tries to download it directly on cloud tho

Theyd have to extract it on their PC and use the manual uploader

[covert lake]

Doesn't it work in the web UI too?

[chilly lake]

it only works from UI if it was trained and recorded

since all models trained with og pretrain have the same tensor, they kinda all have all the original VCTK spakers

maybe need to delete this weight when training a new model

[tired jasper]

would work if the cloud read it, is just a reference index. Using Beatrice you can zip all the models and import zip in VC, it will load voice names.

you also can put index in model dict

but you would need to tell all people you are doing it

idk why they didn't implemented it on beginning

[covert lake]

Any unnecessary files would make the downloader not download

[tired jasper]

it's necessary when you have more than 5 voices in one model, but they didn't though about it in beginning, now someone need to create a convention

[covert lake]

I understand, I'm just saying doing like that would just make more people confused as it wouldn't work

[vital snow]

If i want to train specific language model do i have to tune pre train model to get it sound better?sry i just start for like 1 week

[mortal sail]

how do i delete models

[covert lake]

It would have been better also if instead of the Id it could use a name

[tired jasper]

yeah, best way is to put it on model dict

[covert lake]

Which language are you looking for

[mortal sail]

@icy pendant

[tired jasper]

idk even know why they did faiss index separately

[covert lake]

Is there a way I could do that?

Btw, how's Beatrice going?

[tired jasper]

you need to code it

[tired jasper]

not so good, but not bad at all lol

the metallic noise is annoying

but not all voices have it

[icy pendant]

Overwrite the slot with a new model

Or go to model_dir delete the slot number

[tired jasper]

some you can workaround with EQ and such way of speech

[covert lake]

So it's like a "cheaper" RVC as of rn?

[mortal sail]

ok thanks. btw whats okada?

[vital snow]

Thai

[tired jasper]

no, it's good, but general quality is worse. I guess their aim is to make it faster for realtime, in this way beatrice is much superior than RVC

[covert lake]

So a faster but less quality RVC yea

[icy pendant]

Voice Changer

[solar torrent]

The Okada is a newer voice changer that works just like RVC.

[tired jasper]

if they fix the metalic noise will be not far from RVC to be honest.

[mortal sail]

whats the difference

[feral halo]

Ayo? @mortal sail level 1 !!!

[covert lake]

Ah, you could try with the OG pretrain

You could also make a pretrain if you want but it's hard to do so https://discord.com/channels/1159260121998827560/1233407331405004954

[covert lake]

Does Beatrice get regularly updated?

[tired jasper]

w-okada is the maintainer of the repo voice-changer, which is a tool to run RVC and Beatrice models

[solar torrent]

The difference being the resource usage and supports more models I think.

[tired jasper]

idk, most people here don't even know that is there new version: v2.beta2

i guess it updates slow

but it's not dead

[solar torrent]

Okada is recently maintained, while RVC is kinda outdated since its last update was last year.

[tired jasper]

not that outdated, RVC v2 is pretty good, no need to update too soon

[solar torrent]

[tired jasper]

Go down, go down, go down, yeah

[feral halo]

Ayo? @tired jasper level 12 !!!

[tacit marsh]

Yo everyone

How is everyone in this server

[solar torrent]

Getting graph like a slider.

[lucid musk]

Where do I get the realtime voice changer

[solar torrent]

-rt

[rare sorrel]

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

[lucid musk]

Thx

[tacit marsh]

Afternoon

[tired jasper]

don't use drugs, it's morning, joke

[vital snow]

Thanks also, i have dataset that speaks thai , korean , jp and a bit english
While i speak thai and jp it sounds good but with english it struggle to pronouce
Is this because of too low english dataset?

[tired jasper]

Roadhog from Overwatch2 in the multi speaker model

[chilly lake]

the model learns 1) voice 2) prononcuation of specific sounds

if your dataset lacks in specific sounds, then it wont be able to reproduce them correctly

for that purpose you can use a pretrain and train your voice on top of it

a voice model can fix some accents, but it cant fix bad ones

[vital snow]

Thank you.

[feral halo]

Ayo? @vital snow level 1 !!!

[chilly lake]

also if you use an index, it brings prononcuation from the trained dataset, so that may result in an accent. if you dont use the index it keeps the original

that's the 'retrieval' part of rvc

[vital snow]

Alright i will try

Thank you for your help

[tired jasper]

model seems to starting to stabilize loss =(, some of voices are robotic yet, maybe i should not include too different pitch voices in the dataset for single model, or increase dataset, idk

how much data you use?

[solar torrent]

Imagine the AI Hub becomes the Rizz Hub.

[ionic pumice]

https://cdn.discordapp.com/emojis/657928610870460416.webp?size=48&quality=lossless&name=CliffHappy

[solar torrent]

[hazy ferry]

Where do i make ai art

[mild leaf]

Hello guys! I'm new here. Can you help me which program I have to use for voice change?

[hazy ferry]

idk

[solar torrent]

At #🏙│ai-images, there's a latest pinned message that includes some of most known AI generation websites.

[hazy ferry]

What blud typing

[hazy ferry]

Thanks

[hazy ferry]

Can i create?

Ans

And what do i say for the bot or what it is to make the art

[solar torrent]

Each AI has its own way to generate. For Stable Diffusion, you type some words in prompt box, click generate, and it will generate an image you typed.

[hazy ferry]

Give me an example

[feral halo]

Ayo? @hazy ferry level 1 !!!

[ionic pumice]

you can go to #🤖│bots and ping the for bot to ask it to generate an image

[hazy ferry]

And what do i say

Just

[ionic pumice]

that's probably the easiest thing to do if you don't wanna install anything

[solar torrent]

A chicken at a farm

[hazy ferry]

I can just say : generate me an image of a dog?

[ionic pumice]

yis

[hazy ferry]

o k a y

[half kayak]

Does GPT-SoVITS support other languages like portuguese? If not which tts solutions support it?

[hazy ferry]

What bot

[solar torrent]

Cake and Dramel 🔥🐪

[half kayak]

In the guide https://docs.ai-hub.wtf/tts/gpt-sovits/
It says:

GPT-SoVITS is an open-source repository focused on TTS & cross-language inference, with a Colab port coming soon. Credits to RVC-Boss.

but then says

Currently it only supports Chinese, English & Japanese. More languages are coming soon.

So does it support training in other languages or not?

[grand mountain]

https://tenor.com/view/furry-wiggle-mrrp-nya-colon-three-gif-27633901

[silent flame]

guys how tf do I generate images

[feral halo]

Ayo? @silent flame level 1 !!!

[covert lake]

@elder willow

[covert lake]

To generate images for free, either:

• Use @elder willow in #🤖│bots (It's powered by DALLE3, from ChatGPT+)
• Use Open Source Models like stable diffusion & flux that could be a bit harder but better, what's ur pc gpu?

[silent flame]

ah alr bet

[vague rapids]

anyone know how to install

pretrained voices?

im kinda confused with the process

[elder willow]

hello guys

can i learn ai with c# language

[hazy ferry]

I

[feral halo]

Ayo? @hazy ferry level 2 !!!

[hazy ferry]

Am

Bored

[covert lake]

most ai things are in python tbh

[ashen blade]

Helloo

[feral halo]

Ayo? @ashen blade level 1 !!!

[jagged ermine]

Hello beautiful people

[ashen blade]

Helo

[vital snow]

ello

[hasty elk]

[half gust]

Bro I, did it. I created the world's first entirely If - Else based neuron.😅

[vital snow]

"Unlike with normal models having the dataset for a pretrain too high quality might cause it to be more sensitive to noise when using it to create a new model so I would recommend making it a finetune of the base RVC pretrains.
To make the pretrained model from a scratch, audio with low-mid quality should be considered."
how it will be different from normal high quality dataset?
i only need to remove instrument and keep the reverb , noise , echo ?sry im dumb

[chilly lake]

you will not get anything good by training from scratch

not unless you use a bunch of high end cards and 100+ hours of audio

[vital snow]

i just want to tune from original one

[supple flare]

guys how can i use the model ?

[vital snow]

yeah... my rtx a5000 isnt on the list maybe i just give up and work more on my dataset and train other model

do you have anytrick to cut out noise without suppressor?all my model face this problem

[chilly lake]

having a slight noise in the audios is okay

having distortions / bad mic / someone revving a motorcycle outside is not

with a little noise the model is more adaptable to handling the noise in the inferred audios

[narrow mountain]

use UVR5

[vital snow]

Ahh my house is near highway so there are a lot of vehicles noise

Thanks for advice! I fixed my model's english by get more data
Luckily he do speak many languages

[chilly lake]

UVR can remove some noise, sure, and long as it does not damage the actual voice

[exotic spade]

k

[hazy ferry]

Wazaaaaaaa

[worthy coyote]

is this server under china's support?

[night lake]

💀

No, this server under the support of the shadow government

[worthy coyote]

omg

wen need to burn this whole server down

[night lake]

then another server will replace its place

[worthy coyote]

no way

[night lake]

the shadow gov aint messing around with these e girl voice models

[worthy coyote]

hmmmmm

[amber light]

shhh they must not find out

[snow sedge]

Multispeaker actually works, it's just that the difference between some speakers is not big. I didn't notice a difference between speaker 0 and 1, but 0 and 12 are completely different. Though there're few problems with the UI

Like you cannot add speaker IDs (I guess they were supposed to be populated based on model metadata), and metadata has "yet another field" for number of speakers and the code is not aware of it

[chilly lake]

you can run inference from CLI and just provide a new SID

every trained model has 110 original speakers, well... 109

0 is the one that gets replaced by regular training

[stoic forum]

Hello!

[snow sedge]

The problem is GUI and standardization. I understand that you can just specify an arbitrary SID and expect it to work and that's how I ended up testing it, but that's not how it is supposed to work. And I see that Applio, for example, provides the number of trained speakers (not pretrained) in speaker_id
https://github.com/IAHispano/Applio/blob/1f743e9ab0cc80c8238afafed259e7665fca713f/rvc/train/extract/preparing_files.py#L70-L74

[chilly lake]

yeah, we've started collecting and storing them recently

May need to do some extra work though, that was just an experiment to check whether it actually works

[glacial igloo]

Hey there! Is it possible to convert multiple files at the same time on 1 GPU?

[snow sedge]

DDPN RVC apparently provided speaker_info with SID as key and string as a value. At least wokada voice changer was taking it into account
https://github.com/ddPn08/rvc-webui/blob/b71742809a24cd89eb18081b831c0b1ac11ccb2a/lib/rvc/checkpoints.py#L113-L114

[chilly lake]

seems like they had a json with speaker names?

[snow sedge]

But oh well, yet another dead fork with its own pecularities

[chilly lake]

so what though about all models having all speakers is not exactly right

there are different results if you run inference, but they are not original

i need to try a different sample

[snow sedge]

Having all speakers is not a problem. The problem is that there should be a standard way to at least understand how many trained voices are in the model and which ones you may choose

[chilly lake]

Sure, but there's no other RVC clone that does multi-speaker training that I know of...

[snow sedge]

And that's unfortunate, because everyone was claiming that it just doesn't work. But anyway, adding an object with speakers would be nicer since you'd be able to also label each speaker

Just knowing number speakers based on speaker_id is okay-ish, but you'd either have to figure out and label yourself or model maker must provide a description

[chilly lake]

would probably help to just delete speaker weights so when you train a new model only the trained ones are available

so it would make it easier to just list the speaker dimension as max

and then the model maker can provide a metadata file with speakers

[velvet grail]

Can someone link me the google workspace RVC tool? That or whatever is the commonly used one these days.

[snow sedge]

Already in the model dict, no need for a separate file

[chilly lake]

I mean names, etc

[snow sedge]

Can be stored as object or whatever is easier to serialize/deserialize. Even csv would work in this case. Model config is already stored as stringified array

[chilly lake]

i can be just a normal dictionary, no issues storing them

[snow sedge]

safetensors supports only string as key and value, but it's not like a big issue

[gloomy locust]

someone can help me?

[amber light]

with?

[gloomy locust]

Im trying use the Ai Voice Changer and isnt working, even if i try use others channels and not working

[chilly lake]

for pth this works just fine import torch m = torch.load(r"x:\test\model.pth", map_location="cpu") s = {"0":"Arnold", "1": "Bill", "2": "Mike"} m["speakers"]=s torch.save(m, r"x:\test\model2.pth")

[amber light]

#🔍│help-w-okada

[vestal temple]

Hi

[feral halo]

Ayo? @vestal temple level 1 !!!

[fathom basalt]

Does anyone know how to use the new voice changer?
a video tutorial?

[neat brook]

Can someone tell me how my voice changer sounds

[pale patrol]

wdym new voice changer

[fathom basalt]

The voice changer does not recognize me

When I put the file it tells me, "You need to specify a configuration file" appears.

[glass dust]

hi there

does anyone know whats the best lipsync program , or website?

[south kelp]

miss the old peak days

[buoyant lodge]

how do make one of thoes ai text to speech vids

[golden pond]

hey guys, does someone with nitro who got a xbox gamepass gift trial and not use it could send it for me in dm ? really thanks if you do it 🖤

[tired jasper]

0 is female, 1 is male 🙂 so the difference is on transpose

[slim badger]

anyone know where i can find voice samples?

so i can clone them in elevenlabs

[fleet hawk]

hello nerds

how much better is ov2super over original v2

[feral halo]

Vote Booster: Vote now for a 10% boost. https://arcane.bot/vote
✨ Tip: Use /card to modify your rank card

AI HUB's leaderboard

Want to view more than the top 10 users? Click here

[spare lark]

Does using old version of snap bans tour account?

[amber rapids]

yo

new here

[fleet hawk]

yes

[feral halo]

Ayo? @fleet hawk level 1 !!!

[spare lark]

Not new I am old but i never really show up that much

[amber rapids]

ok brah

[rustic urchin]

hi

[spare lark]

Really?

[fleet hawk]

I am new here

really really

[spare lark]

🤦‍♂️

Can I know if using older version of snap leads to ban account

[feral halo]

Ayo? @spare lark level 1 !!!

[fleet hawk]

heh

@elder willow 'sup

wow

doesnt even work anymore

[ornate ore]

hello

[earnest dragon]

Why would you use a older version

[ornate ore]

can you guyes sub to my channle

it is called Lord Tiler

please

[feral halo]

Ayo? @ornate ore level 1 !!!

[ionic pumice]

it's usable in #🤖│bots lolz

[worn mesa]

[worn mesa]

wut are u offering?

[ornate ore]

you sub to channle

[worn mesa]

[ornate ore]

please

pretyplease

[worn mesa]

done

[feral halo]

Ayo? @worn mesa level 1 !!!

[worn mesa]

subscribed

[lament mulch]

hey anyone who is good at voice cloning and wants to get paid dm me

[hasty junco]

@final cairn слушай, а ты сможешь ещё голосок репринцева намутить?

[ornate ore]

dose any one whant to subscribe

[feral halo]

Ayo? @ornate ore level 2 !!!

[ornate ore]

please

[polar flax]

no, unless you share it in #1159290752195633273

[polar flax]

any reason to include multiple voices in a model? probably?

• switching to another voice on the fly (for like RP-ing purpose), I don't 100% agree to this tho
• the same speaker in different expressions, e.g. normal speaking, screaming, sobbing, whispering. may be useful if the inference could automatically select one of them that matches each part of input voice part.

[tired jasper]

idk to be honest, it's an experiment, i want to compare with solo, and i'm building a script right now to test many models

i want to know if it worth

also it can save much money when training

i have 12 voices in one model, took around 160 epochs to all 12 voices, when it took around 90 epochs for each solo

i want to know where to find the most quality speech audios possible now

so it will be fair test

[polar flax]

I'm not sure how it works for that situation, maybe like? training a LoRA with multiple trigger words for each character outfit, rather than training different LoRAs for each character outfit

[chilly lake]

a specific voice does not take all that long to train

after ~5-6 epoch sounds it already sounds okay

[hardy warren]

.

[celest arch]

Goated server

[vital snow]

fr it gives support more than my dad

[feral halo]

Ayo? @vital snow level 2 !!!

[ionic pumice]

CRAZY ...

[pale patrol]

waht

[finite fog]

erm hi

[chilly lake]

it does not take that long

[peak field]

hi

does ai still work in this server

or nah

[paper wigeon]

Hi

[paper wigeon]

Oh hey you're here too

[vast vortex]

?

[amber light]

what lol

whole server dedicated to ai

[hasty elk]

lol

[tacit falcon]

💔

[native root]

Hey devs! 👋

I'm looking for a developer partner to join me in building Agelix, an AI-driven platform that helps creators and businesses with everything from social media content to sustainable practices. If you’re into hands-on projects, brainstorming new features, and making tools people actually love using, let’s chat!

Shoot me a message if you’re interested or just want to know more.

[ionic pumice]

https://cdn.discordapp.com/emojis/925421909858975784.webp?size=48&quality=lossless&name=LeoReadytodock

[winged sequoia]

#generate the picture of one dog is going to park with here chileds

[amber light]

what

[tired jasper]

for very clean voice input i can't notice any difference between train solo speaker and multi speaker

[buoyant palm]

Hello.

[vital snow]

what to do if voice is squeak then if i lower pitch it will be distorted

[loud kiln]

Do time stretching instead

meeheeehehe

anyone wanna vc

I wanna test something

[hollow summit]

Are There rvcv2 collabs with pitch min and max (rmvpe +) that work

[feral halo]

Ayo? @hollow summit level 1 !!!

[stable ginkgo]

Anyone got a anais tawog AI voice?

[covert lake]

You can search rvc ai voice models at:

• #1175430844685484042
• In #🔍│find-models :
  • Send " @atomic vector search (name of the model)", without the ()
  • 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://applio.org/models
• https://voice-models.com/
• https://thevoicemodels.com/ (for Turkish Models, login required with discord and level 2 on their server)

if there isnt one, you can:

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

[covert lake]

that rmvpe+ is only for inference (use models) btw,
There's:

• Ilaria RVC Zero (ZeroGPU, A100, HuggingFace Space, faster than colab)
• Applio Colab

[hollow summit]

I see but you can’t adjust the frequency off the pitch like the older versions

[rare sorrel]

📚 Audio Guides & Tools

• Creating Datasets for RVC using iZotope RX11, by Cauthess
• Gathering and Isolating Audio, by SCRFilms ❄
• Instrumental and vocal & stems separation & mastering guide, by deton24
• Vocal Mixing Tutorial, by Roomie
• https://mvsep.com/

More commands: -colab, -uvr, -karafan, -overtrain, /help

[tired jasper]

https://github.com/IAHispano/Applio/issues/836

[vital snow]

it maybe dumb question but if i use only singing dataset to train speech model will it work? I always mix both speech and singing so idk

[willow grove]

why is my ms at 41000 (im new to this)

[vital snow]

what r u using?

[willow grove]

"voice changer client demo" im tryinna download villager voice

[vital snow]

what gpu are you using?

[willow grove]

nvidia mx250

[tranquil lantern]

https://tenor.com/view/ai-hub-discord-no-more-fortnite-trash-garbage-gif-10638344671466795947

[vital snow]

i dont know if that gpu can do much try lower your chunk

[covert lake]

damn gpu so old the official specs page on Nvidia site is deleted 😭

I searched it online and it seems to have just 2gb of vram

the minimum would be atleast a gtx 1650

[vital snow]

rip

i have only intregrated gpu but i use rtx a5000 server and connect to it

[covert lake]

You could try the Fork which has way better performance and even a CPU model

not sure if that would work decently tho

In case it doesn't, there's Cloud (remote good pc for those who don't have a good one, YOU CANT DO THIS ON MOBILE):

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

[covert lake]

i have only integrated graphics too

I do most things on Cloud

still a decent pc for some indie gaming tho

[vital snow]

i see...

what you got?

[covert lake]

i3 10th gen with 8gb of ram

[vital snow]

ohhh same here

[covert lake]

I mostly play indies

oh cool

[vital snow]

it still good but i struggle with my ram 8gb isnt enough for me now

[covert lake]

real

[mint wave]

Does anyone have a female voice model for some with a low pitched voice? I wanted to mess around with my friends.

[worthy coyote]

low pitched female voice

then it's just male voice

[hollow quarry]

yo

[worthy coyote]

https://tenor.com/view/hey-there-play-with-me-gif-25986506

yoyo

[plain cove]

poyo*

[elder willow]

hi

[vital snow]

Hi

I use wokoda fork with rtx a5000 i realised it use very low gpu resources but when i try to upper settings it shows red pref but still low gpu usage what am i do wrong? Or i should change my gpu

[feral halo]

Ayo? @vital snow level 3 !!!

[vital snow]

My spec
AMD EPYC 7543

16 GB RAM

RTX A5000 24 GB

[frigid lark]

What's this server about...

[icy pendant]

I dont know about the compatibility with an A5000, I dont know much about that gpu. But what number does the perf display, and what f0, chunk and extra you using

[covert lake]

AI, mostly RVC & Wokada but u can also find other things such as FaceFusion & TTS

[frigid lark]

Im here because i gave a bot permission about " join servers for you" ...
And now it threw me here

Leaving thanks a lot ! :))

[vital snow]

120ms , 2.7s extra, rvmpe
My pref show 70

It uses like 4%

Of my gpu