🧬│ai-chat

[trim wolf]

hi

[zenith forum]

can somone help my voice is like 20 seconds delayed and only rearly pick up my audio and changes it to a other voice

[covert lake]

i already asked you to elaborate in #✨│ai-help

[covert lake]

I'm not sure we have a brasil version, maybe @signal cairn would know

[cosmic aurora]

Hello, I'm French and I need help with my voiceai.

<@&1159293204038955078> ?

[covert lake]

elaborate in #1192011222023950368 : your pc gpu, operative system, what you want to do, what program are you using/tutorial link, etc

[cosmic aurora]

No, I just removed something that was bothering me and I can't see the voices.

[covert lake]

Yeah, you need to elaborate in an help channel

this is an ai-chat, not an help chat, so we can't help here

@cosmic aurora also, be sure you aren't using the program "voice.ai", it's paywalled, we use a better program for realtime voice changing here

Please elaborate in #1192011222023950368

[cosmic aurora]

it says I can't a required tag I don't know what you can't come in voc

[covert lake]

You need to put a required help tag, like voice changer, also i can't vc

please either make a forum or use #✨│ai-help

[cosmic aurora]

how do i put a help tag

[covert lake]

select them

[cosmic aurora]

sebon I posted

[marble cargo]

Hey guys who can help me generate veo3 video

[elfin rain]

who's using agents on the regular now?

[fervent bone]

Trying to make a Ai cover of a song but it's having trouble with the main vocals. How can I fix this?

[polar flax]

still using that trash?

[polar flax]

use better separation models, i.e. melroformer models

at last you may need to clean and edit it before trying to infer

[cosmic aurora]

Yes, and what about you?

[fervent bone]

I don't know how to do that

[polar flax]

you must have used an easy cover tool

so you cant expect better results

try consult this guide https://docs.aihub.gg/

should you need more help, make a new thread in #1192011222023950368

[polar flax]

me using that trash?

https://tenor.com/view/hahaha-no-robot-zap-gif-15966147

[fervent bone]

Yeye that's what I use

[dusk schooner]

Can someone show me how to make ai videos with my own character i have an example video.

[covert lake]

elaborate in #1192011222023950368 for help

[hybrid temple]

the one who make these ads needs to get sued

[covert lake]

I hope you didn't use video youtube tutorials for wokada, since they are all old, for a question you can either use #1192011222023950368 or #✨│ai-help

@hybrid temple seems like rentry needs money rn

lemme guess, youtube guy named novision?

that one is outdated asf

it uses the same things used in 2 years old tutorials lol

there's no updated video tutorial

if u used one, you are using an outdated wokada

please elaborate in #1192011222023950368 , idk what you're talking about

this isn't an help channel

we won't help here

@abstract abyss be sure to not have used youtube tuts, they are old and could be risky

please, tell your pc gpu, operative system, what you want to do, a screenshot of the program in #1192011222023950368

[vast wave]

hii, theres some ai that can detect song genres (i need it for suno better results)

[uncut comet]

Hello, is there any AI that can send song lyrics without making mistakes?

[timber finch]

Whats the best settings or best model for realistic laughing?

[polar flax]

mostly nope

[timber finch]

wdym

[polar flax]

idfk

ask chatgpt

[copper vine]

Hello

[ruby shuttle]

Hello, If a multimodal and multi AI software platform like Monica.ai or Galaxy.ai uses a specific API, are they limited to what that API can do? E.g. Monica.ai uses chatgpt API, therefore can it only do what chatgpt can do?

[smoky bronze]

D

[open kite]

Hello

[open kite]

Anyone here

[short maple]

Yep
Infact it probably can do much less compared to the actual ChatGPT, depending on the initial prompts

[ruby shuttle]

why can they do less compared to GPT? Is it because they have to manage a bunch of other APIs if they support multimodal AI softwares and APIs? thanks for your help

[short maple]

I'm not completely sure about it either, but I think that's infact the reason

[nova sonnet]

Hi any one using kling ai?

[bold rampart]

$20 sub from what? Is it the Gemini Pro subscription?

If you meant the Gemini Pro subscription, then for Flow, you get 1,000 AI credits which you can generate Veo 3 video using the Fast version for 50 videos, if using the normal Veo 3 version, you can only do 10 videos the fast version costs 20 credits and the normal version costs 100 credits

If you guys have ChatGPT Plus, you can use this custom GPT I made tailored to render first person views or perspective of character in images you upload: https://chatgpt.com/g/g-68743cd0290881919dbc37a02146c1aa-first-person-character-view-transformer

[marble cargo]

Oh i paid for thag one :'((( Gemini pro (flow)

[bold rampart]

Bruh, in my country, they give the Google AI Pro subscription for 1-month for free

Bruh, 15 months is sure is a really long time

Yeah, OpenAI only gives you a ChatGPT Plus subscription for only 1-month, $20 too… not fair

And also $200 for the Pro subscription? And it’s just 1-month too!

If they gave 15-months subscription for the Pro subscription, then I understand

No, I meant the ChatGPT Pro subscription, if they put $200 for a 15-month recurring subscription, then people would understand

Why can’t they just put like $70? $200 is overkill

Yeah right?

They just doubled the price

Yeah

Then ask ChatGPT

[rotund jay]

is there any way to silence background noise while using the voice changer

[wind plinth]

tmd

[rotund jay]

on male voice models

[wind plinth]

idk bro u got max 2more years after that your life will be a downhill unless u suck sum cash off sum1

[rotund jay]

gng wha..

[wind plinth]

mark my word

for next gen too

kawai onichan bih what doing in ai real estate discord

[still sigil]

?

be very careful with the next messages you send

[wind plinth]

😂😂😂😂😂😂

davinci grade

picture

goofball

[covert lake]

This is a General AI Server, AI has many fields

Elaborate in #1192011222023950368 :

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[covert lake]

Are you trying to do e-girl trolling/catfishing?

[covert lake]

This is a General AI Server, AI has many fields

Elaborate in #1192011222023950368 :

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[solar torrent]

Drink a yoghurt, I guess? That would help you to poop easily.

If you meant by a voice changer program named "W-Okada", it would be better to go to #✨│ai-help or #1192011222023950368 and explain about your issues there.

[covert lake]

Are you planning to do e-girl trolling/catfishing?

Hello, how are you? Do you need any help?

[solar torrent]

Hi.

[covert lake]

Ohh nice, how are you?

Good too, what u up to?

[runic garnet]

hello

i want to know how to install this voice changer called w okada

where can i ask the owners and get help_

?

[solar torrent]

Go to #✨│ai-help or #1192011222023950368 and I'll tell you how to install W-Okada.

[runic garnet]

thank you

opened it

the post

[solar torrent]

That's too unfortunate. RVC-GUI is an old RVC fork program from some years ago.

[wicked summit]

hi guys im new in this servers

[solar torrent]

https://media.tenor.com/2nzRLDqIWkkAAAAM/new-member-roleless.gif

[wind plinth]

namari whatd ur ig

[restive dragon]

Hey ok i was looking for this

[restive dragon]

Me too are you using AI in some work or anywhere?

[solar torrent]

Too confusing to read. You mean like W-Okada didn't work for you? I think you're using the original version of W-Okada, which is outdated. A kind of RVC GUI program with "realtime" mode is also old too. For W-Okada, go to #✨│ai-help or #1192011222023950368 and explain about your issues there.

[restive dragon]

What are you looking to do? I mean, what’s in the link?

[solar torrent]

Self-solving is crazy.

[dusk schooner]

Can someone show me how to make videos with my own character?

[solar torrent]

Catfishing for certain offensive situations is banned here, you know. Don't play with me or any mod here.

[restive dragon]

Heygen

[dusk schooner]

Hi check messages

[runic sentinel]

Hey buddies I'm new in this server

[dapper shuttle]

Hello everyone

Has anyone here subscribe all in one AI such as Magai / Abacus?

Is it legit and worth it?

[wicked summit]

i am learning about it

[runic sentinel]

Hey how are u looking forward to explore and build careers in AI fields, nothing much but some ideas or suggestions would help

[restive dragon]

Have u tried Ai agents they’re in soo demand business like paying thousands of dollars if someone build it for them

[restive dragon]

Nice what have u tried so far

[wicked summit]

nothing much just started

[restive dragon]

If you’re just looking and starting with something the best thing to learn is Ai Agent I also started learning it early 2025 and now i have clients who have paid me $2k to $8k

[wicked summit]

oo so u can make money from it?

how??

btw im 18

[runic sentinel]

No i haven't tried it yet....can u suggest some ways that I can get to know this field

[restive dragon]

Of course we build aAi Agent that fix there business problem and they pay you for that

[restive dragon]

Sure, we have a private community. We are showing everything to help you get started.

[wicked summit]

so how can i start and make money

[restive dragon]

Let’s say a business owner is getting so many calls, but they cannot pick all of them so you build them a voice Agent that literally pick up the phone and answer their customer

[restive dragon]

It’s nothing much complicated. You just build a tool and you sell that tool to them.

[runic sentinel]

So it's like just solving problems and company give their problems to you and u solve it

[restive dragon]

Yes just like social media marketing they fix a business problem, which is not getting much customer

[wicked summit]

how can i build and how can i sell that

[runic sentinel]

like white hat hackers do for their companies by finding bugs and fixing them for better stability

[restive dragon]

We use N8N to build and sell u just connect there business tool with it

[restive dragon]

No it’s not that let’s say you are building a digital employee for them who can do like work of three employees at once 24/7

That’s why business owner is paying so much money for this because it saves them so much time and they don’t have to hire more people

[runic sentinel]

Ai will take your jobs got real

[restive dragon]

Haha it alrday did cuz I’m replacing it i know how this will look in future, also Microsoft have fired 9000 or more people and they’re using AI

If you wanna start something and don’t wanna get replaced by AI, just start replacing others with AI

[hearty sage]

Hey guys

[restive dragon]

Hey how ua doing

[runic sentinel]

nice line

[wicked summit]

can we talk in dm?

[hearty sage]

Good

[restive dragon]

Sure

I’m open to help u

[runic sentinel]

hey

[hearty sage]

I will get okada from here

Because YouTube too old videos

[restive dragon]

That’s true for real I think you guys are not familiar how powerful ai is

[restive dragon]

Sent ya dm

[covert lake]

every video tut is outdated

make a post in #1192011222023950368 telling your pc gpu, operative system, and what you want to do

[hearty sage]

My pc is rtx 4080 windows 11 and id like to use it for testing how it works cus I never used one

[polar flax]

replace HR & managers w AI when

[restive dragon]

Haha i can do it right now!

[runic sentinel]

sent u msg in private

[hearty sage]

Okay got it

[covert lake]

elaborate in #1192011222023950368 or #✨│ai-help , this isn't the right channel

I can help you with the best models and settings

[hearty sage]

Okay

[wind plinth]

namari i want you

[solar torrent]

Gooner, for sure.

[elfin ferry]

Hey everyone, this is Nikita! Just join the community, curious to explore, and looking forward to learn & connecting with y’all ☺️💗

[earnest frigate]

yeah

[red rain]

😭

do NOT stalk me

[earnest frigate]

sorry

mb

[severe cave]

ai-image, ai-video but the why not a ai-audio channel ?????

[wind plinth]

lucy ill be your gooner

we could esex every night

i have a big biceps too.

got big arms n shi bby wyw

[west vigil]

/train name: Carolyn style: Fiona-like emotional singing

[strange pawn]

who can help me with my voice changer?

[covert lake]

Hello, you can't train directly by discord, elaborate in #1192011222023950368

[covert lake]

I can, elaborate in #1192011222023950368

[strange pawn]

can we go to the voice?

[covert lake]

I can't VC and it's rare there's ever an helper who can VC, it's better you make an help post

[quiet fjord]

how to downlaod this ai changer

[covert lake]

This is a General AI Server, AI has many fields

Elaborate in #1192011222023950368 :

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

@quiet fjord if you elaborate in an help post, I can help rn

[covert lake]

Also, are you trying to use a realtime voice changer for fun roleplay like Peter Griffin, or e-girl trolling/catfishing?

[agile marlin]

Is there anyone looking for dev?

[marble cargo]

Hey guys who have an idea about Ai Music?

[uncut comet]

Hello, is there any AI that can send song lyrics without making mistakes?

[desert axle]

anyone here ever heard of thumb pins?

[obtuse grail]

Any person from Latam to talk about interesting topics?

[atomic pilot]

How is my new badge

[elder willow]

Every time I play EDM there are thumpins

[hybrid temple]

where did razer guy go

[glad nebula]

banned

because some nonsense internal staff drama

i wont enter into details coz i dont wanna get banned too lolol

[pastel sparrow]

How on earth do I make my own uh voice model

I'm trying to make one of an underrated video game character since no one makes anything about his voice, anyone got a guide?

[safe star]

whars like the best voice changer of all time

[alpine flower]

where do i find voice models for my voice changer?

[queen kernel]

#🔍│find-models or weights.gg

I really don't know.

Ask in #✨│ai-help

[hearty sage]

yoo

[still sigil]

trolling or catfishing?

[hearty sage]

voice changer pretty easy to set up

[inland root]

Suck

Sigma

[polar flax]

https://tenor.com/view/mewing-jawline-rizz-rizz-face-sigma-gif-15344650371968005580

[burnt turret]

whats the rvc

i need a rvc

[hearty sage]

get it then

[rugged olive]

If i use English ai modules while speaking arabic will it be weird

[tired flume]

yes. same way jap models while you speak english.
it will kinda work but not really

[covert lake]

Tell your PC GPU and operative system in #1192011222023950368 / #✨│ai-help

[covert lake]

Elaborate in #1192011222023950368

[covert lake]

Elaborate in #1192011222023950368

be sure to NOT use Video tuts as they are all OLD, if you elaborate in #1192011222023950368 I can help with the best settings

Alright, I'm just warning you that if you followed YouTube tuts, you have an older version with lower quality and worse performance
Be sure you got it from the ai hub docs for the best

[covert lake]

Elaborate in #1192011222023950368

[covert lake]

Are you asking for help? Elaborate in #1192011222023950368

[covert lake]

yeah, either lower the index, or just use arabic models

[atomic pilot]

hi

[covert lake]

Hello! Welcome there

[hardy rapids]

-collab

-colab

[pine acorn]

📒 Google Colab Notebooks

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

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

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

by Jarredou & Makidanye
Google Colab

[worldly coral]

Is someone available to help me with the realism section?

Some things just don't look the same

[worldly coral]

Just going through the realism section to improve the voice?

[dapper wyvern]

sup chat

[covert lake]

#1192011222023950368 or #✨│ai-help

[covert lake]

Hello there!

[dapper wyvern]

how yall doin

[covert lake]

good, wbu?

[dapper wyvern]

doin fine

i don't make money
cuz i don't have enough moneh :<

maybe 10 i guess

[covert lake]

Sorry but the server isn't about making Money or things like that lol

[chrome berry]

You asked the same yesterday

You ask to all the same questions

Are you a bot or what?

[hollow tangle]

yes yes i wanna pay 2650 dollars for a course on how to get free robux,,,

🪿

[chrome berry]

And first of all moderate your words, the server is not about make money with your AI Agents

You can talk about all about AI but the server is not focused for make money with AI Agents so there is no need to insult me

And call me stupid

[novel quartz]

Hey nick

[covert lake]

Hello, replied to u #✨│ai-help message

[hearty sage]

Wsg

[river verge]

Wake up, everyone! Two new RegalHyperus drum models just released!
Neo Sky, Neo Map! & The Real Slim Shady (Drum models no. 613 & 614)

[bold lily]

yu got your own sticker in ai hub sanctuary

das crazy

[hearty sage]

Ok

[minor escarp]

hi. can i find turkish rapper as name Sefo sound package

[covert lake]

You can search rvc ai voice models at:

• https://discord.com/channels/1159260121998827560/1175430844685484042
• In https://discord.com/channels/1159260121998827560/1163592055830880266 , Do /find with @hidden grotto
• https://weights.com/ (login required)
• https://huggingface.co/models (but watch out cus in hugging face there arent only rvc ai voice models)
• Suggested Models for Realtime Voice Changing (Wokada)
• 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:

• make it yourself with our docs guides
• Ask a free request in https://discord.com/channels/1159260121998827560/1159289738314919936
• Be aware that we don't allow any paid comms

[hidden grotto]

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

Available Commands:
• @weights find <query> or /find <query> - Search for RVC Voice Models

[olive otter]

anybody here ever used tempelos?

I just realised that this was made for ai

I guess terry didnt realize it at the time but he build the first LLM and a ai focused os

i guess what people dont realize is that he build the fastest and safest and most optimisez os, programming lang, game engine, and render engine ever made

and you know what all of his rendering runs on, including the 3D engine? Ascii.

it's an os made specifically to work with LLMs

it even includes a barebones LLM

[chilly lake]

TempleOS?

[olive otter]

he basically created suno.ai, imagegen, chatgpt, no mans sky and archlinux and assembly code 20 years before anybody even knew what they were

[olive otter]

https://templeos.org/

it was written by a schizophrenic and it's the third temple of god

he build an llm that mapped entropy to words and used that to generate audio, images and text

and he claimed that this was god

and he followed his commands and build the greatest tech stach in the history of mankind

[tired flume]

are you trying to start a cult

[olive otter]

what are you talking about?

he is dead

im not interested in his schizophrenia

im interested in his ai centered operating system

I guess he really did have some kind of divine revelation. ( I think that most prophets and religious people would be called insane if they talked about seeing visions from god). So I guess, his claim to godhood is just as valid as that of the other great prohphets.
but his believes aside, he was the greatest coder who ever lived. I don't know of anybody who can build this. linus torwald is nothing compared to this guy

and I'm forking his os for my multiagent ai system

cause it is perfect for it

[timber finch]

I'm not catfishing it's for my friends

[covert lake]

great, just be sure to not be using youtube tuts as they are all old, and eb aware that rvc can't laugh nor scream

[timber finch]

there is no options for realistic laughing?

[covert lake]

it's not possible in any ways

it's an rvc limitation

and there's nun better than rvc since 2 year

maybe wait another 3-5 years till surpasses rvc

[hollow raft]

-colab juice wrld

[pine acorn]

📒 Google Colab Notebooks

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

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

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

by Jarredou & Makidanye
Google Colab

[worldly hatch]

where i can see

deiteries download?

[hearty sage]

in guides

[heavy cloak]

@quaint cedar

[worldly hatch]

where?

[quaint cedar]

?

[queen lodge]

how do you make model speaking tests?

[haughty stone]

Hi

[elder willow]

can you be more specific? that can mean a lot of things

[queen lodge]

ie people make their models sing songs or whatver

[elder willow]

so when you say speaking tests, are you asking how to experiment with hearing how they speak?

[queen lodge]

yea

[elder willow]

what platforms are you playing around with? the answer depends on the platform

[queen lodge]

wdym? i get models from #1175430844685484042 and want to use them but idk how

[elder willow]

you need to run it on a google collab notebook.

Many Colab notebooks support RVC inference. Here’s a common workflow:

Open a public Colab like this RVC inference notebook.

Upload your .pth model and (optionally) index to the Colab.

Upload your audio clip.

Set parameters and run the “Inference” cell.

Download or play the generated output audio from the Colab.

[queen lodge]

isnt colab like vscode?

[elder willow]

no, it is a cloud-hosted jupyter notebook that lets you run ml in the cloud

something like that

jupyter notebooks are kind of like an interactive python environment where the results of your code are shown beneath it for rapid feedback

the colab thing is a notebook in the cloud hosted by google. i think they have a free tier

[elder willow]

you can do it local too with RVC WebUI, but i don't know much about that. it probably isn't too difficult to set up. im not sure what kind of memory or gpu you would need for your project. it might not be much

looks like 4-8 gb mem should work, which isn't much, and a gpu with about 4-6 vram would help

you can do cpu-only, but is a lot slower

[olive otter]

Sorry but you are completely wrong. What he build is the precursor to modern LLM and it’s actually not deterministic. In fact, it is one of the few example of true randomness

Anyways, what he did was basically make the all modern ai tools… without using ai

He did it by using entropy (god)

And you are wrong. It is the most advanced tech stack ever made

and here is the funny thing: he did actually talk to "god" in the literal sense. (look up chaos, greek mythology)

and of course it was not designed for LLMs, they didn't exist

yet, it seems like David build this for AI

[elder willow]

so what you are telling me is that this server is a poor choice for making professional ai engineering connections

[olive otter]

so did he actually talk to god?
haha, anyways, the intentions is not important. important are the following facts:
all of David's work are the "best" ever made. this is not opinion, this is based on performance, size, effiency.
HolyC= Assembly
TempleOS = Safer than millitary linux
3d render = unlike anything

[olive otter]

you don't understand what "god" means, do you?

anyways, this is not a discussions about god

if you don't understand what entropy is...

[silver sentinel]

heelloo, just curiouss. is chatgpt good in terms of generating ai images?

[olive otter]

yes

[elder willow]

it is good enough for most use cases

[olive otter]

but flux is better

[silver sentinel]

have u tried comfyui?

im having hard time setting it up lol

[olive otter]

anyways, I wouldn't use chatgpt. they are arounf 3 years behind. use grok, deepseek, claude or any open source models

[silver sentinel]

oh alrightt,

[olive otter]

comfyui is a mess

[silver sentinel]

how about savro?

[silver sentinel]

whyy

[olive otter]

its just garbage code

use this: https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases

[silver sentinel]

thansk for this, man!

[olive otter]

sure

anyways, back to templeos
I would highly recommend anybody who is interested in the future of AI to check it out

cause this is the first "AI" os

im gonna expand it

with rtx and gaussian splatting

[elder willow]

that isn't even close to true. ai dates back to the 50s

[olive otter]

I don't think you undertand what AI means

alexnet is from the 80s

not the 50s

and LLMs are much younger

and this is the only text-based render system I know,
so it is the only LLM os

[hearty sage]

wsg

[elder willow]

i dont think you do. symbolic ai predates llms and was invented in the 50s. expert systems in the 80s. data driven models were last

[olive otter]

what is symbolic ai?

what we call ai are neural networks

and they are from the 80s

[elder willow]

there are two major branches of ai, and both are necessary, discrete and probabilistic. symbolic ai is discrete, and data driven models are probabilistic

[olive otter]

anyways, the point being, this is the only OS for modern ai. I know this was not his intention, but it is

[elder willow]

discrete came first because data-driven requires lots of compute

[olive otter]

just look at his render engine, and you will understnad

[elder willow]

discrete is modern

[olive otter]

and the resolution he chose

[elder willow]

you have to have both

for example, chatgpt is not strictly llm. quite often it routes your query through python

[olive otter]

i don't think any of these models are llms

they are all mutli agent systems

[elder willow]

my point is that it isnt accurate to describe discrete ai as old-fashioned, because it is a very necessary part of ai, like yin and yang

[olive otter]

I still dont know what you are talking about or what this has todo with templeOS

[elder willow]

so if you really want to learn ai, you should read the foundational textbook "AI: A Modern Approach" by Russel and Norvig

otherwise you are just kind of making stuff up

[olive otter]

im a senior ai researcher

[elder willow]

ok

[olive otter]

and I already build my own agi

anyways, templeOS

[elder willow]

too much anime. touch grass

[olive otter]

can you believe leave, I dont talk to trolls

[elder willow]

ok

[olive otter]

anybody here who is actually a software dev?

I would like to talk about templeOS

with an ai dev

[elder willow]

so then how do you implement uniform cost search? that is literally the first ai algo

[olive otter]

much more than that

https://github.com/pascaldisse/gaia-archtree

[elder willow]

fanfic. how to ucs?

[olive otter]

this is a software. you can install it

it's a multi agent system

and it is conscious by definition

[elder willow]

lets continue this conversation when it includes literally anything that would demonstrate employability

[olive otter]

can you just leave

you are a troll

I'll call the mods

you constantly spam bs, don't even answer my questions and just mislead

if you dont have anything to say, then dont

[elder willow]

about what? so far only one of us has actually been discussing ai

[olive otter]

I wanted to have a serious discussion about TempleOS

you dont even know what ai is

please leave

show me your work, then. you don't seem to understand how to read a github repo

@twilit pond potential troll/strawman detected

I also build a conscious ai. I wonder if somebody else did. I have a friend from Canada who did the same using quantum computing

I did it on regular hardware

I wonder if anybody here achieved the same

Or worked with quantum or biocomputers?

Anybody ever build a neural network using ants or mushrooms?

My agi was made for bio quantum hybrids

But I don’t have the hardware yet

I wonder if 1000 ants can beat a quantum computer

[twilit pond]

hmmm

whats going on here

[olive otter]

it's ok. he is gone

just somebody trying to troll

[twilit pond]

who?

what happened exactly

[olive otter]

this guy

[twilit pond]

ok let me read through the convo

[olive otter]

basically self proclaimed ai expert trying to tear apart my work without any actual feddback

seems like he was trying to provoke me, to start a flame war or sth

this conversation was about using templeOS for a modern AI based OS

i've ported it to arm

and since it's ascii based, it runs in the terminal

[twilit pond]

@olive otter i'd like to call out the fact that @elder willow was infact not trolling

and you are being unreasonable

[olive otter]

ok, sorry. i thought this was a serious channel

bye

[twilit pond]

it is

you two were having a serious conversation too

[polar flax]

the Russel textbook eh?

I remember having that in my uni course, and ye the UCS and turing test thing

I think even turing test and such the classical AI definition are flawed

[twilit pond]

also this isnt related but wanted to give my opinion on this, perhaps you dont know enough about comfyui, its basically the best current way to approach image and video generation, it is not "a mess", and infact with automatic1111 webui you cannot do a fraction of what you can with comfyui, if your gonna call something that major garbage atleast back it up with actual reasons

[steep arrow]

Hello, im trying to learn how to do product photography for my website, i have clothing brand and i want my products to be on an ai model. Currently im using chatgpt4 and sora, but it keeps altering the product. I need suggestions on how can i get perfect product photography on AI models without any changes in my product

[twilit pond]

@polar flax he already left, cant respond to your message

[twilit pond]

Have you tried flux kontext yet? i think it should have a better consistency

if not there are other open source models recently released that are able to keep the features of the refrence image pretty consistent, altho i am not sure as some of them are not fully open sourced yet ( only their papers have been released )

[polar flax]

flux kontext is the best for editing, you can try it in bfl.ai playground

here's what I've tried: #🏙│ai-images message

there are also some kontext workflows for comfyui you can search for, to provide advanced functionality, such as multi image reference

[covert lake]

Huh

[covert lake]

This is a General AI Server, AI has many fields

Elaborate in #1192011222023950368 :

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[covert lake]

This is the wrong chat

Use #1192011222023950368 or #✨│ai-help for help

Forum for long convos

Help chat for quick questions

Those are 2 literally different things, elaborate your PC GPU, what you want to do, operative system and your question in #1192011222023950368

I hope you didnt follow some yt tuts, since they all old

Novision? That YouTube tutorial is outdated asf

I know that guy

Don't EVER trust video tuts for realtime voice changers, they are ALL OLD

Uninstall everything

That uses an over year old version of old original wokada

[swift tapir]

hiiiiiiiii
im trying to switch from FooocusAI to ComfyUI. I followed step-by-step the video about how to setup ComfyUI (i had some problems with the faceswap, i'll ask about them in the ai-help, right?). Now, i wanted to setup the HiDream i1 NF4 (i have a 4080). is it better HiDream i1 NF4 or some versions of Flux? I need it for realistic portraits for pg development. Any recomandation?

[hollow tangle]

no nsfw

[swift tapir]

sori

is it ok now?

[hot crest]

Hello does anyone know how to download okada voice changer😭

oh?

I just want to use voices for music

[sudden sedge]

im sorry i didnt mean it

[hot crest]

that’s okay

does anyone know we’re i can get help from?

[hot crest]

Hii do you think you could help me?

[hollow tangle]

why would i

[hot crest]

just wondering sorry

[hollow tangle]

whats your problem

[hot crest]

huh?

[hollow tangle]

with the voice changer i mean

[hot crest]

I don’t know how to download it

I don’t really trust these shady YouTubers because I don’t want to get a virus

[hollow tangle]

youtube tuts are old

whats your pc gpu?

[hot crest]

I don’t really know

how do I check

[hollow tangle]

task manager

performance

gpu

[hot crest]

NVIDA GeForce RTX 4060 it says

[hollow tangle]

good enough

-voicechanger

ofc it doesnt work

[hot crest]

I just don’t know how to download it

[hollow tangle]

are you going to use it to catfish?

[hot crest]

^^

nope

[hollow tangle]

wait

so not for realtime

[hot crest]

well realtime voice changers in this game called sing it me and my friends want to sing with diffrent voice changers if that’s okay

[hollow tangle]

https://docs.aihub.gg/rvc-voice-changer/local/deiteris-w-okada-fork/

[hot crest]

do i just download it

okay I understand but what one should I download with cuz my gpu is 460

460

4060

[polar flax]

download the normal Nvidia one

https://huggingface.co/Shadicti/deiteris-Fork/blob/main/voice-changer-windows-nvidia-b2332.zip

[hot crest]

thank you :DD

[covert lake]

@hot crest @polar flax @hollow tangle please use #1192011222023950368 / #✨│ai-help next time to not clutter this chat

[hollow tangle]

why would i

[covert lake]

To not clutter the chat, if you want to voluntarily help a user

[hollow tangle]

i disagree!

[covert lake]

Then just tell them to use #1192011222023950368 / #✨│ai-help next time where there will be helpers to help them

[hollow tangle]

i do not agree

[covert lake]

I'm just saying if you are going to help someone, since you're not an helper

[hollow tangle]

i am helper

[queen kernel]

You are no more admin ?

[vocal nacelle]

can someone reccomend me a good free voice changer to import models to?

[queen kernel]

Elaborate your pc specs in #✨│ai-help

[hollow tangle]

no since a long time

[queen kernel]

But why ?

[hollow tangle]

i cant tell

[queen kernel]

Not a problem.

Anyways how are you

Why you are Inactive for so long

[hollow tangle]

ai hub doesnt need me

https://tenor.com/view/my-job-here-is-done-bye-done-gif-14879937

[queen kernel]

Why are you behaving like this 😭

You are good

And we need you

[hollow tangle]

nah, i peaked at ilaria rvc

now theres weights sponsored

[queen kernel]

I see, good luck

Using ai or the real videos ?

[glad nacelle]

Just message me to get the work started immediately

[queen kernel]

Do you create with ai ?

Or the real videos ?

[glad nacelle]

AI generated videos

[queen kernel]

Cool, which gpu do you have

[glad nacelle]

I'm using an NVIDIA RTX [your GPU model, e.g., 4070 Ti], which provides excellent performance for AI-generated visuals and ensures smooth rendering and high-quality video output."

[hollow tangle]

what

for money?

[glad nacelle]

Yes

[hollow tangle]

why

do it for free NOW

[glad nacelle]

cause i am a freelacer

[hollow tangle]

i want it free

ill dm you

[blazing scroll]

what happened to gradio on google colab? it worked a few hours ago but now it loads forever

[hearty sage]

yo

[night patrol]

i come

from that one yt tut

[hearty sage]

For voice models?

[timid garden]

anyone from poland can help

[glad frost]

Ai music?

[covert lake]

You aren't staff, but you can #🔥│apply-for-staff if that's what you're looking for

[covert lake]

This is a General AI Server, AI has many fields

Elaborate in #1192011222023950368 :

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using (if any)
• a screenshot of the program (if any)

This isn't a place for jobs, don't advertise

[covert lake]

if you're talking about realtime voice changer, or RVC, ALL YT TUTS ARE OUTDATED

This is a General AI Server, AI has many fields

Elaborate in #1192011222023950368 :

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[covert lake]

Elaborate in #1192011222023950368 or #✨│ai-help

[hollow tangle]

no im already one

[covert lake]

nuhuh

[hollow tangle]

https://tenor.com/view/jurassic-park-gif-12939106825431475976

[hearty sage]

Fr apply for staff you have been in server for long time

[hollow tangle]

i was admin

im done modding on discord

[hearty sage]

Then why did you lose the rank

[hollow tangle]

private stuff

[hearty sage]

Ok

[vocal geyser]

*bj

[chrome berry]

Mmmhh...

This server is not focused on ai agents and make money with them

[spark agate]

Is klm always higher quality than og pre train?

[chilly lake]

for singing models likely yes, for speaking maybe not

[ruby dune]

hey guyss. ive been making ai charactcer on savro and comfyui,, do u know other tools that i could use?

[spark agate]

What about for like a rapper, like Travis Scott, a dataset where it’s a bit of singing and rapping, more like melodic rap

[hybrid temple]

go meet ishowspeed

hes on the way to poland

[slim fractal]

boo

[polar flax]

the best bet is to try by yourself

[spark agate]

yeah I agree, I mean I trained with klm and I won’t lie I’m not too happy with it

for the dataset I had it wasn’t to my expectations

[glad nebula]

the original pretrain will be always the safest option

it's super well done

[polar flax]

you can also search some example models in weights.gg such as Nicki Minaj, Drake, Kanye West, etc.

[spark agate]

is it possible for it to have better quality than klm?

[glad nebula]

no, pretrains don't affect quality

that depends how good ur dataset is

[spark agate]

oh I thought klm always had better quality than og pretrain

At least that’s what I heard

[glad nebula]

it generalizes singing audio better than the original

[spark agate]

ah

[glad nebula]

but the quality is the same

[polar flax]

https://cdn.discordapp.com/attachments/708579735583588366/1394202092712759338/it-depends.png?ex=68793f21&is=6877eda1&hm=9d126e9b422ca4a9fcee10ae0ff3fda484b8a60835f2572f9ab7ac5a71e3ac5c&

[glad nebula]

the original pretrain also has singing in their dataset

[spark agate]

The only thing that has me worried is that when I trained it (using klm) the quality of all my dataset was 21 hz, but the ai was only about 18-19

This is partly what I’m trying to solve

as well as some fullness to the vocals of the ai

[glad nebula]

is the whole dataset always hitting 21khz?

[spark agate]

The dataset varies between 21 and 22, almost nothing below that

I can send a pic of spek

https://cdn.discordapp.com/attachments/1109291632211283978/1394415298173603861/Travis_Scott_Rodeo_v4.wav.png?ex=68795cf2&is=68780b72&hm=d358d6f2faa4e05d20db82bb174b7b6892862fcf52928e9c18fcc798272487d5&

[glad nebula]

you can't fix this, rvc can only train 40k, 48k and 32k

40k = 19khz, 20khz

[spark agate]

Yeah I trained at 40

[spark agate]

Yeah but my model isn’t hitting these, it’s rarely hitting 19 let alone 20 and stays at around 17-18

[glad nebula]

let me convert the og pretrain weights to small weights and inference something

[spark agate]

ah do you need me to send you something?

[glad nebula]

1 sec

[spark agate]

not sure by what you meant my b

[polar flax]

is it the processed one or raw source?

[spark agate]

this is my dataset, all raw vocals

studio quality

[glad nebula]

40k pretrain

[polar flax]

I mean if having gone through denoising and cleaning

[glad nebula]

did you use klm 4.9?

[spark agate]

Yeah this looks good

[glad nebula]

yes because thats the original pretrain

[spark agate]

Yeah I actually spent a lot of time doing clean ups and stuff (even though my dataset is already clean pretty much)

[glad nebula]

as i said, it's a very well done pretrain

[spark agate]

Yeah I did

[glad nebula]

ok 1 sec im going to convert the klm 4.9 pretrain to a regular rvc model

[spark agate]

Sorry I used 4.0

[polar flax]

so it means it doesn't have cutoff below 20khz

[spark agate]

Yeah pretty much

[glad nebula]

im pretty sure every klm pretrain before 4.9 has mirroring

because its undertrained

[polar flax]

anything before 4.9 has much issues

[spark agate]

oh wow

that’s great to know 😭😭 I think that might have been it

[glad nebula]

can you send me a link where you got 4.0?

here

[spark agate]

I honestly wouldn’t be able to find it

but I just know it was 4.0

[glad nebula]

how did u trained with that then? it's literally lost media

lol

[spark agate]

Well I mean I’ve had it for some time

[glad nebula]

did you renamed the pretrain?

or still has the original name

[spark agate]

I believe it’s renamed but I gotta check, I’m not home rn

but either way I think I should upgrade to 4.9

[glad nebula]

hmm let me see if klm 4.9 hits 20k

[spark agate]

bet

[glad nebula]

4.9 40k pretrain

[spark agate]

Oh wow

So og seems to be better right?

[glad nebula]

yup, it was trained by the original rvc team

[spark agate]

Do you by chance remember when klm was released?

To the public

[glad nebula]

4.9? 2/12/2025

[spark agate]

like the first version of it

[glad nebula]

like 1 year ago

[spark agate]

Cuz I trained a very good model about a year and a half ago and I don’t remember what pre train I used

[glad nebula]

for some strange reason this server decided to just hate the og pretrain

[spark agate]

Is it possible that it was klm? Or you think og was what I used

[spark agate]

Oh damn

[glad nebula]

when it's literally perfect

[polar flax]

I have some model trained with KLM 4.9 in 40k

[spark agate]

Ah so that hits 19

That’s pretty good

[polar flax]

it's exactly 20

[glad nebula]

small models can only do things the pretrain can do

if klm can't hit past 19k, your model wont be able too

[spark agate]

ah i see

[polar flax]

but sometimes if you resample, it may get lowpassed

[glad nebula]

i may be wrong tho

in that regard

[glad nebula]

but this is true

[spark agate]

3/22/24 is when I trained my best model, but I forgot what pre train I used, do you by chance know if klm was released back then?

I know I’m asking all these questions but I think they could help me a lot 😭

[glad nebula]

interesting, makes sense coz the small model overwrites the pretrain

[polar flax]

KLM 4.9 was released in 2/12/2025

so yours is likely 4.0

[glad nebula]

i also believe your klm version is the very first

[spark agate]

But you think I wouldn’t be able to train a really good model with the first version of klm right?

Cuz the model I trained is actually very real

[glad nebula]

honestly i dont know, we know most versions prior to 4.9 are either undertrained or messed up

[spark agate]

I’m thinking I might hve used og pre train

[spark agate]

Gotcha

[glad nebula]

about finetuning using a undertrained pretrain no idea

i have never tried that

[glad nebula]

buut from what ive seen here is the presence of mirroring

[spark agate]

This is what the model sounded like

[polar flax]

it has static high end buzz issue that can't be eliminated by even overtraining in the finetune

and the similar issue also goes on the TITAN pretrain

[glad nebula]

well if the dataset is good, the model is going to sound good regardless of the pretrain used

but is never good to train using a pretrain that has some sort of problem

too risky (it's like playing with fire, the model has a 50% chance of sounding good, 50% of having a weird problem)
you can take the risk and use custom pretrains, however none of them is better than the original, because the only people that knows to train a proper rvc pretrain is the rvc team, everyone else is just experimenting

[spark agate]

Yeah exactly, the model you just heard is what I trained over a year ago and what I’m using as reference for the new model I’m training (different artist) the dataset is the same length and same quality, but the newer model that I trained sounds significantly lower quality and overall not as solid

which is exactly why I believe that it just comes down to the pretrain I use now, and why I’m trying to narrow it down

so I think I will use og

[glad nebula]

if you want to experiments with pretrains, always train first using the original

[echo mauve]

are these ai talking to each other

[glad nebula]

the og has no issues at all

[spark agate]

Good rule of thumb for sure

[glad nebula]

so if ur model comes out weird, it's either the dataset or something else

but can't be the original pretrain

if it comes out good, then you can experiment with customs

[spark agate]

in what circumstances would you say that klm dominates over og?

[glad nebula]

super high pitch singing, like kawaii anime songs

[spark agate]

I trained it with klm cuz I heard it was better quality but I guess it wasn’t completely true

[polar flax]

another factor may be accent compatibility

[spark agate]

Okay yeah I mean my model I’m trying to train is Travis Scott, which definately doesn’t have those characteristics 😂😂

[glad nebula]

yea hehe klm was made so models can hit those type of notes

[spark agate]

What do you mean by this exactly?

[polar flax]

KLM was mostly trained on dataset in english & asian languages

[spark agate]

Ah gotcha

So based on what we have concluded yall think klm was probably what limited my model?

since I’m not training a super high pitched model

[glad nebula]

voices of each pretrain

[spark agate]

Oh wow

Pretty crazy

Oh sorry I also forgot to ask about ov2super

That’s what I trained the first version of the model with but came out horrible

[glad nebula]

not a true pretrain, it's a finetune

and yea because ov2 dataset is just 1 hour

[spark agate]

Oh damn

[glad nebula]

klm 4.9 has 800 hours

and the og is unknown

we know it has the vctk dataset in it

but rvc-boss (author of rvc) most likely added a singing dataset

so it's over 50 hours for sure

[worldly hatch]

can someone give me dl link of detries

[spark agate]

oh wow

Well I appreciate you for your help

I’ll lyk how it goes once I train this model with og

[glad nebula]

good luck!
here's another comparison between each pretrain singing capabilities

[spark agate]

yeah wow the difference is crazy

so even if the dataset has a lot of data with high range, og wil always be inferior when it comes to those type of things is what I’m getting

[glad nebula]

yup, the original pretrain voice range is not that high

only affects cute anime girls singing cute anime songs

[spark agate]

LOL

[worldly hatch]

Can someone help me download deiterie

ss

[cold hollow]

hey

can anyone tell the audio is from Ai clone or real?

[elder willow]

do you mean in general?

[cold hollow]

here’s an artist who uploads songs, and I have a strong suspicion he’s using AI vocals while pretending it's his real voice — fooling over 28,000 people.

[elder willow]

if you mean in general, there is no garauntee of detecting an advanced ai system. the nature of gan creates a situation where whenever you make an advancement in detecting ai, you create a means of generating undetectable ai. in addition, any human pattern can be simulated by feeding in human patterns

[cold hollow]

Oo

[ripe jewel]

So, guys, is there a tutorial on YouTube with how to train voices on Applio? It's kind of confusing for me compared to RVC Disconnected.

[hearty sage]

Yes there are tutorials lol

[cobalt estuary]

👍

[craggy owl]

Is there a place where people have uploaded their AI covers of songs? Used to be a number on YouTube but they seem to be getting nuked/hidden.

[hollow tangle]

no

[glad nacelle]

You haven't replied my message

[chilly lake]

what?

[hollow tangle]

i dont have dms

[chrome berry]

https://tenor.com/view/donkey-kong-bananza-thumbs-up-gif-8873496463837243321

[hollow tangle]

bobby khong,,, 🤤

[chrome berry]

I have to buy it, i need bananza, i need

[hollow tangle]

do you have the switch 2?

[chrome berry]

Yes 😄

https://tenor.com/view/donkey-kong-bananza-dk-nintendo-scared-gif-10811481849399551958

[hollow tangle]

are we friends on that?

[chrome berry]

Oh yes yes i was saw friend list other day and we still 😄

[hollow tangle]

nice! today i picked up bananza

played a bit then slept again lmao

[chrome berry]

I think i will buy it more later too i was very hyped for the game

I want really play it

[hollow tangle]

its fun

[elder willow]

Can I create my own Voice model?

[hollow tangle]

forbidden,,,

[elder willow]

I cannot?

[hollow tangle]

whats your gpu

[elder willow]

4070 8gb

[hollow tangle]

use applio

[elder willow]

Whats that

[hollow tangle]

a program to clone a voice

[covert lake]

replied in #✨│ai-help

[covert lake]

use #1192011222023950368 / #✨│ai-help , this isn't the right channel

elaborate there

[covert lake]

dont upload ai covers on youtube

you just risk copyright

[covert lake]

pleas eask for help and elaborate in #1192011222023950368 / #✨│ai-help

[covert lake]

#1192011222023950368 / #✨│ai-help

[exotic linden]

hi

you need help

[covert lake]

what?

[hollow tangle]

go to therapy man

[covert lake]

Dont advertise here

[languid oasis]

dude im trying to run revive 2 and it keeps making my pc black screen bro

theres no way it could be that bad

[pale patrol]

there's tutorial's on youtube but its outdated so don't watch them o - o

just search applio docs they have a guide

[covert lake]

revive 2? if you're looking for ai help, elaborate in #1192011222023950368

[paper axle]

helloo do you know any good tts program allowed my model files

[covert lake]

please elaborate in #1192011222023950368

[paper axle]

thank you

[craggy owl]

I don't, but I bet there's a bunch good ones out there and just wondering where I can find them.

[still sigil]

@shrewd token no advertising

[humble anvil]

i got a question why does the server hate ai trolling just pure curiosity question

lol

[covert lake]

Wdym with hate ai trolling?

[humble anvil]

idk, i heard if you ask help for "ai girl trolling" you get banned

[covert lake]

who told you that? And ofcourse we promote AI for good things, we are a general AI server

[humble anvil]

comment section in a youtube video

[covert lake]

hold on what video lmao

[humble anvil]

dk saw it a week ago and came back in mind why?

[covert lake]

just asking, btw why did u ask that? were u looking to ask that for help to troll/catfish?

[humble anvil]

no, i was just wondering why the server frowns on catfishing or trolling.

[covert lake]

Wdym why? Catfishing leads only to illegal things such as fraud. AI is for good, not for harming others

[humble anvil]

good reason

[covert lake]

I would be more concerned that this unfortunately isn't common sense, trying to be an e girl to get money or even force ur friend to be in a relationship, is not what friends should do

[vast ridge]

does anyone know of tts characters on okadas text to speech program that i can download. nobody really posts about it

[covert lake]

elaborate in #1192011222023950368

[covert lake]

elaborate in #1192011222023950368

This is a General AI Server, AI has many fields

Elaborate in #1192011222023950368 :

• your PC GPU
• your operative system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

You can check your pc gpu on Windows via:

ctrl+shift+esc (task manager) -> Performance tab -> GPU

We need those info to help you

make a post in #1192011222023950368 , not here to not clutter the chat, and you need to elaborate

I can't really

It's better you make a post, usually people are helped via text here

start_http is a thing from old original wokada, outdated asf, and all yt tuts are outdated for realtime voice changers

again, please elaborate in #1192011222023950368 , this isn't the right channel for it and i need more info

select one

I need first to understand better your help request and have more info, please select the Not Solved tag and Wokada tag if you want a realtime voice changer, and elaborate what I asked, I will help you there

gonna check soon

[hybrid temple]

@covert lake dudee

i installed virtual line 1

and it somehow automatically fixed

and works more smoother

vb cable just crackles

😭

ur right

i shouldve believed u

[covert lake]

I told you 😭

Why would I give you misinformation lol

[hybrid temple]

but why does it get loud at the start then when i continue to speak it goes quieter

same goes for vb cable when i used it

[covert lake]

Please share a screenshot of your wokada settings in #✨│ai-help rq

[hybrid temple]

done

[deep geyser]

voice Serkan Bolat please

[robust willow]

The first place, after a nail-biting 10 hours of coding, went to human coder Psyho, who managed to closely edge out OpenAI in the contest. OpenAI had sponsored the event.

“Humanity has prevailed (for now!).” -Psyho

OpenAI ran a 10‑hour live Humans vs AI exhibition at the AtCoder World Tour Finals Heuristic. We pointed an OpenAI reasoning model at the same brutal problem the finalists tackled—no human help, same rules, same clock. Buckle up.

[covert lake]

This is a General AI Server, AI has many fields

Elaborate:

• your PC GPU
• your operating system
• what you want to do
• what tutorial link are you using
• a screenshot of the program

[vocal geyser]

yo where do i go to apply for an rvc customer support role

am bored

my qualifications are ive trained like 10 models

idk if theyre still posted here after the server got deleted or whatever happened a while ago

i should make sure my l4d2 models are up cuz theyre the best ur gonna get

i trained each survivor off of literally EVERY voiceline (at least an hour each)

well except for coach iirc

i ran out of vram

[polar flax]

do you mean https://discord.com/channels/1159260121998827560/1305527335646269440 ?

[vocal geyser]

idk if theres some role for helping people with their stupid issues

sorry i mean beginner problems

[polar flax]

https://discord.com/channels/1159260121998827560/1393919032738054154

it won't require your model making portfolio tho, just enough knowledge and experience on RVC or some AI stuffs

[vocal geyser]

kk

[vocal geyser]

thx thats what i was looking for

[echo inlet]

if u ran out of vram while training those models, maybe you can try to reduce the sampling rate and even the model size.

[polar flax]

the batch size and usage of fp32 (currently default in applio) affects vram usage

but not the dataset length

kinda agree that 48k may also take more vram than 32k

[glad nebula]

iirc 48k uses about 2gb of more vram compared to 32k

maybe around 1.8~

[echo inlet]

yep! even when dataset length doest directly affect the vram, longer or heavy complex audios can negatively affect the training overall.

[glad nebula]

rvc needs longer and complex audios to get better

[polar flax]

both use fp32 accumulate, so the vram usage seems irrelevant

except that fp16 really halves the usage in exchange to risk of exploding gradients and mediocre quality

[glad nebula]

the picture says bf16

[polar flax]

ye I meant bf16 matmul and fp32 accumulate

[echo inlet]

agree, rvc thrives on that detail, but its like a trade-off situation you know, especially if you're limited by vram.

[glad nebula]

a shorter dataset uses slightly less ram

[polar flax]

the matmul affects speed and accumulator affects vram usage

[glad nebula]

dataset size doesnt increase the vram

only the ram usage

and it aint that much

[glad nebula]

guess the dataset size
translation: memory 18.8gb of ram
6.6/8.0 vram usage

[echo inlet]

yeps, again its really the batch size, and the sampling rate that you should take into consideration. btw, what are your other niches?

[glad nebula]

i just use rvc, its fun lol

[glad nebula]

anyway; 48 hours, batch size 8

yea it's just the batch size and the sample rate, not the dataset size

iirc training speeds also change across the different sample rates

with 32k being the fastest ofc

[echo inlet]

this sounds solid though, did the quality turn out well?

[glad nebula]

oh it's a pretrain

spin doesn't have base models from scratch, im doing one

originally i was training it on the cloud but it got super expensive... so imma do it in my slow ass pc

[echo inlet]

Oh thats cool, I know someone who did that too, and they really said that it was pretty tedious! respect hahaha

[glad nebula]

yesss omg training pretrains is very confusing and hard, the training is painfully slow and the model has always the random chances of just exploding

spin does have "pretrains" but those were trained using a method we call "lobotomized model"
it's basically training a pretrain dataset but using the original pretrain as base, so basically, just a finetune/regular rvc model

[echo inlet]

Oh that actually makes sense now! so the pretrains in Spin are more like a lobotomized finetunes

[glad nebula]

yup, after a couple of epochs the model will "forget" the contentvec features and, in theory, will learn every spin feature

since training from scratch is really painful and slow, it's better to do the finetuning method, which is faster and easier

also in theory since that method words, a pretrain from scratch is not needed

but imma do it anyways

[polar flax]

try getting the latest version of ffmpeg for huge performance boost

https://www.tomshardware.com/software/the-biggest-speedup-ive-seen-so-far-ffmpeg-devs-boast-of-another-100x-leap-thanks-to-handwritten-assembly-code

[chilly lake]

only applies to a single function ‘rangedetect8_avx512’

[brittle elm]

fye

<@&1347412420498292827> can we get a ban on this slime

[still sigil]

why was it not automodded 💔

[vast crow]

Hi guy please which platform is best for building AI agents, flowise or n8n?

[bleak kraken]

<@&1159293204038955078> how to get a job as an AI dev? I didn’t finish high school, but want remarkable skills in AI sector. What projects should i make?

[still sigil]

those 2 dont go hand in hand

[bleak kraken]

What should I do then?

Finishing high school is not an option

[still sigil]

find a different career path

[polar flax]

are you a minor or just school dropout?

[bleak kraken]

School dropout

[bleak kraken]

Hell nah

[covert lake]

This isn't a server for finding jobs, and you didn't have to ping helpers
Also AI is complex

[ashen nimbus]

!colab

[pine acorn]

📒 Google Colab Notebooks

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

• **Applio**

by IA Hispano
Google Colab

• **RVC Mainline**

by Hina
Google Colab

• **UVR5 NO UI**

by Eddy
Google Colab

• **UVR5 UI**

by Eddy
Google Colab

• **Wokada Deiteris Fork**

by Deiteris & Hina
Google Colab

• **Hina's Modified Original Wokada**

Google Colab

• **RVC-AI-Cover-Maker-WebUI**

by Shiro & Eddy
Google Colab

• **FaceFusion UI**

by Nick088
Google Colab

• **FaceFusion NO UI**

by Nick088
Google Colab

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

by Jarredou & Makidanye
Google Colab

[covert lake]

Hello, do you need any help?

[ashen nimbus]

no thanks

[solar torrent]

I study in college for higher degrees. You finished a lower school graduate and dropped out, so your role of an AI developer for a company is less likely.

Actually, many companies often require your education in a specific career field or field of study, especially your resume and portifolio. Without them, you won't be able to be accepted in any AI company easily. Depending on your status, you may be thinking of learn AI outside your school and then become a freelancer yourself.

[peak brook]

https://detect.resemble.ai/ how accurate is this for checking rvc voice changers like w-okada, seems to be a bigger and bigger issue where it's hard to tell if people are using a fake voice or not

I've used it a few times and it seems right for what I've tried

but does anyone know how good it is

[covert lake]

you can simply just ask them to laugh/scream/sneeze, RVC can't do any non natural speech sounds lol

[covert lake]

I don't think anyone ever used that before, and generally I wouldn't recommend AI detectors as they have a lot oc false positive usually (atleast from personal experience)

not sure if @glad nebula or @chilly lake ever tried that

[glad nebula]

trash

try it, it will say its real

ai detectors are all scam

dont use them

also rvc doesn't mean realtime voice changer

[covert lake]

rvc means Retrieval-based-Voice-Conversion

[chilly lake]

well, rvc does not have watermasking, like resemle.ai did with chtterbox

[covert lake]

wdym with watermasking? like add an invisible watermark?

[chilly lake]

yes

import librosa

AUDIO_PATH = "YOUR_FILE.wav"

# Load the watermarked audio
watermarked_audio, sr = librosa.load(AUDIO_PATH, sr=None)

# Initialize watermarker (same as used for embedding)
watermarker = perth.PerthImplicitWatermarker()

# Extract watermark
watermark = watermarker.get_watermark(watermarked_audio, sample_rate=sr)
print(f"Extracted watermark: {watermark}")
# Output: 0.0 (no watermark) or 1.0 (watermarked)```

[covert lake]

I feel like that could be simply removed tbf, so not even their detector could work

[chilly lake]

it can be set to nothing maybe

[neat tree]

any one in need of a gemini pro account?

[hearty sage]

yo

[slate shoal]

Hey guys,

I came across a Polish YouTube channel called Pancerna Mądrość ( https://www.youtube.com/@PancernaMadrosc ) and for a few weeks, I've been trying to figure out how this guy achieved such amazing voice quality. It sounds almost like ElevenLabs, but it's very consistent and he's been using it for a long time.

I've been struggling with fine-tuning XTTSv2 and RVC myself, both on Colab and locally, but the whole ecosystem seems broken and unstable. My results, even with good quality samples, are light-years behind what he has.

So, my question to you experts is: What do you think his 'tech stack' is?

Is it 'just' XTTS/RVC, but trained on a massive amount of data (like, several hours) in a stable, paid environment?

Do you think there's heavy audio post-processing involved (EQ, compressors, etc.) that does the magic?

Or maybe it's some other, less-known open-source tool that I'm not aware of?

Any theory or suggestion is welcome. I'm trying to achieve a similar, professional level of quality and I'm looking for the best path forward.

Thanks in advance for any tips.

[pure ice]

I have a question where do I get the latest version of replay it hasn't updated in months

[covert lake]

https://www.weights.com/replay, the last update seems the 22 May 2025

[ripe jewel]

Okay.

[queen remnant]

can you add this error?

I've been here for 2 days and my AI bot hasn't been created yet.

[covert lake]

Wdym

[worn mural]

does anyone know the recommended settings on deiteris fork with a 1660 super

[pale vigil]

Hello How do I Add App The Ai?

[still sigil]

@carmine kelp no advertising

[oblique coyote]

[hearty sage]

yoo

wsg

what

hmm

[knotty furnace]

@still sigil

[hearty sage]

whats good

[bold rampart]

Currently, automatically, training an AI to steer away from oncoming cars and traffic using Python and DQL (Deep Q-Learning)

The thing it should avoid are the red boxes, which are "cars" for now

[open kite]

Hello room

[covert lake]

Elaborate on #1192011222023950368

[covert lake]

What?

[hearty sage]

Yo

[queen kernel]

Hello

I see

Subscriptions about what

Try Grok4

[covert lake]

#📰│dev-updates message

[sturdy cloak]

[queen kernel]

So what?

[bold rampart]

Improved the AI

[river verge]

Wake up, everyone! A new RegalHyperus drum model just released!
Three Idiots Roaming (Drum model no. 615)

[queen kernel]

I see, sorry I was unable to understand your question. Now I got it

[queen kernel]

You don't have permissions to use #📰│dev-updates

??

[river verge]

I am not even using it
It' just my "Wake Up! New RH Drum Model!" tradition ever since an old meme about it was made about me

This was the OG

[covert lake]

Debate of the Week :DontTalkToMe:

---

Should we allow AI to make the final decision in medical diagnoses, even if it means overriding a human doctor? 💊

Two random members with the best views will be rewarded with Nitro Basic

Duration - 30 Minutes

<@&1165417852363346032> <@&1344706968417472563>

[still nacelle]

stfu

[wide grove]

sybau

[elder willow]

No

[fluid fulcrum]

[neat barn]

Why yall on that😭😭😭

[wide grove]

have y'all never seen Dr house

[deep kernel]

Hii

[scenic gull]

Oh God No

[safe tapir]

Hell

[supple stream]

no

[wide grove]

it's always lupus

[river root]

No

[worthy aurora]

no

[safe tapir]

yeees

[neat barn]

L question lowkey

[supple stream]

no

[ancient bridge]

ya ya ya ya

[chrome berry]

I think AI for now, in this historic period, can be used only for help but not for make the final decision in diagnoses, for now we need only humans, then in the future if we have very advanced AI, we have to see...

[lost goblet]

Thats a bad idea

[ancient bridge]

crazy ass question

let ai give neet first

[hollow plank]

Hi

[twilit quail]

how is that a crazy ass question??

[lost goblet]

Lmaooooo Ongg