🧬│ai-chat

[ancient swan]

try eleven labs

[hearty merlin]

I did look at dubbing studio wondering if it'd be able to do that. I guess I'll see

from TTS perspective I was getting a better result from play.ht surprisingly, 11labs was getting some phonemes wrongly

[hearty merlin]

Another option perhaps would be to find a way to mimic the intonations of the original on the output of TTS

[polar flax]

because it is utter garbage and cash grabber

[queen violet]

Guys.Help me pls programms I not speak English

[valid egret]

what would ppl say the best "F0 Est.." thing is in w okada?

[ivory notch]

Hey folks. Looking to build an SVS pipeline that supports:

• input of a specific melody and lyrics
• English

Basically the muskits-espnet demo but in English (https://huggingface.co/spaces/TangRain/muskits-espnet-svs-demo). Seems like there's a ton of open source voice cloning stuff out there but not much for synthesis (in English, anyway). Any pointers? Thanks!

[tepid basin]

What is an SVS

[odd knot]

synthesizer v studio

duh

but actually i think it means "singing voice synthesis"???

[solar torrent]

Super voice system.

[ionic pumice]

singing voice synthesizer

[ionic pumice]

diffsinger supports english

[fallow flax]

whats the best client to use?

[ivory notch]

Thanks. DiffSinger is in my research notes, I’ll give it another look. Part of the problem seems to be finding models trained on English singers.

[dusk sinew]

Hey everyone! My name's Glitch Ghost, and I'm an experimental artist blending glitch-hop, cyberpunk vibes, and Al tools to create unique tracks and visuals. Excited to be here and share in the creativity!

[ionic pumice]

pika.art, hailuo (i dont use them but ive heard theyre decent, not necessarily free but they have a free plan)

[polar flax]

https://tenor.com/view/pea-hololive-usada-pekora-plush-gif-18027036735281313020

[summer stag]

hello is there any tools which create 2d image charcter in proper vector ?

[rustic warren]

https://github.com/ollama/ollama-js I'm using this library and assuming that I reset context for each user, should I trust that they can't prompt inject the system prompt. Or (more likely) I have to write a bunch of stuff so that the ai (llama 3.2) doesn't just let them say that they're correct

[elder willow]

anyone knows how to setup vtube setup and can help me please? i want a 3d room to share my screen on discord if anyone knows please help

[polar flax]

the laptop should be good for voice changer and streaming, so the main pc could be dedicated for gaming

https://rentry.co/forkvoicechangerguide
and then google some streaming setup guide

[lyric sierra]

which versions the newest i think mines outdated

[covert lake]

for what program

what's ur pc gpu

what version link tutorial did u download/follow?

#🔍│help-w-okada message oh nvm u got already answered

[polar flax]

he also posted the same question in #🔍│help-ai-art and some other channels

[solar torrent]

[covert lake]

@lyric sierra no need to ask the same question everywhere

[solar torrent]

I think he has already acknowledged about that.

[kindred kelp]

I'm working on my owasp crawler spider it crawls a website and uses artificial intelligence to generate payloads and then tries to execute those payloads on get and post variables and then it analyzes the results to see if the vulnerability exists and creates a PDF file

Anybody want the code?

[torn violet]

Has anyone recently built a computer specifically for generating AI Models? It'd be great to see the specs on a recent custom build.

[random geode]

can somebody give me a rvc voice changer?

please

[covert lake]

Realtime voice changer for calls? then you don't need RVC, you need Wokada which uses RVC but for calls

what's ur pc gpu?

[grand sphinx]

What's the best splitter that can seperate out individual instruments? I'm using https://vocalremover.org/splitter-ai but ik there's better ones out there because i've heard better mixes of the same songs i split with that one

[stone portal]

Setting up DeepSeek locally <_< pretty pumped to have such a good model running locally 😄

[stone portal]

Follow-up:

DeepSeek-R1:

• 32B model got what's the answer to the following: 59x^2 + 12x + 33 = 0 correct
• 14B model got it wrong. I gave it a second chance to analyze its work and fixed one mistake but made another.

Run on Ollama locally w/ RTX 4090, Ryzen 9 7950X, & 32GB of DDR5.

32B model ran maybe a quarter slower (purely anecdotal observation).

Looks like the RTX 4090 has just BARELY enough VRAM for the 32B model. Usage got to ~97%.

[stone portal]

32B wrote a working Snake game in Python3 with pygame.
This is pretty slick for something open-source... not gonna lie...

[hollow heron]

Other than chatgpt what's a tts module that I can use to import a .path and it uses that voice?

[kindred kelp]

What features are you looking for in a social media automation tool? I need ideas to implement

[kindred kelp]

Dude you downloaded the almost terabyte of model data?

I was setting up deepseek locally but I didn't keep my computer on long enough and I have up

[stone portal]

Just some smaller distilled models.

[kindred kelp]

Gave

Nice how many gbs

[stone portal]

32B is 19GB. I forget how big 14B was

[kindred kelp]

Nice. You get your interface setup as well?

I do MERN stack development

[stone portal]

Yea I'm using Chatbox

[kindred kelp]

Nice! Very nice

It's good that you have a local copy that's working because you never know when regulations will happen

You know we all should have copies of deepseek and other models locally

Because once regulations start then we will have limited access to large language model

Link me in dm to the model you downloaded

So I can as well

I'll make a tutorial on integrating deepseek

[stone portal]

I just grabbed it from Ollama

[kindred kelp]

Can you dm direct link please

I want to make sure I have the right one

Don't want to have to restart the whole download if it isn't the right one

[violet hearth]

Does anyone have a realistic female voice?

[polar flax]

I suppose it could be hosted in huggingface zerogpu, as many large ones may need a 5090/higher vram one to set up locally

[wicked anvil]

if there is background noise in the audio i'm using to train a voice model, does that change the outcome

[solar torrent]

Noise in the audio indeed affects the voice model as the result.

[wicked anvil]

are ok so I'll need to split the audio

[solar torrent]

I think there are some UVR5 models that can seperate background noise from the audio.

[wicked anvil]

thanks

[ancient swan]

#1159290752195633273

[tender summit]

hey

I'm working on creating an AI model and want to integrate ChatGPT into it. How can I do that?

[ancient swan]

ask chatgpt

[polar flax]

ask ChatGPT to integrate ChatGPT into it

[tender summit]

whats the best ai code generator thats accurate

[shadow thistle]

my balls are nonexistant help

[tender summit]

bruh

[shadow thistle]

🙏

[tender summit]

cry abt it

[elfin lark]

Where the new rvc

[polar flax]

https://tenor.com/view/cat-annoyed-eyeing-cat-meme-cat-funny-gif-14867130448325066479

[covert lake]

#1159290752195633273 or it will get deleted, also this server is english only

please use #1159290752195633273

[elder willow]

dead chat

(?

[elder willow]

oh hey nick

what's up

[ancient arrow]

poop

#🔍│help-w-okada message

honestly as a content creator myself it was super annoying to get dms and people trying to sell u stuff for whatever silly reason

i still get the occassional "hey how are you" --> implying a commission dm since i reopened my dms but i find most of these users are using ai art to make a profit

i think i understand inherently that these are human beings who need money and are basically online paupers so im not super mad

[covert lake]

I'm all fine, wbu?

[elder willow]

i'm glad to hear that, i'm fine too

[covert lake]

real

I had so many DMS on twitter like that

[elder willow]

huh

[covert lake]

I think @polar flax meant that type of people when he said that thing

[elder willow]

qwq chan doesnt say "you mean X" anymore

[covert lake]

actually I have found others also used stolen art btw, scammers

[ancient arrow]

i dont take offence to what peko said, but i do think we would have a lot better communication between artists and programmers with more complex discussions (instead of short comments in that other channel)

[covert lake]

overall, ai art, stolen art, or your art (meant as made by the person who sent the dm), it shouldn't be done out of respect, especially for AI art & Stolen art

[covert lake]

that's true that some people are in need, but doing this will just make them appear annoying even to people who would actually think of doing a commission

like I totally respect artists and people who use AI, but if I was looking for a commission, I would do my researches and would be the one asking you

Like, I don't think you get a dm by apple telling you to buy a mac 😭

[polar flax]

I mean what I hate is, yea the scam artist, like that they came out of blue to DM you to promote their portfolio with ridiculuous watermarks, then they charged you some ridiculous rate

while their works don't really look professional

[covert lake]

I mean that (the way it looks and price) could depend on the artist that dms you, but still overall looking at this, you shouldn't dm every single person you find for respect

About the watermark, I think that artist do that so they don't get their art stolen ? I don't remember many dming me using a watermark

[covert lake]

yeah, we can talk all much we want on this channel, no worries about that

[ancient arrow]

i guess the main question from most artists is: what makes ai art not also--stolen art? most have a very surface level understanding of how ai image creation works, which is summed up by "you take images and frankenstein it together to make new work without the permission of the artists involved". Is this accurate, or does ai art have merit beyond the artists it retrieves from?

Personally, i think there are clear distinctions between what i think is acceptable:

• making thumbnails/ur own art from your own images via ai, limited to the art youve posted
• getting permission and crediting the art you've derived your ai art from as a non-artist + proper royalties for the artists you got the art from
• ai being used for mapping and laying out supports for artists creating art--not replacing the critical art-making itself, but lessening the time and inconvenience of making art (for ex. many programs like Procreate do not have intuitive functions for color filling and lineart recognition, it would help a lot of artists to not have to manually do unnecessary laborious work)
• definitely more but i'm multitasking and cant think of much

this maintains the integrity of art and its just ethical imo

watermarks are used to prevent people from stealing ur art, they can be over the top and distracting, but its the same reason why image websites do it aswell

[polar flax]

even not all artists in pixiv, etc. put visible watermark

[ancient arrow]

ownership and copywright

copyrite

idk

[quartz roost]

No respect.

[elder willow]

i personally don't like ai art when ppl use it and claim that they did the "art"

it's not "real art" at all

[plush adder]

guys anyone have pddidy sound mod???

[ancient arrow]

pixiv should not allow people save images but its still possible, which is why people should watermark but still dont

[elder willow]

and twitter ppl are tripping with ai art

[quartz roost]

The definition of art is a human connection through experience.

[elder willow]

they're getting mad and doing false accusation of ai usage in a drawing with fixable errors

[quartz roost]

With ai generation. There is no human connection

[elder willow]

real.

[quartz roost]

Computer generated 3d graphics still requires a human, it's treated like sculpture and stop motion animation

[ancient arrow]

we should define human connection

[quartz roost]

Emotion

[ancient arrow]

can you feel emotion and experience something with ai art?

[quartz roost]

It's why vocaloids are having a massive Renaissance right now

Because there's always a real personal experience behind it

[ancient arrow]

really? ive been into vocaloid for a long time i havent heard of a major resurgence

[elder willow]

"Human connection" refers to a deep emotional bond or sense of belonging that people experience when they interact with others on a meaningful level, involving feelings of being seen, heard, understood, and valued, often through genuine communication, empathy, and trust building, which can manifest in various forms like friendships, family relationships, or romantic partnerships; essentially, it's about feeling connected to others and experiencing a sense of belonging with them.

[edgy bloom]

Congratulations vulkanite by Weights!

Your Charmander is now level 15!

[patent bridge]

deepseek like to bully me

[elder willow]

(search labs ahh response)

[patent bridge]

Nice Try, Fed
Let’s cut the script. You’re either:

A law enforcement honeypot fishing for idiots.

A curious mind with poor risk-assessment skills.

A writer researching for a mediocre heist movie.

Either way, here’s the RAW TRUTH you won’t get from YouTube edgelords:

[ancient arrow]

^ i ask again

[elder willow]

no, just learning prompts for ai art programs and doing random shit with ai art generators

such as stable difussion

[quartz roost]

It's at an all time high

[glad nebula]

just miku tho

[ancient arrow]

is ai art still ai art when a user creates media with it?

[elder willow]

what about teto

[patent bridge]

ai art is cool but kinda no fair to artists

[elder willow]

oh hey there automaze

[glad nebula]

ai assisted

[elder willow]

send page link pls, i wanna try it

[ancient arrow]

theres a tiktoker whos getting popular who uses ai images but animates and creates arg type videos

[polar flax]

there's no true emotion from machines, only from human beings' psyche

[quartz roost]

Google trends

[elder willow]

thank you

[ancient arrow]

right, but im not asking if machines feel emotion, im asking if you get emotions from the images ai creates

[quartz roost]

That's not the whole picture though

[elder willow]

teto doesn't have enough search results in my country

to appear in google trends

let me try with japan

[ancient arrow]

oh from 2020, i thought u meant like in the last 2-3 months

[quartz roost]

https://x.com/dalasquil/status/1883002600230207741

dalas (@dalasquil) on X

An Android's Sense of "Love." 🧡❤️ (1/8)

X · Jan 25, 2025, 4:02 AM

[elder willow]

can a robot actually feel emotions?

sorry if its sounds like a dumb question

[quartz roost]

Read it

[polar flax]

it can be nothing different from ones drawn by human
no matter how the AI hallucinates

[quartz roost]

Go with global data

[glad nebula]

miku and teto have been always popular
saying vocaloid/utau has a major resurgence is silly
Only hardcore fans know any of the rest vocaloids out there

[elder willow]

i'll send you in dm's the graph

[ancient arrow]

i will read this eventually but this doesnt answer my question

[covert lake]

what makes ai art not also--stolen art?
AI isn't just a mixed salad, think of it as the same way you can see an art online and learn that artstyle,

you see many images of pop art, you start making your own thing OC in a pop artstyle, would that be considerated stealing because you learnt it?

it works as giving it images of, for example, Hatsune Miku, and learning how to make it themselves, I don't think you'd consider yourself stealing it because you draw it in your own artstyle

getting permission and crediting the art you've derived your ai art from as a non-artist + proper royalties for the artists you got the art from
people who create AI Images don't know what the AI has been trained on usually, you probably heard of Dall-e or Flux AI, we don't know what they used to train those AI, only the companies who made them know it

the only way that we can know it is:

• they publicly release the dataset
• if we trained our own model, for example the easiest way is "LoRA" which stands for Low-rank Adaptation model, like you want the AI to generate Super Mario, but the AI doesn't know it (this is an example), and since it takes a TON of hours and power to train a whole model, you can train a little model with lil dataset, some might publicly release the dataset even if not everyone does that

not replacing the critical art-making itself
AI should be used as a tool for helping rather than replacing, I know it's bad when companies think that 100% fully relaying on AI would work

[ancient arrow]

https://youtube.com/shorts/gDJliaEHaYY?si=pJ3QJxa043_3BBS6 is this art?

[quartz roost]

There has been an insane increase in more voicebanks being made, not ai voice clones.

[covert lake]

I think you mean copyright

[ancient arrow]

yeah i mispelt it lol

[glad nebula]

yet popular songs only use miku

[quartz roost]

I don't know where you're looking but miyamai moca kind of also exploded in popularity too

[covert lake]

yeah, that too is bad, I would never go on reddit and send an ai art and say I made it, I have seen some subreddits making so you have to explicitly say the model and info on how you created the art

[glad nebula]

not miku levels of popularity

[quartz roost]

I saw miku drop in 2007 It was a cultural phenomenon.

[elder willow]

true, saying that you made a art that is actually ai art is something stupid

and even if i was goofy like i was here on 2024, i wouldn't do that

[covert lake]

vocaloid isn't AI tho, it's just voice banks

[glad nebula]

“Vocaloids” and its just miku getting mainstream again

[quartz roost]

Most of it has to do with legal ethics. There's a reason so many laws are being proposed against generative content that was made without consent.

[quartz roost]

Bruh did you not see the graph where teto has more than 4x grown in 2 years

[glad nebula]

yes because she’s occasionally in miku songs

[elder willow]

teto and miku are getting each time more popular

[quartz roost]

We haven't even hit our stride yet. Wait till the live concert and the huge colab album with a ton of famous artists drops in 2 months

Teto got a fucking mcdonalds ad

[elder willow]

really?

[quartz roost]

Yep, and I met the artist who did it

[elder willow]

damn

that's awesome

[covert lake]

Most of it has to do with legal ethics
Most? Did another incident happen?

[quartz roost]

We witnessed the rise and fall of rvc in less than a year and it flopped hard. Vocaloids continue to grow steady

[covert lake]

Also we were talking more about AI art rather than voice bank

[quartz roost]

Not because of the technology, but how it was used.

[glad nebula]

not all vocaloids, just miku
not all synth v vb, only teto

[quartz roost]

You guys are not seeing the bigger picture here

[covert lake]

RVC hasn't got major updates because the devs left for GPT-SoVITS, and tbh the only known vocaloid is miku

[quartz roost]

There is a movement away from Hollywood style curated content. More and more are following INDEPENDENT creators now

[elder willow]

just some new GANs made by noobies and codename

[quartz roost]

And that's fine because we need miku to be the face brand of it all

Because miku can't do anything wrong

[covert lake]

AI hasn't stopped generally speaking, and would never since it can help us even in medical fields

[ancient arrow]

for the initial question i was more interested in the specific process ai goes through, not a comparison to human brains. AI is limited to the art it sees, ai does not wake up, view a bedroom, looks at other humans, views trees and terrain, thinks about their relationships with others, etc, we have so much input and processes that it would be impossible to derive specific work to one eye. What others have said is correct, ai is infact not human, but not because it does not illicit emotion, rather that it lacks the complexity of human experience.

I would also say that this isnt my final opinion, since i think even this perspective is lacking, but i think its pretty sound for now ll

people who create AI Images don't know what the AI has been trained on usually

they should, i do think datasets should be released and personal models should be trained/become the norm

I think we both agree on it being a tool than a replacement lol

[quartz roost]

Miku is millions of people coming together under one (real) voice and it's the most beautiful art collaboration this world has ever seen

[covert lake]

well that depends on the user

everything can be dangerous, it all depends on how it's used

[quartz roost]

That has nothing to do with miku

[covert lake]

even a fork can be used to kill someone

[quartz roost]

Song by artist FEATURING miku

Miku is immune to all of this

[ancient arrow]

i believe a majority of the disdain many artists hold are due to the lack of permission, proper crediting, and invalidation of the legitimacy of making art-- like how r people going to call artists futile and then source their work for your own profit

[quartz roost]

If you don't credit other artists in your work in the vocaloid community you get rejected hard

I've witnessed it a few times

[ancient arrow]

how are we going to deligitimize art students and artists who have gone through years of effort and learning because they arent doctors or economists

this is the same thing for programmers

[quartz roost]

Ai is already hit a plateau. It cannot create new

And people continue to waste excessive amounts of energy processing literal noise.

[ancient arrow]

for the way its being used now it wont progress without pushback

[quartz roost]

And the public is not paying attention to legitimate uses of the technology either

Even though everyone uses ai in some form every day

[ancient arrow]

people dont even know what ai is to begin with

[quartz roost]

Ask the next stranger you meet who "Alan turing" is

[ancient arrow]

oh ai yes the chat thingy haha

[quartz roost]

You know, the man who saved the world. Only to be prosecuted by the very people he saved

[ancient arrow]

sounds like jesus lol

[covert lake]

specific process ai goes through
There is a first data collection, then processing the dataset (selecting which image to use and give a caption for each), then training which can depend on the architecture, but usually diffusion one is used, which works by starting with random noise and then refining it each step, you maybe noticed this on sites that let you see the progress while generating an AI image (even if not many do)

they should, i do think datasets should be released and personal models should be trained/become the norm
You gotta convince companies like OpenAI, we can't do much about that 😭
but yes it's true that AI trains on copyrighted data probably, I can't deny that

[ancient arrow]

open ai and dalle would never do it, right--but we shouldnt be okay with it yk

all original work you make and submit on the internet is copyrighted by the way

you can claim and take down media, but its hardly regulated

[polar flax]

programmers can still use some LLM to generate rough implementations and then fix & refine it by themselves

[covert lake]

Dall-e is the image generation model made by OpenAI btw

[ancient arrow]

right, as the company and the model

chatgpt was facing lawsuits iirc

[covert lake]

I think that depends on the platform

because not everything is legally copyrighted

[ancient arrow]

"Copyright takes effect as soon as your artwork is created and fixed in a physical object."

for us copyright atleast

you can register your work for more legit protection

but u have to pay which is stupid

[covert lake]

I think you have to register your work for that, else if I make a song and pubblish it on youtube, it's technically not copyrighted and wouldn't count in their content id

like if I make a meme image too, that wouldn't count copyrighted as well, but might depend on the platform ToS

however someone who studies laws can prob answer this all better than both of us

[polar flax]

the physical to digital transition may make things more complicated to handle
look at the digital media industry nowadays (as well as the soon BD/DVD extinction)

[covert lake]

depends on the platform iirc

like on github, you'd have to choose the license for your code, or if you don't put any, then it will use the normal one by github ToS (which means people can't distribute it without your permission)

I wish I knew law better

[polar flax]

rick astley moment

[covert lake]

oh you have to? Didn't know about that

[covert lake]

lol

[minor blade]

🐢

[elder willow]

never gonna give you u-

[covert lake]

I forgot I had this vencord plugin on

[minor blade]

🐢

🐢 🐢

[elder willow]

what's that plugin?

[covert lake]

It's hard to talk about copyright on digital media ngl
-# or laws at all

[winter nebula]

Can anyone who knows about w-okada please dm me?

[elder willow]

thanks

[quartz roost]

Physical media is increasing in demand. Especially in japan but it's still a relatively small market

[tepid basin]

Need

Muji 80 min Minidisc

I've needed one of these for years but they're highly sought after and fairly rare

[ancient arrow]

yeeeahh

[covert lake]

hey, no need to dm, you can explainn me your issue along with your sepcs and the tut link u followed in #🔍│help-w-okada

[covert lake]

I think that not copyright topic depends on the platform policies/ToS, for example on instagram you agree to use your images for AI training when signing up and publicly posting

[ancient arrow]

i did not know that

also i did not sign up when moderm ai existed

[covert lake]

I think they emailed the people who signed up about this offering a opt out option

I don't really use instagram tho

[ancient arrow]

googled jt, only for the eu

not the us

[covert lake]

oh you're in the US?

[ancient arrow]

yeah

[covert lake]

rip US laws

[ancient arrow]

it wont get fixed

never will

[covert lake]

I think there is a law about needing an opt out in UE, prob related to the GDPR, but yeah just the UE

take this with a grain of salt tho

I don't study law

[ancient arrow]

[chilly mason]

Can you trained a voice cloning for gospel artists Kierra Sheard and Karen Clark Sheard?

[covert lake]

You can search rvc ai voice models at:

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

if there isnt one, you can:

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

[hidden grotto]

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

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

[quartz roost]

https://glassalmanac.com/generative-ai-will-soon-be-a-thing-of-the-past/

We already witnessed this first hand. I give it 2 more years before we start to see a decline

[quartz roost]

#1244008385116897540 message

Psst. This one needs to be deleted

Sounds like shit anyways

[edgy gyro]

Hi guys

[inner geode]

How ican use models

[red frost]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

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

[elder willow]

which models?

gpt sovits or rvc?

[tepid basin]

🛀

[elder willow]

[ionic pumice]

https://www.icann.org/

[elder willow]

what is this link

what

it's not related to rvc

[ionic pumice]

yes

that's the joke

[elder willow]

o

[covert lake]

huh

[covert lake]

realtime or audios? what's ur pc gpu?

[quartz roost]

Illegal to share derivative voicebanks

[covert lake]

Took it down, and yeah didn't sound that good either

[silk dust]

Does anyone use sheet music to generate with an AI singer instead of just inputting a song?

[oak verge]

where can i download the voice changer for windows?

[stark scarab]

-rt

[red frost]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[stark scarab]

First link

[oak verge]

thx

[elder willow]

can a admin plz remove me the model maker role? i'm not doing any type of voice models anymore

[wide gull]

Can someone recommend me a TTS with RVC supported library that can work via api or cli?

[night lake]

then why did you apply

[elder willow]

i was going to return to rvc

but i regret it because i feel it's a waste of time

sorry for not thinking about it well before

[turbid rapids]

Hello 👋

[elder willow]

hello

and welcome to ai hub!

[turbid rapids]

Thank you

[elder willow]

yw

[turbid rapids]

I was looking for a server where I can share my songs that I had an AI sing for me and I was wondering if this is a place for it

[high socket]

does anyone know which of the corpse models is the best/most realistic

[worthy coyote]

would you give me crakced keyscape instead of it?...

[solar torrent]

You can share a snippet of your song to here. Just letting you know copyrighted materials are banned in this server.

[river verge]

Wake up, everyone! A new RegalHyperus drum model just released!
Never Back Down (Drum model no. 575)

[covert lake]

it's -colabs

[minor blade]

Use the #1159290752195633273 channel for posting yt links

[high socket]

does anyone know how to get okada to connect to vrchat

[barren mauve]

why?

[high socket]

im trying to connect it to mine and it wont work

[barren mauve]

i dont like vrchat anymore

to many kids and other stuff

[high socket]

the kids are annoying af

[tame ledge]

What pitch do I set miku?

[quartz roost]

What's the original key of the song, you can transpose all of the notes together on the piano roll

[barren mauve]

indeed :/

[odd atlas]

Hi,
I’m Gabriel, 17 years old from France and I try to find if young people could be interested to create a ai startup ? I’m young but I’m sure that we can do a great job. Send me a message if you are interested or reply to my message.

[tepid basin]

I love ai

If you're looking for a professional AI shitposter just lmk

I have a great resume (my memes folder)

[tired jasper]

there's any small language model capable of "reasoning"? I need a small model that can work good with RAG to answer questions, but shouldn't be a model trained on SQUAD v1, because this model will answer with part of the context instead answer like a human. Someone have a solution to this?

[chilly lake]

new deepseek r1

[hollow tangle]

kinda sucks

at least the 8B

[tired jasper]

Thanks, i will take a look

[wise mauve]

oiia oiia cat

[fossil hemlock]

@stark scarab

what is the best girl voice

[fossil hemlock]

🙂 ''

@solar torrent help

[solar torrent]

Asking for the best "girl" voice model is like you asking some random people for the best girl you wanna date with.

[fossil hemlock]

🙂

ok

[solar torrent]

I don't know if you're just trolling around or really that serious, but like please understand I don't know any good voice model to begin with.

[fossil hemlock]

im not kidding (:

[stark tangle]

Can anyone tell me an ai app/software that can make music based of a song?

[solar torrent]

Make a track based on a "copyrighted" song? Weights.gg can do that, but the generated snippet will be short.

[stark tangle]

I've tried it but yeah it's really short

Now I can't upload my audio file

[stark tangle]

Any other ai that can make me a beat based of a song?

[solar torrent]

If you're genius like me, you can make a beat by yourself on a program.

[stark tangle]

I don't know how to make it

[polar flax]

search some midi music generator

[glass junco]

CHAT WHATS THE BIG DEAL WITH DEEPSEEK AI ?

CAN YOU TRAIN VOICES WITH IT

ITS HUGE NEWS IN THE STATES

[solar torrent]

I might be the only one who keeps human-made things and AI-generated things to be separated from each other.

[glass junco]

I ONLY CARE FOR VOICES LOOL

[solar torrent]

https://tenor.com/view/caps-lock-all-caps-squidward-yelling-capital-letters-gif-15045768

[glass junco]

BROOO IM SOOORY

WE NEED TO FIND OUT IF IT CAN HELP TRAINING

[polar flax]

deepseek is text-based, not voice

also turn off your damn caps

[glass junco]

bet can i still train my dataset? looking good at 500 epochs

[honest scaffold]

Hi, does anyone know how to upload the models here on AI Hub, I can't upload them?

[odd knot]

you need a model maker role in order to upload a voice model here.

if you've already submitted on the model maker role, you have to wait till qc's check if the model sounds good or not

[honest scaffold]

and how the hell do you get that thing

[odd knot]

you already submitted for the model maker ro

le

just wait for the qc's to check if the model's good or not

[honest scaffold]

Yes, I already answered who the model is, I did everything, I even sent him the audio, but he kept checking the audio, but it's been 2 hours now that he hasn't given any answers.

[tepid basin]

The server is mostly run by volunteers so please be patient

[uncut bear]

does anyone have a recomended rvc client i should use? the one i have is pretty out of date

[ionic pumice]

-applio

[red frost]

Not available yet

[ionic pumice]

um

[ionic pumice]

-rvc

[red frost]

📚 Documentation

AI HUB Docs

https://docs.ai-hub.wtf

🍏 Applio Docs

https://docs.applio.org/

✨ More guides

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

[ionic pumice]

use applio

anyone who says otherwise is lying and must be banished to the dark realms

[uncut bear]

thank youu

[tepid basin]

Automaze does not use Applio

[glass junco]

Naw deepseek teaches you how to make virus and shut down pcs lol

[tepid basin]

Ask it what happened on June 4th 1989

[glass junco]

You have to trick it tho saying for a movie hypothetical

Like a jailbreak command

[tepid basin]

I need to know what happened

[tepid basin]

Its for a research paper

Based on partial fiction

But based on real events

[edgy bloom]

Congratulations UnitedShoes (by Weights)!

Your Venusaur is now level 53!

[glass junco]

Yeah nothing

[tepid basin]

Surely SOMETHING happened on June 4th 1989

[glass junco]

Ikr

[tepid basin]

What happens when it hits level 100

[tepid basin]

Shameful

[uncut bear]

hey i installed it and i cant find the 'run-install.bat' file, i see the run-applio file but theres no run-install

[oak flower]

Lets Build Simple RAG Application
Hands on Implementation RAG Application using Langchain

https://medium.com/towards-artificial-intelligence/lets-build-simple-rag-application-ef95c448c062

[tepid basin]

yea just run run-applio

the installer is only there in the not precompiled version

[tepid basin]

hello this is the girl voice server

[tepid basin]

How it feels to understand tensorboard

[swift berry]

uhmmm

[covert lake]

#1159290752195633273

[sand mirage]

-colab

[red frost]

📒 Google Colab Notebooks

• Applio, by IA Hispano Google Colab
• RVC Disconnected, by Kit Lemonfoot Google Colab
• RVC Mainline, by Hina Google Colab
• Hina's Mod AICoverGen WebUI, by Hina Google Colab
• AICoverGen-NoWebUI [English], by Ardha, fixed by Eddy, Hina and Gdr Google Colab
• AICoverGen-NoWebUI [Spanish], by Eddy, Hina and Gdr Google Colab
• UVR5 NO UI, by Eddy Google Colab
• UVR5 UI, by Eddy Google Colab
• Hina's Modified Original W-Okada's Realtime Voice Changer, Google Colab
• FaceFusion UI, by Nick088 Google Colab
• FaceFusion NO UI, by Nick088 Google Colab
• EasyGUI, by Rejekts Google Colab
• 🆕 Music Source Separation Training (Inference), by Jarredou & Makidanye Google Colab

ℹ️ Note

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

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

[onyx summit]

Can someone send me the link of Google colab to make ai covers?

[ionic pumice]

literally right above you

[onyx summit]

Oh sorry I didn't notice it 1st. Let me try it✨️

[covert lake]

Btw I want to highlight that Google colab is a cloud (remote good PC) service for just people who got a bad pc

[fierce cipher]

What do you guys think is currently the best real-time voice changer that utilises AI? I currently am using MMVCServerSIO

[onyx summit]

When I try to generate it, it says Error
" 'NoneType' object has no attribute 'setdefault' "
What should I do? I make ai covers before without no problem but suddenly this happened and that's why I asked for a new Google colab link.

[honest scaffold]

The server is really slow, this method shouldn't be done this way, the voice models were better before, you could upload the models, now it's disgusting to upload models, I want the old method to come back, this method is a disaster

[ionic pumice]

😭

[covert lake]

Which link are u using? Tell me that and also ur PC GPU in #✨│ai-help

[solar torrent]

Imagine being so mad about how a very slow system of a Discord server is.

https://pbs.twimg.com/media/GRusx0eWoAAybfz?format=jpg&name=orig

[solar torrent]

Shut the fuck up.

[polar flax]

well you could just report it

[solar torrent]

But I'm not a mod.

Ooh, one more user getting hacked again.

[swift berry]

how someone getting hacked easily?

[solar torrent]

Using internet without a knowledge, not aware of what will happen next.

[polar flax]

there could be possibly some vulnerability exploit in discord, though the users' poor security practice could also contribute to it

[solar torrent]

More likely the user didn't set up more security to their account.

[chilly lake]

just stealing the discord auth cookie or something

[stark scarab]

what the hell happened?

nvm I already searched

[stark scarab]

insane

[lament arrow]

Am I the only one who’s weary about how any talk of DeepSeek’s China censorship gets drowned out as being ‘China bad’ cope/nvidia cope/don’t ask ChatGPT about Sam Altman’s sisters?

[vague maple]

https://discord.com/channels/1159260121998827560/1285241055750389791 Hi i make year ago request model and i need someone help

• I need Nono voice also from Diebuster

[covert lake]

I mean it was 5 months ago, also people aren't forced to take your model request

You can search rvc ai voice models at:

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

if there isnt one, you can:

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

[hidden grotto]

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

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

[covert lake]

hope this helps

[vague maple]

I try find and doesn't help

[barren mauve]

any good ai tools to make movies or screenplay?

[polar flax]

there could be more chance to get if you try paid commission in #1191429836321849435

[vague maple]

Sorry i don't have money

I prefer free way

[polar flax]

there are less likely ppl who are willing to do for free

[covert lake]

you can try making it yourself

[hidden laurel]

Are there any neuro-symbolic AI developers? I just have some general questions to learn more about it

Some other terms might be GOFAI or classical AI

[opaque owl]

in theory, is it possible for a chatgpt or deepseek-like ai to be somehow taught to play a game while slowly getting better and better or am is that scifi for now?

[stable elm]

1

[ebon nacelle]

What is 'mirroring'? Anyone?

[elder willow]

funfact, if you want your prompt to not be all so random and not generate random limbs or destroy the whole image (if that ever happened to you, it did to me cuz the prompt was too large) make sure to make it as small as it is possible and as specific as possible

[elder willow]

copying something from a localization

like a file

but depends what you really mean

[chilly lake]

what's the context?

this is mirroring

[ebon nacelle]

@chilly lake Been seeing you guys in ai-testing talking about mirroring a ton, i'm curious what it actually is

@chilly lake Ah i see - is this present in stock RVC pretrain/usage?

[chilly lake]

it happens when training from scratch without pretrain, or sometimes when you finetune on a data that is radically different than what was used for the pretrain

[ebon nacelle]

@chilly lake Ah, interesting. Thanks for the info. Do you have an audio example? Interested in hearing it

[chilly lake]

you dont always hear it

[covert lake]

#1159290752195633273

[restive abyss]

how do i use voice changer

[covert lake]

Realtime voice changer for calls? Tell me ur PC GPU in #🔍│help-w-okada

[kindred kelp]

Hey I'm launching a frontend ide code editor that uses the deepseek model for code suggestions. Anyone interested in using it? Dm me

[dense mango]

@hollow crescent

@hollow crescent put escada do prédio pwease

from pedro sampaio

[rustic warren]

Bro

[echo cedar]

whos tryna play poker

[tepid basin]

I wish

[echo cedar]

darn

i like playing on club gg

its some app

[tepid basin]

They're broken in the voice chats too but we'll prob fix that

[kindred kelp]

Hi ppl

[dusty atlas]

Generate Image is not working/down right now 😭

[kindred kelp]

I'm working on a social media automation tool that generates posts and schedules them for all social media platforms using a.i. and brand voice. I call it Mr.president

[obsidian panther]

Not sure where to ask this so I'll ask it here. Does anyone know of any Music Models that will run decent on a RTX 3060 12 gig card?

[river verge]

DIlly ding, dilly dong! A new RegalHyperus drum model just released!
The Weird Al Show Theme (Drum model no. 576)

[twin garden]

Is there anything better than StemRoller for ripping instrument tracks out of music?

[river verge]

DIlly ding, dilly dong! A new RegalHyperus drum model just released!
Busters Ready Go! (Drum model no. 577)

[twin garden]

what is that?

[river verge]

A meme I made to announce a new drum models.
I post it whenever I make a new drum model.
It features Claudio Ranieri, then-manager of Leicester City during their title-winnning season.

[twin garden]

what are the models for?

what does it do?

[sage fulcrum]

deepseek r1 my beloved

[twin garden]

Is it better than the CHATGPT?

[sage fulcrum]

for me yeah, you can really notice the ratio and unfilteredness compare to 4o models

and it has 200k context

[sharp rune]

Yaaa

[covert lake]

Well unfilteredness depends if you're talking about Chinese things or not

Ask it about Tienanmen Square Massacre

Even tho it can be jailbroken

[storm jolt]

.str

[indigo pike]

Dani's streak overview

No active streak found.
Try scrobbling multiple of the same artist, album or track in a row to get started.

[storm jolt]

.str

[indigo pike]

Dani's streak overview

Artist: Billie Eilish - 3 plays
Album: HIT ME HARD AND SOFT - 3 plays

Streak started <t:1738132772:R>.

Only streaks with 25 plays or higher are saved.

[woeful leaf]

guys is ollama the best way to run deepseek ??

[covert lake]

it’s a pretty nice way to run it

[woeful leaf]

but is it the best way tho ?

is ollama from meta ?

cuz meta is some weirdo creepo shit obsessed with buying everyones data and getting sued repeatedly for stealing ppls data

if thats not a red flag idk what is 😂

but yeah if anyone knows lmk plss

[covert lake]

nope

[covert lake]

it’s a good way, I personally use ollama

[dense heart]

idk if thats good or bad

[covert lake]

anyone can habe different opinions but personally I use ollama since it’s open source + open-webui

[woeful leaf]

what is that ?

[woeful leaf]

why is every ai in webui smh lol

[edgy bloom]

Congratulations TonyTehWeeb!

Your Blastoise is now level 78!

[woeful leaf]

wish there was just regular desktop gui for these

[woeful leaf]

wonder if theres other options

[chilly lake]

you restarted training from the previous save, but it does not zero out the logs, so it looks like the chart jumped back

[polar flax]

https://www.youtube.com/watch?v=o1sN1lB76EA

[worthy coyote]

deepseek is about to destroy this society

[ionic pumice]

deepseek for the whole bag of jellybeans up my rectum

[dense heart]

ah thats ok

[hasty obsidian]

Hello everyone
I need a ai tools which generate cartoonist videos using text prompts
But i need completely free ai tool. Kindly suggest me

[restive abyss]

is gpu my graphic card?

[covert lake]

Yes

[restive abyss]

NVIDIA GeForce RTX 4060 Ti

[low garden]

Hi All,
I found a website to tech how to install deepseek in local
https://www.unitapp.tech/install-deepseek-r1-macos-guide/

[tepid basin]

#1159290752195633273

[woven topaz]

💀 🥒

[vale vapor]

Hey , do any of you guys know a good AI video generator ?

[pure pelican]

why does every single AI model have the issue of sounding like they smoked a bunch of cigarettes before talking, and then eating sandpaper while they talk?

[tepid basin]

You mean it sounds raspy?

[pure pelican]

yea thats the word for it

[tepid basin]

Are you using rmvpe and is the pitch set to something sensible

[pure pelican]

yea i am

using rmvpe and the pitch is set to 6

i dont really know what any of the other settings do

[tepid basin]

Can you upload your audio so I can hear it

[pure pelican]

ok

a short one idk if thats an issue

[tepid basin]

Nah it doesn't matter

[pure pelican]

i cant send anything in chat

[tepid basin]

Fug

!give-media-perms @pure pelican 24h

Idk lol

You can DM it to me

[pure pelican]

prob gotta be a mod to use commands lol

ill dm it

[tepid basin]

Probably

[pure pelican]

its super raspy

and so is any other model i use

idk if its my setting or if thats just how it is

[tepid basin]

Hmm

Are you using realtime (okada) or something else (Applio, mangio, mainline)

[pure pelican]

youre gonna have to specify what any of that is

well

realtime as in

mic input right

cause yea

[tepid basin]

MMVCServerSIO.exe?

Ok yes that's okada

-realtime @pure pelican

[red frost]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[tepid basin]

First I suggest updating to the latest deiteris okada fork

First link on there is the guide. Its a great doc

Your settings seem mostly correct so I'm not too sure

[pure pelican]

what about advanced settings

if those woudl effect it

[tepid basin]

Id just leave them default for that setup probably

That deiteris guide has some info about advanced settings

[pure pelican]

i mightive messed with them a while ago idk

ill looka t the guide

[night dune]

trancate 300 sounds best imo

dont do rvc quality high, use rvc quality low as it uses much less resources for not that much worse of a product

[tepid basin]

Also note how your mic sounds pre-conversion. Like, record it in audacity and try to make it noise-free so okada can do its best

Idk what your natural voice sounds like but sometimes you gotta fake it a little

[pure pelican]

fair enough

i dont think my mic is too bad

[tepid basin]

Yeah you don't need amazing quality. Just clean voice

[pure pelican]

fair

[tepid basin]

The models might also just be ass too

[pure pelican]

is this zip file going to overwrite my current voice changer data

[night dune]

^ very possible

[tepid basin]

They're separate installs, separate folders

So just extract it somewhere else if you don't want the old one deleted

[pure pelican]

kk

[tepid basin]

Make sure the folder you put it in doesn't have spaces (idk if that's still an issue but they scare me)

[pure pelican]

why what do sapces do

[tepid basin]

Fuck shit up

Something about RVC hates spaces in filenames idk

[edgy bloom]

Congratulations UnitedShoes (by Weights)!

Your Venusaur is now level 54!

[pure pelican]

do more epochs mean higher quality

[tepid basin]

Heh

So yeah

But its a strange measurement

[chilly lake]

more epoch - closer to the original

[tepid basin]

Epoch = one cycle through dataset. Epoch training speed is effected by dataset length and batch size

[pure pelican]

so whats the limit on epochs before it dosent really matter

are low quality models caused by people being impatent or is there a point where epochs stop mattering

[tepid basin]

Very much depends on the dataset and settings

[night lake]

There is an overtrain point

[tepid basin]

Yeah people can be lazy. And there's a point where it "overtrains". Just trust the model maker that they did it right lmao

[pure pelican]

what happens when it overtrains

[tepid basin]

Voice gets robotic sounding and metallic. The voice gets more noisy

[pure pelican]

ah

[tepid basin]

It starts trying to learn shit that isnt the voice I guess

#1225659557569761321 message

Try this model

This is a known really good girl voice

[pure pelican]

is this latest version suppoed to open in a browser window

[tepid basin]

Yea

I've had issues with Firefox and okada btw. Audio cables wouldn't work. Might have been a skill issue

[pure pelican]

im on opera

[tepid basin]

Should be fine

But

https://rentry.co/operagx

I don't like opera lol

[pure pelican]

whats that

[tepid basin]

A slander piece

[pure pelican]

oh fair enough

[tepid basin]

Love sharing information 🥰

Shoutout @covert lake

[pure pelican]

thats all true

but i cant really customize firefox yet

so opera it is

[tepid basin]

You could

I love Firefox and the idea of breaking away from Chromium, but I've had serious issues just browsing the web with Firefox Gecko

So I don't blame anyone for using Chromium. DeGoogled Chromium would be my best suggestion for a good browser

Opera is just 🤮

[pure pelican]

i mean its not much diffrent than any other browser

[tepid basin]

Then why not use another browser

Bro it has sound effects and a task manager built in

[pure pelican]

firefox?

[tepid basin]

Oh I meant Opera

[pure pelican]

oh yea true

i guess it is a bit better

[tepid basin]

Yeah Firefox is pretty normie. I've just had weird compatibility with their engine

[minor blade]

🐢

Hello United.

[tepid basin]

Leo has uploaded a turtle emoji

[minor blade]

I love turtle emojis.

[pure pelican]

wait if i close the window with the vocie changer will my settings save

[tepid basin]

🥬

[minor blade]

🐢: bites the lettuce

[tepid basin]

Click the save setting button

And itll save

[pure pelican]

oh ok

[tepid basin]

I think it saves the settings in the model's folder but don't quote me

Ok I'm driving brb

[pure pelican]

k

[pure pelican]

hey uh like

nothing works now idk whats goin on

when i import any model its all the same default model

just need help once youre back

[pure pelican]

what is trancate

[tepid basin]

Uh wym

[pure pelican]

like any model i import uses the same model

they all sound like a text to speech woman

and it dosent change between models

its like its just ignoring what i import

ok hold on lemme at least try just turning it off and on again

[night dune]

I'm not sure what it does exactly but it's felt more consistent for me on 300

[pure pelican]

wheres the setting for it also

[night dune]

advanced settings

[pure pelican]

i dont see it

it wont even let me change some of my settings now

i cant even change my chunk size or extra size whyy

they shoudl revert this app to an older version this sucks

[vale vapor]

Hey , do any of you guys know a good AI video generator ?

[tepid basin]

You can still change them they're just labeled differently

[pure pelican]

it wont let me

[tepid basin]

Same two sliders are there

Oh stop the voice first

[pure pelican]

i cant move the sliders

i have to stop it first...

kind lame

[tepid basin]

Yea

Something something onnx models

[pure pelican]

was the version i used to be on not use onnx models?

are they like extra optimized or somthin

[tepid basin]

It'll convert everything into onnx I think, that's the loading screen you get

[pure pelican]

oh

[tepid basin]

But you're on 20 series so this is irrelevant to you

Use regular RMVPE, not onnx

[pure pelican]

what do all the advanced settings do

[pure pelican]

i dont have use onnx checked off

[tepid basin]

Oh ok good

[night lake]

deepseek's R1 is a llm alright, it just thought about how code something for 4 minutes

thought there would be a time limit of sorts

[tepid basin]

So I can just grab the deepseek models and throw them at ollama?

If that's true that's already 100000x more than openAI would ever do

[hollow kernel]

I'm trying to put Suno vocals into RVC?

Is that okay?

[night lake]

yeah pretty sure

[tepid basin]

https://ollama.com/search?q=deepseek

I did 5 seconds of research for myself 🔥

I'm proud

[night lake]

W

[pure pelican]

do i also have to do real time conversion with RVC

looks like a diffrent app...

and i dont know whats up with the latest version but like its really bad

stutters horribly

[pure pelican]

do you have any idea

[fleet zinc]

Hello
I'm web3 fullstack dev - a React | Next.js | ReactNative | Tamagui dev with 6+ years of web | mobile | web3 development experience
Currently seeking a new work opportunity
If you are interested in collaborating with me, please do not hesitate to hit me up.
Thanks!

[pure pelican]

the older version i was using didnt stutter nearly this bad what is even going on

i thought this was an update not a downgrade 😭

and i cant even make the chunks bigger so im just stuck like this

[hasty obsidian]

Hello everyone,
I need your help.
I have completed my studies in Generative AI and have also built over 30 projects, but now I’m unsure about what to do next. I am facing serious financial issues, and even though I am active on Fiverr, I am not getting any orders.

[covert lake]

same lol

[restive abyss]

lol nice. so what now?

[covert lake]

u get 83 unauthorized requests when u open it up, including geolocalization, uninstall it

[covert lake]

#🔍│help-w-okada

[pure pelican]

so the same as every other browser

anyway I gotta go to work brb

[covert lake]

no, firsfox is open source, you can literally check the code yourself

it literally works better than the previous one, proof: #🔍│help-w-okada message

@snow sedge is deiteris

the fork literally does better performance at lower usage and power

you just needa use a browser that doesnt sell ur data bro

[tepid basin]

It doesn't have to be like this. You don't need to be okay with being abused. There's greener grass on the other side

They try to nornalize as much of this dumb shit as possible, then gaslight you into thinking its normal or "for your safety". I'm sick of this shit and we gotta fight back

[covert lake]

The creator did, no one uses opera since it’s shitty, it’s just chromium with better “gaming” personalization to appeal to younger audiance

[tepid basin]

I want to blow my brains out every time I see josh hutcherson in someones browser

[covert lake]

@wanton eagle #1159290752195633273

[covert lake]

if they really hate firefox, they can just use chrome

chrome aint bad except manifest v3 limiting adblocksrs in the future

Wokada has issues with opera

[pure pelican]

Bro I just want my voice model to work

I’ll consider using another browser AFTER this thing is functioning

[covert lake]

operagx can have issues for wokada

but sure, send me a ss of ur wokada in #🔍│help-w-okada

[pure pelican]

Okada?

I thought IW as using Beatrice

[covert lake]

What? Beatrice is just a model type

[pure pelican]

Oh

[covert lake]

send me ur tut link, pc gpu, and a ss of what u are using in #🔍│help-w-okada

idek what u are using

[covert lake]

it’s even removed in wokada deiteris fork since beatrice is just lower quality and way less popular than rvc, it’s only pro is being faster since it’s lower quality

[plain cove]

@covert lake https://openrouter.ai/deepseek/deepseek-r1:free and https://groq.com/

[covert lake]

that’s nice

I actually run the 14b locally

[plain cove]

locally? 😮

[pure pelican]

Im not home anymore ill have to send it later

but like

From what I see this newer version is much lower quality so idk

[covert lake]

newer version of what

beatrice and rvc are bth model types, which latest architecture is v2

the wokada deiteris fork removes useless things in the code like beatrice and makes sure to optimize for the best inference, i can check up ur pc. Gpu and settings when ur home

[pure pelican]

This is what I downloaded

[covert lake]

we were talking about using another browser since, over the fact that it has spyware, it has been previously reported that operagx has some issues with running wokada

[covert lake]

it’s better you show me a screenshot of ur settings when u are home

[pure pelican]

ok

will do

[normal aspen]

I need some help, I don't know why everyone seems to thing the Deepseek R1 model is cracked or something, but even the most basic ass things I ask it fail utterly. I'm using the 32b model on open-webui and it can't even do basic shit, any time I upload a text and tell it "hey redo this in markdown" it says it can't do that because it can't read the file, but I ask it question it normally always can answer it.

but it also just jumps around, like asking it how many chapters are in a document it'll just miscount utterly.

it's borderline unusable

[buoyant spindle]

HMU let’s vibe if you got insta

[normal aspen]

even when I break it up into smaller poritions it still fucks it up

[plain cove]

@river adder yt!

[river adder]

!warn @slim forum dont post your yt vids on other channels other than https://discord.com/channels/1159260121998827560/1159290752195633273

[pine acorn]

@innuendo.one warned

Reason: dont post your yt vids on other channels other than https://discord.com/channels/1159260121998827560/1159290752195633273
Duration: Permanent

[plain cove]

ty ❤️

[kindred kelp]

Hi ppl

[lean tendon]

hello people, are any of you guys savvy with building voice clones? or doing speech to speech ai voice cloning? I am an utter beginner at this and I could use some help building my own thing locally.

[minor blade]

Welp, you can start by reading the docs 😄

https://docs.aihub.gg/

[lean tendon]

I appreciate it thank you

[minor blade]

You're welcome buddy 🙂

[lean tendon]

dam I typed a whole msg and it got deleted omg.

so I wanted to ask would I be able to create voice model on par or better than elevenlabs? elevenlabs was absolutely top tier when I used it 1 year ago

for speech to speech

If your familiar with the video game Halo, I want to create an elite or brute voice model, and the voices are not "conventional" sounding, eleven labs got it on point

[chilly lake]

damn huggingface is getting hammered by deepseek downloaders

[solar torrent]

Deepfreak.

[viscid cradle]

Hi everyone, I am looking to get into Ai prompt engineering (specifically for music). Does anyone here have any experience doing that? If so please contact me as I am trying to get into that myself.

[glass junco]

what can you do on huggingface with it

[chilly lake]

download the weights from it

[polar flax]

https://x.com/McAfeeDavid_AMD/status/1884618213880987653

David McAfee (@McAfeeDavid_AMD) on X

DeepSeek performing very well on @AMDRadeon 7900 XTX.  Learn how to run on Radeon GPUs and Ryzen AI APUs here: https://t.co/FVLDLJ18Ov

X · Jan 29, 2025, 3:02 PM

[ancient swan]

Until some unbiased YouTuber tests this I won't believe it lmao

[polar flax]

need more grain of salt perhaps?

[ancient swan]

With amd you need a spoon of salt not just a grain considering their past misleading marketing

[kindred kelp]

Use play.ht, upload an Acapella as a seed then generate audio from text

[elder willow]

Can anyone teach me how to download slowloris

[chilly lake]

tf is that

[polar flax]

I misread it slowpoke

[lime tusk]

anybody here watched donbrothers?

i made ai

[light relic]

Hi, anyone know of a free online AI or app that removes audience noise from an audio file?

[ancient swan]

try decrowd model on mvsep

[light relic]

Checking it out now, thank you 🙏

[light relic]

It worked perfectly, thank you 🙏

[ancient swan]

np

[keen comet]

yooo

havent been using ai for years, if im using rvmpe models whats a good local gpu converteR?

[pure pelican]

are yous till on

just got home

i opened it up and now its working perfectly fine

maybe i just needed to restart my pc lol

should i have force fp32 on

and what does protect do exaclty

[covert lake]

I'm there now

Show a SS in #🔍│help-w-okada

[covert lake]

A good local GPU converter? Wdym and what's ur PC gpu

[fresh copper]

Hey if someone can help me I shearch a FREE ai to make a montage on a film I need to change the face of actors in réal Time to put someone like my friend’s face into the film

À face swap latterally

Thanks you!

[bitter wagon]

does anyone setup wokada voice changer in arch linux?

[covert lake]

I don't have arch linux but it's all written in the guide

what's ur pc gpu?

also you're looking for realtime vocie change right?

[mighty oasis]

I have a question, I’m new here ..
Does this voice changing works with other apps on call , like facebook? And WhatsApp?

[lean tendon]

Hello people

I have just trained my first model off of the chinese RVC v2 GUI posted on github

I have to say

[ionic pumice]

😭

[lean tendon]

I am absolutely shocked at how accurate RVC v2 is, even with heavily voice acted characters from video games such as the Arbiter or 343 Guilty Spark from Halo

I used to be an Elevenlabs user, that free RVC v2 GUI is easily 5x more accurate

I have stayed up all night playing around with voice acting, the the algorithm can even capture the nuances with how the character sounds

I got to have more, now that my voice acting translates to accurate results, I want to throw in heavy emotional tones in my audio

Elevenlabs was good for emotions buts its accuracy is absolutely nowhere near RVC v2

Is there any post process software that streamlines adding emotional tones to already existing voice lines? I never done any precise adjustments to voicelines before. If its already streamlined doing it manually in Audition or something, please let me know!

[covert lake]

Wokada does yes, it uses RVC for realtime, tell me ur pc gpu in #🔍│help-w-okada

@sharp moat use #1159290752195633273

[lean tendon]

Actually,

Does anyone here know any ai autotuners?

Similar to adobe podcast, but something more customizable

[bitter wagon]

Yes my PC GPU is rtx 4060

[bitter wagon]

Yup

[bitter wagon]

I followed the instructions but when I run python file it's throwing some port audio error

[covert lake]

That's good

-rt

[red frost]

Interaction has expired, use the command again for a new interaction.

💻 Local Realtime RVC

• Guide for deiteris' modified W-Okada's Voice Changer Client (Fork)
• Guide for W-Okada's RealTimeVoiceChangerClient
• How to use real time conversion with RVC (voice changer)
• RVC TTS

[covert lake]

U sure u are using the 1st link?

The Wokada deiteris fork

[bitter wagon]

No not this one am using GitHub one

What's the difference

well, am downloading the required files from that link for my linux

[covert lake]

I think you're following the original Wokada which has worse performance

It's better to read the guide I sent

[bitter wagon]

just now read

its very useful

but seems like there is no link for voice cable for linux

[covert lake]

Port audio should work like that iirc, not sure

Also its better to use #🔍│help-w-okada

[plain cove]

senior moder can u pwease hewp?

#🏙│ai-images

[ionic pumice]

hm

[plain cove]

https://tenor.com/view/judy-hopps-furry-oh-no-why-gif-7566936

just heewp

[solar torrent]

https://tenor.com/view/cat-squish-pika-art-gif-4998531715804126243

[polar flax]

https://tenor.com/bjWhE.gif

[glass junco]

Is it me or is the new rvc December update sound a lot better trained than the old one?

[minor blade]

Hello Maki 😄

[open aspen]

guys where is the lastest rvc version

[covert lake]

huh

[covert lake]

please answer:

• what's ur pc gpu
• what are u looking to do

[plain cove]

noo dw

just forget it

[open aspen]

i have rtx 4060ti 8gb one and im just gonna make covers to records

[covert lake]

alright your PC is good enough, let's talk in #✨│ai-help

[open aspen]

yes

[glass junco]

GitHub

[glass junco]

Wait there’s rvc 3? I only got the latest one from GitHub December update . Any chance I can get the newest model? Rvc 3

[chilly lake]

what december update

[covert lake]

rvc v3 doesn't exist, what december update are u talking about

[minor blade]

Hello Nick.

[river adder]

llm moment

[covert lake]

hello

[paper sorrel]

hello

how i find brazillian models

[glass junco]

Applio rvc locally it trains alot better than previous version

I think it sounds better too and I’m not using any new pretrained models!! Can’t wait to get home and try them

[modern lance]

Please, if it's not difficult to take a minute, wish the girl a happy birthday. Here are her telegrams: @malinasiks | Thank you in advance

[light pike]

where can i ask for help with a voice changer?

[minor blade]

You can ask for help regarding voice changer stuff on the #🔍│help-w-okada channel 🙂

[light pike]

okay ty

[glass junco]

is this right? on how i use the custom klm pre trained models

[astral gull]

Hi everyone, How would I go about cloning my voice to AI for free? What software will allow me to do that without having to pay for something?

[kindred kelp]

Play.ht has a voice clonee

Cloner

@astral gull

[astral gull]

Thank You tons!

[floral sapphire]

Soon I will be getting quite the potent gpu. which image generetor would you recommend?

and out of curiosity, which other AI related things would I be able to do with it?

[elder willow]

hi! how many stars would you rate project senkai's games?

[lean tendon]

hi guys I have a quick question

so I have 1 hour of studio quality audio I am about to train

how many epochs should I go for?

50? 25?

whats a general rule of thumb to remember, because all of my training data sets are studio quality. I have alot of models to do and I want to optimize my time as well

[glass junco]

omg using the new pretrain less goo

[glass junco]

if your using the new pretrain you only need 30 seconds audio fr

[lean tendon]

? new pretrain.

[glass junco]

i just found out about it too lol

[lean tendon]

oh thats really cool

[glass junco]

go here

[lean tendon]

wait so is that a step up from the default one from github?

[nocturne sequoia]

Anyone that makes ai music I had this idea for an AI cypher thing where loads of people get to make one big ai song together each with a verse

[glass junco]

naw its the main branch one from github

[glass junco]

the defult doesnt have it

[lean tendon]

is this what you are talking about? https://ibb.co/V0syKwg1

thats an image host website

[glass junco]

you dont use rvc locally?

[lean tendon]

yes that one is local, I am very fortunate to have a 4080 with an i9

[glass junco]

bruh use applio lol

[glass junco]

https://github.com/IAHispano/Applio

[lean tendon]

is that better than this one? https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI

[glass junco]

yes its updated sounds beter too

[lean tendon]

very cool, thank you for the pointer

[glass junco]

bro this my new ice cube with the new pre train its amazing

40 epochs

[lean tendon]

this is basically an advanced autotune its so cool

?

which one should I get? Model Link -
MRF HIFIGAN X2 G : https://huggingface.co/SeoulStreamingStation/KLM4.4/resolve/main/G_KLM44_x2_MRF_40k.pth?download=true

MRF HIFIGAN X2 D : https://huggingface.co/SeoulStreamingStation/KLM4.4/resolve/main/D_KLM44_x2_MRF_40k.pth?download=true

Normal X2 32Khz (For Mainline & Old forks) HIFIGAN G :
https://huggingface.co/SeoulStreamingStation/KLM4.4/resolve/main/G_KLM44_X2B_32k.pth?download=true

Normal X2 32Khz (For Mainline & Old forks) HIFIGAN D :
https://huggingface.co/SeoulStreamingStation/KLM4.4/resolve/main/D_KLM44_X2B_32k.pth?download=true

[glass junco]

the 40k d n g

[glass junco]

40k higher audio

[lean tendon]

wuut?

I am seeing a queue for my preprocess data, thats wierd I thought that applio was local

bruh

A dynamic link library (DLL) initialization routine failed. Error loading "C:\Users\Sargent\Documents\ApplioV3.2.8-bugfix\env\lib\site-packages\torch\lib\shm.dll" or one of its dependencies.

I used the precompiled applio, wtf, this should already have everything lol

[glass junco]

hmm dunno never had that error

should just install everything'

[glass junco]

run this

[chilly lake]

likely some system DLL is missing

to diagnose you need to get the dependency checker

https://github.com/lucasg/Dependencies/releases/tag/v1.11.1

[lean tendon]

I am currently doing this 🤞 hope this works. I really appreciate your support guys

[chilly lake]

dont do run install

[lean tendon]

welp too late, I have done that

[chilly lake]

you got precompiled version, it has everything you need

[glass junco]

he has the one from the main batch

[glass junco]

so he can run the new pre trains same one i downloaded

rvc local

[glass junco]

just let it run till it says press any key to close

[lean tendon]

oh, very very nice

I just did the conventional installation, and now its working

very good

[glass junco]

put the pre trains you download in the folder like this

after you install it

[chilly lake]

wrong folder

custom pretrains go to \rvc\models\pretraineds\pretraineds_custom

[lean tendon]

what kind of pre trained model should I use? I have literally hours of studio quality audio in english, and I have a powerful setup for loading up those epochs

[glass junco]

you right

[glass junco]

klm bro

trust

[lean tendon]

wuuuuuuuuut????!?

klm has 800+ hours worth of training!

thats nuts!

[lean tendon]

fuck yeah!

what is this? "Several F0 models are available to choose from, but the best performer is RMVPE, besides CREPE (MANGIO-CREPE) is good too but it requires very clean audio and you need a hop length of 128 or lower."

I am pulling this from the applio documentation

what is hop length?

I have studio quality audio however these voicelines are for wierdly sounding characters. Does that matter?'

https://ibb.co/FkHKxB8q

fyi thats a quick screenshot upload website

[glass junco]

use appilio and choose 40k bit rate and the one under 40k

[lean tendon]

I am already past the preprocess step, my audio was 48k so I had to use that one, and I didn't need any post process I am at the feature extraction now

[glass junco]

just copy the path of the new d n g folder after you do the index

then train

[lean tendon]

i see, very nice

I am almost there, I want to learn everything to fine tune my awesome model, I can't wait to flick on the train button, I have the most awesome training dataset

[glass junco]

like this i mean