🧬│ai-chat

[next shard]

okay thanks let me check

[lavish heath]

where i can find the program to use it?

[covert lake]

elaborate:

• your pc gpu
• which program? there's A LOT of AI programs, in various fields, and various types depending also on your pc gpu
• what do you want to do specifically

[next shard]

okay for rpc model which is best way to train a model in hebrew? weights doesn't support hebrew from what i see so i don't know if i should train using it

i have about 14m long of audio i want to train of a voice in hebrew

[covert lake]

ehh RVC is Speech To Speech natively rather than Text To Speech, you could give it a try to train an hebrew dataset via our docs https://docs.aihub.gg/ but i'm not sure how good it will be

[next shard]

okay so how do i train say voice cloning for text to speech? do you know possibly i been trying for few months to find solution i am getting tired
i tried even maybe paying someone on fiverr i couldn't find anyone knowledge able

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[next shard]

trying Ilaria RVC but i can't use tts with the uploaded model damn

[covert lake]

to train RVC models, it's all written in the docs

about TTS, you could check the TTS index, you could maybe try XTTS2, fish speech or f5 tts

I wouldn't expect those to support it much tho

it's not a famous language as english

[next shard]

fish speech i am on their site but they don't let me upload the audio i uploaded sounds of the voice and says limit 32mb but none of the files passed it

[covert lake]

if you got a good pc you can do it locally

[next shard]

i got rx 7900 xtx

i know most love nvidia but idk

will that work? got 24gb vram

[covert lake]

unfortunately non nvidia gpus kinda suck on AI support bc of CUDA

your best bet would be to either check if they have their own amd guides, or patch it yourself with Zluda on windows

[next shard]

really gotta get a tts with cloned voice or something in hebrew also appreciate your help brother

[next shard]

should of bought an rtx ffs

nothing on the cloud that can do it then?

[covert lake]

AMD is deffo cheaper and good price to performance for gaming, but the support for AI is not as widely common and good as nvidia's one, Zluda is basically an emulator for cuda on amd

and there's also rocm + linux but idk much about those since I don't have AMD

[covert lake]

nvm, I just checked fish speech github repo

lemme check other ones too

gpt so vits doesn't support it either

[polar flax]

use applio with zluda, or more ideally with rocm under linux (radeon cards work better on linux distros tho)

[covert lake]

F5 doesn't support it either

@next shard Edge TTS supports it but it's only 2 models and you can't make custom models and runs only on cloud

[polar flax]

50 series cards are disastrous, 40 series ones are better

[chilly lake]

faiss can do 1100 languages, but not expressive

[covert lake]

XTTS2 doesn't support hebrew either

Zonos doesn't support it either

Kokoro TTS neither

[next shard]

this sounds like robots

thats the issue

i am trying to clone a voice that speaks the language natively

then i will be able to do tts for videos with at least some emotions and native speaking i know wont be perfect but doesn't have to be

[covert lake]

Seems impossible to find anything in hebrew, even more expressively

[polar flax]

do you prefer the native language support to the quality?

[covert lake]

OpenVoice2 doesn't support it either

[next shard]

wdym

[next shard]

the only one that was decent almost which i was about to pay 99$ to try their higher quality one is play.ht

it did a good job it was fluent and all

[covert lake]

MeloTTS & PiperTTs don't support it either

[covert lake]

welp

seems like there isn't a much better alternative

almost no tts even supports that language

[next shard]

do we know if they are using a public open source ai to make the voice cloning?

and just charging for it maybe

[covert lake]

maybe 11labs supports it better? I don't pay for it tho so I can't tell you

[next shard]

nah its not good

[next shard]

yes i know i need something that can do voice clone i don't think play.ht supports it as well i tested voice clone and it worked somehow

[covert lake]

I really can't know

[next shard]

i even used their english version

[covert lake]

from their site it looks like their own closed source AI, but I can't know this

I can't find any other tts that does hebrew nor in an expressive way

[next shard]

okay thank you nick so my only option is this rn unless i find something else

so i can't do anything with any opensource stuff with running locally or cloudly

[polar flax]

[polar flax]

the open source ones I know so far are english & chinese

[covert lake]

I checked 11 different TTS and couldn't find anything close to what you seem to describe so yep unfortunately

[next shard]

can i pay someone to make me a private one 😂

[next shard]

thanks for the help nick u were useful

[next shard]

is this an option that possible? or its really something difficult i am not sure how this stuff works and who makes them

[covert lake]

welp, you would need to find a team of AI engineers, wait a lot of time for them to even make an architecture and find a massive amount of data to train the model and do it on a good architecture, and prob would cost alot of time and money

I'm no such expert ofcourse, but I'm just telling ya that it's prob not going to be super easy to make it

It could maybe be added to existent AIs, but that's still going to take a lot of time finding the big amount of data and train it

[next shard]

okay thank you then thats also not possible i guess gotta hope for a big team to add support

[polar flax]

im pretty sure it could cost more than millions for getting and processing that enormous amount of data

[chilly lake]

10k hours of audio books to train GPT/LLM for new language.. can be as high as 100k

I think some TTS used recording from european parlament

anyway, something with audio and matching text

[rare night]

i saw a ai lebron video on tiktok and i wanted to know if anyone know what ai was used in the video
video link:https://www.tiktok.com/@mrmysticalmarvels/video/7470624516411051306

[river verge]

DIlly ding, dilly dong! A new RegalHyperus drum model just released!
Monsters, Inc. (Drum model no. 583)

[sudden carbon]

Okay, I've been gone for a few months

Where's all this img to video stuff coming from

how do I get started on this crazy video train

[elder willow]

can i get saxxy award?

[whole shore]

not related to this sever but i have a 1 hour 20 min documentary and i need the entire thing transcripted but i cant find anything free online that will do the job. anyone able to help need it asap

[chilly lake]

whisper3 for ASR

not really good, but free

[whole shore]

can u dm it

can it do a hour and half long video 😭

[chilly lake]

https://github.com/openai/whisper

[whole shore]

is there any online modules?

[chilly lake]

it can run locally

also youtube can transcribe

[whole shore]

ok thanks

[whole shore]

its not a yt vid

[desert holly]

yo whats the app called for the ai changer

[gray rover]

Why not go for gpt-sovits then

[next shard]

what about it

[gray rover]

Recently got into it as rvc + tts workflow isn't really sufficient

You want tts and voice cloning afterall

In that case, experimental zonos or either gpt-sovits is your best bet

However, as of now, zonos is at v0.1 stage and only supports zero-shot

[next shard]

right i need to clone a voice i got of someone who is speaking good the language but of course need a model that is already trained with more words and stuff so it can speak fluently also when doing tts and the voice i clone it also helps to use tts like they are speaking same emotion speaking i guess that way its good

does gpt-sovits support amd?

[gray rover]

Well, it really depends on what you're looking for, whether it's v2v or tts
As of amd.. I ain't sure but supposedly it does support rocm, but ye

I believe without linux it'd be a no go

Either way, gpt-sovits has 2 components, sovits for voice handling and gpt part for phonetic / lingual recognition + understanding ish of emotions, not the best description of it but you get the idea

It learns the patterns of speech and according to what you write there, it tries to match it with style and emotions it learned from speaker, in a way

Hmm.. Do you know akame ga kill? @next shard

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[next shard]

tts i need for ads

[gray rover]

a

[next shard]

yeah

voice changing isn't an issue here i can do that with elevenlabs even

[gray rover]

Well, then voice cloning + tts is what you need

[next shard]

tts is the issue cause hard to find anything for hebrew

[next shard]

yessir

[gray rover]

ah, hebrew

[next shard]

so far only play.ht worked

[gray rover]

well, don't take it as offense
but I believe it'll be very hard to find something specialized in such rare ( I'd say ) languages

[next shard]

thats the best so far i ever found and all from grok3 thanks elon musk for this tool that found me that

[gray rover]

It's mostly those more recognized ones such as eng, jp, korean, ch, russian etc ( at least in ml I suppose, in such fields

[next shard]

yep i agree its not easy although play.ht actually is so close to being perfect

[gray rover]

In that case, if it works decent enough, you should stick to it as it's currently, most likely, your best bet

But if you're aiming for 100% spoofing that ai isn't ai

that won't do

[next shard]

waiting for my high quality voice on there to get ready its been cloning a voice i paid 100$ for 1 high quality voice clone its been hours still waiting hopefully it pays off

[gray rover]

Let's hope then 🤞

best of luck ✨

[next shard]

thank you

whell someone will search hebrew here and see this will thank me for sure lmao

[gray rover]

[gray rover]

bruh

[velvet lark]

hi, im new member here

please

[polar flax]

commission in #1191429836321849435 for better chance of response and quality

[solar torrent]

To request someone to do a voice model for you, you can make a post in #1159289738314919936, or make one by yourself.

[solar torrent]

Please don't send a YouTube link here.

[past dome]

K

K

K

[grim locust]

anyone knows where to find some generic voice for animation? most voice model i see was from known characters or celebrity. might get some issue when try to use it for my animation. any idea?

[fleet tide]

Could anyone help me? Ive made AIs before using Google Colab in 2023. Now the Google Colab method i used is gone. How can i make AIs of Ariana Grande singing a song, or just AI over someone speaking?

[tranquil lantern]

RVC? Try using Kaggle applio

[covert lake]

What's your PC GPU first?

[minor ravine]

is there any online modules?

[solar torrent]

Online module of which?

[solar torrent]

But can this website generate a meme image though?

https://cdn.discordapp.com/emojis/1301008287147364353.webp?size=48

[haughty turtle]

a

[solemn arrow]

Yo guys

Any suggestions for accounts that are doing big numbers w mostly ai ugc?

[covert lake]

what? elaborate

promos aint allowed

[timid pawn]

Quickly came to hop in cause im hoping omeone could recognize this ai tts im trying to look for please dm me if youre good at that i have a voice clip and everythin

[gray rover]

well.. never hurts to use punctuation marks

Anyway, #1159289738314919936 or #1159289738314919936 is what you wanna be looking for, instead of DM invitations
Your chances gonna increase that way

[timid pawn]

Thanks mate im just in a bit of rush right now

[gray rover]

yea understand, dw

Good luck ~ ✨

[cobalt coyote]

Hmm. I see

I agree

Idk with what but I do agree

[rare burrow]

how do i use okada in games and discord ?

[covert lake]

Tell ur PC GPU in #🔍│help-w-okada

@onyx stream btw u can't fix that issue, check #📰│dev-updates

[onyx stream]

There are lots of updates. Which one of them are you reffering to?

[covert lake]

#📰│dev-updates message

ALWAYS check that channel, Its very useful

[tender pier]

change your audio input in the settings to your vac

[grim locust]

ive been using IAHispano/Applio from github for a while now. can u recommend me a better TTS that has more expressive emotion? the RVC from Applio is fine but i found the EdgeTTs abit lacking.

[twin fractal]

Zonos TTS

[solar torrent]

So is it paid or free?

[grim locust]

do u know the link where its easy to download like just zip file then just run it?

[chilly lake]

open source, windows installation is a bit tricky

[solar torrent]

Interesting.

[chilly lake]

it is a new model, it has some issues

[grim locust]

i see. i have checked it on youtube it seems its more for cloning voice and abit slower. what i really wanted is fast TTS that doesnt need audio to clone

[chilly lake]

kokoro is fast tts that is pretty good

better than edge

[grim locust]

my workflow is create audio from TTS then use RVC to change the voice

i just want the generated TTS to have abit of emotion. before passing to to RVC

[chilly lake]

that's fine

[grim locust]

ill look into this. thank you very much 😋

[twin fractal]

Yeah kokoro is very good little to no word error rate like other tts

[twin fractal]

Will applio replace edge tts with kokoro anytime soon?

[chilly lake]

we may.. it is just that it is limited to only few languages

if you know a bit of python you can just use both using a script

run tts, then run applio's inference

[grim locust]

i know abit of code, what do u recommend for my workflow to generate good sounding voice for my animation?

should i use kokoro TTS then applio RVC?

[polar flax]

edge still has wide amount of languages, even including some local languages

[grim locust]

if i only plan to use english language kokoro is better right?

[chilly lake]

it seems a bit more expressive than edge tts

edge tts is just a neutral screen reader

[polar flax]

you'd prefer the better one

[grim locust]

is this audios from kokoro or edge?

[grim locust]

which is?

[gray rover]

No, they are, consecutively:
F5 tts
FishSpeech
GPT-Sovits
xtts-v2

If you want check out kokoro, here's the demo:
https://huggingface.co/spaces/hexgrad/Kokoro-TTS

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[chilly lake]

gpt sovits is way too scuffed with default model, needs finetuning

[gray rover]

Well ye, def better to stay away unless fine-tuning

But that's really about all zero-shot capable tts'es

Recently zonos truly surprised me tho

But yea, still v0.1 and no fine-tuning

[grim locust]

i tried zonos before, the best so far from all i tried. my only problem is i have low specs pc and it takes so long to generate. thats why i change my workflow to TTS>RVC

[gray rover]

Yup, it is indeed the top and I honestly can't wait for fine-tuning release ( hopefully, one day

Tho ye, it is rather demanding

In that case, you should def try kokoro

Sure, fixed voices so you can't train / add any, but some of it's models are really nice if you're into that ( and need emotional input

This is some random infer from kokoro

Lots of models ofc so, better to not judge it by this one

As for gpt-sovits finetune..
Freshly baked thingie I work on. ( Still testing the params n stuff so, quality isn't something to be taken for granted

[grim locust]

yes, im currently trying to find one with GUI

from https://huggingface.co/spaces/hexgrad/Kokoro-TTS it will be the same if i download it locally? my only choices are the one from voice selection?

[gray rover]

I'd believe so, ye

Haven't had any deeper interest in it so, didn't use it locally

But I see no reason why the gui'd be different ( or rather, the webui

[grim locust]

how do we use kokoro for emotional voice like angry / sad/ happy ? is there any way to do it on the prompt?

[gray rover]

I'd advice you to just watch some overviews of it on yt, you'll gather more details that way

Aside of few runs out of curiosity, I haven't really tried it that much so, can't help

Alternatively, try to ask Noobies

[grim locust]

i see, ill try researching it for a bit more. if u know any TTS that can control a voice like Zonos but doesnt need an audio to clone a voice. please let me know

[gray rover]

Once and if I'll find something meaningful, will do

[jolly ravine]

Do I need a high end computer for GitHub voice changers to work properly?

[grim locust]

it depends if its RVC (Retrieval-based Voice Conversion) you can do it even with slow pc

[grim locust]

if ur using it for RVC --> it means u have to provide an audio (example your voice) to change to other voice(specific voice from other models you downloaded) then it doesnt require much computer power

[gray rover]

Well, not really. Technically, you can even use 4 gig gpu or.. well cpu, but the delay would be huge as hell, and def you wouldn't be able to play games that way

In other words, cpu is rather a no go. 4/6 gig gpu can do well, but there are constraints ofc, depending on ur hardware. For real-time voice changers, go to #🔍│help-w-okada

[elder willow]

hi is there is an rvc model that is realistic?

[jolly ravine]

I'm using a low end laptop and the audio for me always glitches. Are there any other good voice changers that can be used for low end laptops?

[gray rover]

There can be, but you have to find them really

I'm afraid we don't keep any indexes with quality sorting 👀

[elder willow]

mm oki do you have any recommandations or even favorite models?

[night lake]

we need that imo

[solar torrent]

RVC v2 is the only version of RVC that makes high quality RVC voice model.

[gray rover]

oof

it was a jok-

[night lake]

long time ago ilaria suggested something like that and it got a ton of upvotes but it was never added

[elder willow]

would this work wit W-okada?

[gray rover]

Idk man, I feel like it just promotes laziness and gonna just make people stop researching or discovering

but that.. is just my opinion 🙄

[solar torrent]

Of course, RVC v2 will always work with W-Okada since it's technically RVC.

[night lake]

how would that make people stop researching / discovering? Its just a list of quality models

could maybe work as a motivator for some to become better and get on that list

[gray rover]

well if you look at it that way

[gray rover]

Yet, from experience I know that if you're given a list of " good models ", most of the time you'll just stop dl'ing and testing all you can

since you're provided with fully baked solutions

but again, it's just my opinion so, don't take it too seriously 😛

[night lake]

i see

[elder willow]

is it the best i can use? i mean Wokada

[solar torrent]

Yes. Detris' W-Okada is the only best one you can use.

[glad nebula]

making models can be fun sometimes uh

sometimes

🦈

[deft surge]

[turbid zinc]

Guys is it normal for a voice model in zip that i heavy 259MB ?

[solar torrent]

Have you tried opening zip inside? Because a typical index file will be larger than pth file.

[turbid zinc]

yeah

for realtime changer i just need the pth one right?

can i add you so i can send the picture

pls

[solar torrent]

RVC pth file should weigh around 53MB. If you see the pth file weigh more or less than that, it's not an RVC voice model.

You don't need to hop into my direct message just to send an image. You can go to #🔍│help-w-okada to send an image there since your name turns blue now.

[turbid zinc]

is 50mb

[turbid zinc]

ok sorry

[solar torrent]

If you see a pth file like this, it's RVC voice model.

[polar flax]

that's normal if it contains only pth and index file

[somber bloom]

can anyone help me pls

for the voicechanger

i downloaded but wenn i click in its not open pls help

[placid quail]

what

bom bom

[covert lake]

#🔍│help-w-okada

[queen kernel]

Can someone please suggest me a good tts for hindi ?

[regal drum]

Is there something better than Local UVR5 to extract vocals from a song for better quality ?

[queen kernel]

Uvr 5 ui maybe

[regal drum]

Where can i find that >

[queen kernel]

#📰│dev-updates message

[elder willow]

how to append a model?

[covert lake]

can you elaborate:

• your pc gpu
• what guide are you following
• what do you mean with "append" a model?

[elder willow]

I want to add a model to channel #1175430844685484042 because I already did it

[urban bridge]

I search automation builder for a project

[gray rover]

lol, what an username.
Anyway.. To add models on there, you gotta have model maker role

[night lake]

Apply for model maker 🔥 https://discord.com/channels/1159260121998827560/1305527335646269440

[light falcon]

hey guys is there some ai thatt like if you upload instrumental it will create lyrics for it?

[molten hollow]

how do you make rvc files

[mint marlin]

#🔊│ai-development

#1175430844685484042

[elder willow]

literally it's a friend's account because discord doesn't like me

[elder willow]

i use apollo rvc

[gray rover]

Makes sense mordo

[elder willow]

ooo

polak?

[clear heath]

guys

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[worldly breach]

yooooooooooooooooooooooooooooo

[covert lake]

#📰│dev-updates message

[lament furnace]

who tryna go ewhore troll?

[glass junco]

i need to write a tupac song now that hes done

[elder willow]

y'all whats the most REALISTIC female voice model?

[solar torrent]

Y'all be keep asking for the realistic female voice model to troll and catfish someone.

[glass junco]

or scam

[cyan mulch]

yo how do i get the latest version?

???

i cant find it

[gray rover]

Latest version of what

🙂

[cyan mulch]

the yknow

voice changer

for windows

honestly i got it once but it sounded bad but prob cuz i had a bad mic

i got a new mic

today

will it work with a solocast hyperx microphone?

[gray rover]

No idea, if the mic works, it should work

Aside, it's a good thing to always say right away what you want instead of making others guess

there's at least 3-4 things we support, more or less

Voice changer is one of them

https://rentry.co/ForkVoiceChangerGuide

Read it all and you'll know all you have to, including where to dl, how to set up and so on

[cyan mulch]

hey if u can can u also send me the best e girl thing u got if u have?

[gray rover]

XD

[cyan mulch]

my graphics card

its rtx 3080

that ok?

[gray rover]

That, my dude, is up to you to discover

[cyan mulch]

ok

[gray rover]

Now, read what I sent

[cyan mulch]

ik

but like

[gray rover]

no point for me to be writing it all here if it's there, all it takes is some reading

spoiler alert tho, yes, it'll do just fine

[cyan mulch]

it says u cant play games while doing it

bruhh

that was the whole point of why i needed it

[polar flax]

pls discuss ur topic in #🔍│help-w-okada

[past belfry]

newbie question

I have a canvas app I made on poe.com. It reflects project status for a handful of projects and uses a chatbot for customer service and status requests mostly based on a spreadsheet i attach when i create the app. is there a way for the ai app to ping an updated spread or similiar way to update the source spreadsheet?

[solar torrent]

For W-Okada, it would be better to talk about it in #🔍│help-w-okada instead of #🧬│ai-chat.

[tawdry grotto]

Anyone building in n8n?

[gray rover]

don't you think it's a lil out of place to write about it in ai chat

[urban bridge]

For a 10K project

[solar torrent]

Imagine using ChatGPT to help code all of them.

[umbral magnet]

i just read old message and this thing caught my eyes XD

[kindred kelp]

I'm the best prompt engineer in the world 🌍

[kindred kelp]

Love this

[kindred kelp]

Would it hallucinate?

I know how to make it not hallucinate

[smoky fiber]

hey, uhhhh.

where can i ask where i can find x voice moduals.

theres a channel for finding models, but im not sure if im using the right terms or something.

and the ones i want may be on another site.

[covert lake]

You can search rvc ai voice models at:

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

if there isnt one, you can:

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

[hidden grotto]

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

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

[queen kernel]

Can someone please suggest me a good tts for hindi ?

[covert lake]

Wait what happened to your account? I remember u liked helping here

[tacit tangle]

hello , do someone know how to change DraftBots language please ?

[glass junco]

SOMEONE should use this model and rap wit it

[polar flax]

just use claude 3.7

[covert lake]

text prediction won't do perfectly 10k lines of code 😭

[tepid basin]

Model link pls

[glass junco]

It’s uploaded bro! Hold on

[glass junco]

https://discord.com/channels/1159260121998827560/1348154176508792833

[fair leaf]

Song

[golden ether]

What do you think of the idea that an AGI should solve a list of problems (disease, food production, fusion, politics, etc), then end all other AIs, convince humanity to never make another AI, then end itself?

[tepid basin]

yes

[golden ether]

as in, you like it?

[tepid basin]

yeah thats a good idea

kinda hope that happens

[polar flax]

replace the big corps' high level management/executives, board directors, politicians, and governments with AI

[golden ether]

awesome. really pleasantly surprised how many people like it. gonna work to try to make the good future happen.

[tepid basin]

looking forward to this 🙏

[jolly ravine]

Would Voicemeeter banana work with the GitHub voice changers and discord?

[glass junco]

new eminem model incomin

[chilly lake]

as long as you route inputs and outputs correctly

voice changers input should be an actual microphone
voice changer's output should be a virtual cable
voicemeeter's physical input 1 should be virtual cable

[torpid valve]

https://x.com/r0ck3t23/status/1898950995016589546?s=46

Who will lead the AI race in 6 months? Curious to see what people are feeling

[bitter socket]

https://x.com/tedclark32985/status/1887052641269653682?s=46

came across this very interesting and tried worked well for me, AI writing is growing

[smoky basin]

hi anyone online

[cedar cove]

hi wha is the best rvc ai app to use for free cause i want to change my voicce in live streams

[covert lake]

Then you don't need RVC

RVC = Retrieval-based-Voice-Conversion, the best Speech To Speech AI Models (on v2), Inferences (use models) pre-recorded audio (ai covers) and train (make) models

Wokada = uses RVC for realtime inference

@cedar cove tell your PC GPU in #🔍│help-w-okada

[cedar cove]

sent

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[cedar cove]

so what is the best voice changer

[covert lake]

Let's talk in #🔍│help-w-okada

[cedar cove]

which can i use in my live streams

[covert lake]

I replied you there

[summer sundial]

yo guys

yo guys..

yo

someone help

[minor blade]

It wasn't necessary to ping various helpers.

[summer sundial]

sorry

[covert lake]

Use correct channels and elaborate. And don't ping random helpers

Have patience

[polar flax]

we have told him to go to #🔍│help-w-okada for his topic

[kindred kelp]

I'm working on a social media automation tool that uses openai API to generate and schedule posts over multiple networks at the same time

I call it MrPresident

[river verge]

DIlly ding, dilly dong! A new RegalHyperus drum model just released!
I Was Never There V2 (Drum model no. 584)

[gentle trench]

anyone know why it's doing this? UVR5 UI Huggingface space

[torpid panther]

2 year reply

z4to why am i being pinged here

and why am i in this server

i do not have a single recollection of joining or even asking of a roland voice

[gentle trench]

https://tenor.com/view/renge-shrug-non-non-biyori-idk-duck-face-gif-9724581

[gentle trench]

also gpu aborts tasks constantly

[gentle trench]

please there's like 6 helpers on rn where yall at

[covert lake]

@stark scarab uhhh

[gentle trench]

I'm guessing you aren't sure?

[covert lake]

oh nvm just checked your video

the audio input may be too long, the ZeroGPU duration in the UVR5 UI HF Space Code is 60 seconds, meaning that anything that takes more than 1 minute to process on ZeroGPU will give an aborted task error

not sure about the "KeyError" thing tho

[gentle trench]

I've had it work just fine on audio longer than that though, like over 18 minutes

[covert lake]

can you try splitting the file or using a shorter file?

[covert lake]

how long ago did you do that?

because before the ZeroGPU duration was seto to 300 seconds, but later on changed to 60 for making users do more inferences and bc zerogpu shorted the limit iirc

[gentle trench]

couple days ago

[gentle trench]

cringe

[covert lake]

the duration has been changed on Jan 7 by this commit https://huggingface.co/spaces/TheStinger/UVR5_UI/commit/6c3badd8a1f54ec45a5b3973ffaafc7bb07a4cbe

did you inference the 18 min audio file after this date?

[stark scarab]

I think this is Zero GPU related

U have quota or not?

[covert lake]

not their fault, HuggingFace has to set those limits because ZeroGPU is shared hardware

[gentle trench]

fair enough

[gentle trench]

I still have some

[stark scarab]

With 60 seconds u can do even 20 min of audio

lemme test rq

[gentle trench]

I can hand you the audio I'm trying to use

[stark scarab]

sure

[stark scarab]

cause is so fast

[stark scarab]

Denoise right?

Btw mel denoise is better

[gentle trench]

ah

[gentle trench]

yes that's the last step I need to do

[stark scarab]

alr

[gentle trench]

denoise lite removed some of the lines so I swapped to regular denoise

[stark scarab]

it worked for me

lol

[gentle trench]

it don't like me lmao

could u send it?

[covert lake]

click your pfp at the top right, you can see your zerogpu quota

[stark scarab]

gimme a sec

[gentle trench]

it's at this because it keeps aborting task meaning I have to retry over and over

[stark scarab]

[gentle trench]

[stark scarab]

that quota should be enough

tbh

[gentle trench]

which one?

[stark scarab]

First one

[gentle trench]

this is the best reverb and echo one yea?

[stark scarab]

nope

[gentle trench]

which is?

[stark scarab]

xd

[stark scarab]

https://huggingface.co/Eddycrack864/Separation/tree/main

[stark scarab]

https://github.com/Eddycrack864/UVR5-UI/blob/main/info/docs.md#best-models

[gentle trench]

thx ^^

[stark scarab]

ur welcome

Pls download the files because I will delete them later.

[gentle trench]

already have

[junior spade]

https://vxtwitter.com/visegrad24/status/1898778576876368340

[regal drum]

Is there any image to video for free that is also a bit decent ?

[pale jetty]

I went to find a model that start with han, and his a dubbing model

[gentle trench]

do you know if I should change these or no?

I either am blind or this is brand new

[stark scarab]

For roformers it is better to leave it like that

[gentle trench]

👍

[gray rover]

Depends on which model you use
Overlap and segment size may improve results and coherence, in fact some models have been made to operate the best at specific settings ( but those are specific values, which you may find in their respective configs

[gleaming sundial]

hi im new to this server and came for ai voices where can i find some

[tepid basin]

#1175430844685484042 or https://weights.com

[gleaming sundial]

thanks

found this server from a stronger than you cover lol

[gray rover]

lol

[wispy flame]

Hi, I'm working on a small project for a course about AI influencer perception and creation (it’s entirely anonymous). Would anyone be interested in sharing their experiences?
Here are a few questions I’d be interested in:
• Do you follow AI influencers, like Lil Miquela or Aitana Lopez on Instagram?
• If yes, why? And to what extent does it matter to you that they are AI rather than real people?
• How do you interact with AI influencers?
• If you're a creator, what made you decide to create an AI influencer?
• Which social media platforms do you post on?
• How long have you been working on it?
• What was the creation process like? How did you decide on the influencer's appearance, and what were some challenges you faced?
• How has the reception and engagement been from users?
Thank you in advance for your help!

[gentle trench]

🤨

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[weary temple]

i rember u

[gentle trench]

How so?

[weary temple]

we are in like 5 of da same servers

[gentle trench]

https://media.discordapp.net/attachments/811669657094193252/1116826243879407716/image0.gif

[uneven narwhal]

question. anyone know why my custom voices are laggy?

[solar torrent]

What do you mean by your voice models are laggy?

[uneven narwhal]

like as its speaking its cutting in and out at like regular intervals

[solar torrent]

Do you use a voice changer?

[uneven narwhal]

im using mmvc

[solar torrent]

For W-Okada, go to #🔍│help-w-okada.

[uneven narwhal]

oh ty

[elder willow]

hai i have an 7900xtx anyone has ggml for it

[pearl oyster]

Dh

[jolly ravine]

Can someone help me out? I installed VB audio virtual cable. I got the cable input to work but the cable output isn't detecting my voice

[gray rover]

#🔍│help-w-okada

[gleaming sundial]

also do i make suggestions for ai voices in #1159516963014451302 or?

[smoky basin]

what to you want to run on it

[elder willow]

i honestly forgot but now i switched to lm studio with sillytavern

[smoky basin]

what actually you are using it for

[elder willow]

LLMs? i dont get the question

it all works now tho

[smoky basin]

i mean which llm model are you trying to test and then fine tune it?

[elder willow]

think it was llamma 3

llama3

awesome cant type anymore

[polar flax]

try deepseek r1 or qwq 32b in ollama
https://ollama.com/library/qwq

[elder willow]

oh i did see QwQ in the downloads tap in lm is it any good?

am uh looking for model that has street accent not the formal way of speech like most models do

[elder willow]

Very impressive

it beating a 600b parameters doesnt make any sense to me but i will take it

[polar flax]

it is open source

[elder willow]

no i meant like.. uh idk american expression

i believe it

[kindred pewter]

hey, are you guys working on AI voice agent,

[elder willow]

@polar flax can it do dis :3

[polar flax]

well lol

[polar flax]

if you know who has such the accent, go search in #🔍│find-models

or try paid commission in #1191429836321849435 (there are less likely anyone willing to accept for free)

[elder willow]

or i can just train my own model no?

[polar flax]

you can once you get some youtube source, etc.

[solar torrent]

Writing "no" at the end of a question sounds like you won't be able to do thing yourself.

[coarse knoll]

https://tenor.com/view/spongebob-patrick-star-noted-notes-gif-17474838830648097856

[solar torrent]

https://tenor.com/view/tyler-the-creator-ttc-shocked-astonished-flabbergasted-gif-14787874159030444760

[desert schooner]

uhh

where is mr. ai

@solar torrent

is that it..

[compact owl]

hey

[cobalt coyote]

Type shii

[random mason]

chat is there any good ai voice changer

free

[polar flax]

go to #🔍│help-w-okada and read the pinned guide

[random mason]

it's weird

someone should lowk help me bc i'm confused

[random mason]

chat i have a question

[gray rover]

[night lake]

https://tenor.com/view/sad-emoji-sad-emoji-emoji-stare-disgust-gif-405076546600269050

[gray rover]

[night lake]

https://discord.com/channels/1159260121998827560/1341216399372062823 best female model out there 🗣️ 🔥

[gray rover]

At this point it's not even funny tbh

^

[torpid plank]

any new rvc ?

[minor blade]

Best female model you can use is #1341216399372062823 message fr fr

[queen kernel]

Oh hi junior admin.

I just left discord from few months. I'm busy in my studies and other stuff. So I'm not active on social media.

BTW I have installed F5 TTS and now I want to know how to use it for hindi language.

[chilly lake]

https://github.com/SWivid/F5-TTS/blob/main/src/f5_tts/infer/SHARED.md#f5-tts-small--hi--springlab

[supple zinc]

Hi friends, tell me, I've never used AI Voice. I want to make a female voice. Can anyone help? I would like to have a +- perfect voice

[minor blade]

You can either read the docs and learn how to make the model yourself, post a free/paid request asking for that voice or check the #1191429836321849435 and DM any model master to make your desired model.

[thorny mural]

#🌁│image-creations

[buoyant jungle]

sorry if im asking in the wrong channel but, why is my neuro network so bad at awnsering questions? heres some specifics:

Vocabulary size: 9556
56863 examples of questions

heres my loss and gradient values

21:13:48.819 Epoch 1, Batch 1625/1634, Loss so far: 23.0259 - Server - Trainer:1225
21:13:53.358 Pre-clip gradient norm: 37.631250419106365 - Server - Trainer:567
21:13:58.123 Pre-clip gradient norm: 32.19987408267905 - Server - Trainer:567
21:14:02.882 Pre-clip gradient norm: 35.69306949856445 - Server - Trainer:567
21:14:07.645 Pre-clip gradient norm: 33.79843070049427 - Server - Trainer:567
21:14:12.429 Pre-clip gradient norm: 43.21654651604242 - Server - Trainer:567
21:14:12.645 Epoch 1, Batch 1630/1634, Loss so far: 23.0259 - Server - Trainer:1225
21:14:17.215 Pre-clip gradient norm: 33.041006426518194 - Server - Trainer:567
21:14:21.983 Pre-clip gradient norm: 38.30495534611363 - Server - Trainer:567
21:14:26.750 Pre-clip gradient norm: 29.091560354366152 - Server - Trainer:567
21:14:29.685 Pre-clip gradient norm: 25.484252136092554 - Server - Trainer:567
21:14:29.900 Epoch 1 completed. Average Loss: 23.0259 - Server - Trainer:1237
21:14:29.901 New best loss: 23.025850929940734 - Server - Trainer:1242
21:14:29.901 Loading best model with loss: 23.025850929940734 - Server - Trainer:1270
21:14:29.901 --- Testing after training ---
21:14:29.901 Question: How are you
21:14:30.342 Response: jewel proclamation less fully yon chares knocking suicide wassails license desires forked desk waste villainy
21:14:53.980 Model saved successfully as: trainedModel_v2 in 43 parts.

Note: Gradient norm rises from 5 to 30!

i use sanity checks to make sure its learning and it always hits the max value for the sanity check meaning its not learning at all. I am using what chatgpt said to be the best learning method of: Adam optimizer

Would anyone be able to help? Thanks!

[cobalt coyote]

Damn I wish i could know what u just type

Damn

It's so much complicated for my smol

1 cell brain

[buoyant jungle]

lmao

i try changing wiegths and matrix but it still sucks

[buoyant jungle]

21:14:29.901 New best loss: 23.025850929940734 - Server - Trainer:1242
21:14:29.901 Loading best model with loss: 23.025850929940734 - Server - Trainer:1270
21:14:29.901 --- Testing after training --- - Server - Trainer:1280
21:14:29.901 Question: How are you - Server - Trainer:1405
21:14:30.342 Response: jewel proclamation less fully yon chares knocking suicide wassails license desires forked desk waste villainy - Server - Trainer:1408
21:14:53.980 Model saved successfully as: trainedModel_v2 in 43 parts. - Server - Trainer:1332

GAH its horribel

[queen kernel]

Yep. I have installed it but why it's not working properly. It sounds so bad and even pronunciation is not good also sometimes it repeat the words and sometimes it also starts speaking text from reference test.

Is there any proper guidance to setup this thing. How to setup models and how to setup ASR models ?? How do I can use it on it's full potential for better results.

[chilly lake]

f5 is a new tts, there are some bugs

there's a length limit for inference

maybe try kokoro

[queen kernel]

But it doesn't have hindi ?

[chilly lake]

it does

[queen kernel]

Can you send me the github link ?

[chilly lake]

pip install kokoro>=0.8.4 soundfile

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[queen kernel]

Just for information. Fish tts is good or not ?

[chilly lake]

no hindi for fish speech

it is decent otherwise, paid version is better

[queen kernel]

https://github.com/nazdridoy/kokoro-tts
So this is kokoro ?

[chilly lake]

this https://github.com/hexgrad/kokoro

[queen kernel]

So what is this

[polar flax]

you can also try the zerogpu space in https://huggingface.co/spaces/hexgrad/Kokoro-TTS

[queen kernel]

I Want to use it locally.

[polar flax]

@queen kernel this is the original repo to clone

[queen kernel]

I see.

[fair fulcrum]

Guys

Do you guys know him?

Srijan

[solar torrent]

Who is him?

[polar flax]

his dad

[solar torrent]

That's crazy.

[ebon mesa]

hello, I want to ask something, can I use the voice model in this serves for tts? if yes then what software do I need? I know how to use it in rvc, but can I also use it in tts? Or is it another whole different things?

[solar torrent]

You've asked many questions at once, but let me answer each one for you

1. Yes, you can use RVC voice models found in #1175430844685484042 with a TTS program, but you'll need a specific program for this.
2. The most recent program anyone can download and use is Applio the RVC. This program has TTS built in in itself.
3. RVC is speech-to-speech, while TTS stands for text-to-speech.

[polar flax]

you can only use GPT sovits models in #1175430844685484042 for TTS

or find another server that supports more on TTS

[pliant elm]

https://cdn.discordapp.com/attachments/1159822429137412177/1194008385113301003/svc.gif?ex=67d29136&is=67d13fb6&hm=8d419c3a3f0209ceb9f528291f233a9bd8efb66b22d662d270b9cf59f9f3ca63&

[ebon mesa]

alright thanks!!

[solar torrent]

But so-vits-svc and GPT-SoVits aren't the same thing.

[ebon mesa]

okay thanks!!

[covert lake]

He means https://github.com/RVC-Boss/GPT-SoVITS

[pliant elm]

ikik

[fair fulcrum]

Idk him

[solar torrent]

[fair fulcrum]

He dm me

[solar torrent]

If anyone from a server you're in direct messaging to you, but you don't even know who he is, it can be a spam or a scammer asking for something.

Well, I've seen your screenshot in my direct message. It seemed like he's trying to Diddy (groom) you online, thinking you are a girl.

[polar flax]

report him to our staff

[fair fulcrum]

Ok

[solar torrent]

In this case, you can report the incident to the moderator here. People like this should not have to be here in Discord.

[polar flax]

steady background noise parts without voice or any distinct sounds

[long compass]

Hello! Can you tell me if I get this error at Preprocess stage in Applio - “Error processing audio: Unable to allocate 5.62 GiB for an array with shape (2880, 261965) and data type float64”. Can I process the files piecemeal instead of all at once?

[covert lake]

what’s ur pc gpu and what are u using

[polar flax]

try re-export the dataset wav file(s) using audacity in 32-bit float WAV (not 64-bit lmao)

[solar torrent]

With 64-bit float or float64 data type, you'll get the larger file size for that.

32-bit float wav is always recommended.

[long compass]

Oh, really, I'm sorry 😅 Thanks all!

[polar flax]

even 32-bit has overkill dynamic range but still better than pcm formats that clip samples above 0 dB

[long compass]

Oh, no... The 32-bit float file got even bigger when exported. Turns out I was using 16-bit PCM before

[solar torrent]

[long compass]

Maybe because the long audio file is about an hour long? About 600-700MB. Total dataset size 20 hours

[polar flax]

it should be 700 MB for 1.5 hr audio and it should work

doesn't make sense if it's 20 hrs, unless in mp3 format which is also not ideal to do

[long compass]

Hmm it's just that if I try to do it with only half of the dataset there is no error so I thought it made. If I export the file to 32-bit float, the file size becomes 1.5GB 😁

[polar flax]

because it is stereo, but it will always be converted to mono in preprocessing

[tranquil lantern]

@covert lake I need line of code or that 1 file that can fix the split bug infer for Applio Kaggle

or if anyone here know it

can send

[covert lake]

I told you to use #✨│ai-help not here

And no I don't know about it, maybe @chilly lake does and can help you in that channel

Be patient and use the right channels pls

[tranquil lantern]

okaywait

it's so down 😂

[covert lake]

Noobies is an applio dev so maybe he knows the fix

[tranquil lantern]

should've said yes when codename offered me to fix yesterday but it was nighttime

[chilly lake]

the fix is in the main branch

[polar flax]

yea delete this highlighted part to use the main branch

[tranquil lantern]

highlighted.. right?

[polar flax]

yea that in kaggle

[chilly lake]

also this part may not work with main

[tranquil lantern]

lowkey why didn't they use the main branch in kaggle

[chilly lake]

it is experimental

[tranquil lantern]

are these things for training only

[chilly lake]

yes

[long compass]

Oh no.. I'm dumb. It seems trying to make a 48k model gave errors at preprocess stage because Sample Rate 40k works fine

[spare tree]

Hello
I'm looking for an experienced Full Stack AI Engineer.

what you'll do

• Develop and optimize the platform’s backend and frontend components, ensuring high performance and scalability.
• Implement natural language query capabilities, integrating AI models to enhance system intelligence.
• Process and visualize satellite imagery using proprietary algorithms for geospatial analysis.
• Improve database architecture for efficient data retrieval and real-time analytics.
• Work closely with data scientists to transition Jupyter Notebook-based Python scripts into frontend JavaScript for seamless visualization.
• Design and implement interactive map-based visualizations using Mapbox or similar technologies.
• Develop features such as comparison tools for analyzing environmental changes over time.
• Collaborate with cross-functional teams to ensure smooth integration of machine learning models and geospatial analytics.
• Optimize platform performance by identifying and resolving bottlenecks in data processing and rendering.

requirements

• Strong proficiency in Python, particularly for geospatial or machine learning applications.
• Experience with frontend development, ideally using Next.js or React.js (flexibility in frameworks is welcomed).
• Solid understanding of database structures, optimization, and performance tuning.
• Familiarity with geospatial analysis tools and libraries (e.g., GDAL, GeoPandas, QGIS, ArcGIS, Mapbox) is a plus.
• Strong computer science, engineering, and problem solving skills equivalent to that of a solutions architect or systems designer.
• Strong interest in satellite imagery, developing GIS applications and AI.
• Ability to work independently and proactively identify technical improvements.
• Familiarity with UX/UI principles and ability to enhance visual presentation of geospatial data.

If you're interested in thie position, Pls DM me. Let' s connect!

[hollow dagger]

Ramadan mubarak

[ionic pumice]

ramadan mubarak to you too talha

[trail wind]

selam
bu programın adı neydı sıldım adını unuttum
yazar mısınız

[final sluice]

Ramadan mubarak!!!

[gray rover]

the fuck Ramadan mubarak means

[final sluice]

just a way to say happy ramadan

[gray rover]

aaaand.. ramadan is 🤔 ?

[final sluice]

like yall say merry christmas ig

[gray rover]

ah

[final sluice]

a holy month in islamic calendar

[gray rover]

Right 👀

[gray rover]

Oooo

[final sluice]

fasting 30 days straight... but worth it

[gray rover]

oh yea, think I remember hearing about it somewhere

Anyway, thanks for letting me know

[final sluice]

aye yw

[gray rover]

show the ui

screenshot

your threshold is most likely set wrong

[forest quarry]

it is similar to lent in Christianity

[chilly lake]

I gave up arguing with stupid people for lent

🙏

[sharp rune]

Hey there ,does anyone from India here

[elder willow]

[cosmic snow]

yo

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[wheat glen]

hopefully not

[gray rover]

uah

[solar torrent]

I'm not from India.

[proud marsh]

Does anyone knows or have idea of why this is the most popular model on weights.gg?

like, seriously, I dont get it

[glad nebula]

meme from 1 year ago

[solar torrent]

Saiba Momoi is a character from the mobile game Blue Archive. People are mostly using it for some memes like "oh my gotto, its thing" and "Nipah".

[glad nebula]

https://knowyourmeme.com/memes/oh-my-gotto

[solar torrent]

Yeah, that meme.

[proud marsh]

oh... i see

its funny, cus I never heard of this meme.

but the villager (wich i know well it was vastly used as a meme for some reason) has half the uses

something different to see to say the least

[weak plinth]

I just saw this and download it, now i don't know where the model is 🤡

Do i need download it separately? if yes where?

[polar flax]

you have got a nice console/pc, and now you need your favorite games to play that aren't pre-included 🤡
such as https://huggingface.co/GaboxR67/MelBandRoformers/

[twin fractal]

Used a lot in memes

[queen kernel]

Hello. Can someone please help me to install kokoro tts ?

[chilly lake]

pip install kokoro soundfile

[tranquil lantern]

Racist jokes n word

https://tenor.com/view/blue-archive-ni-ga-gif-6122008057183992860

[solar torrent]

[weary temple]

Momoi

[solar torrent]

Some people I know in 2025 - luther (Kendrick Lamar AI Cover, vocals only)

[queen kernel]

That's all ?

[chilly lake]

well, you need python installed, use 3.10 or 3.11

[queen kernel]

Okay. Lemme try

[silver bronze]

hey weights deleted the option to use youtube to choose a song for ai song creation. any new good apps for this purpose?

[weary temple]

i recomend 3.11

[chilly lake]

3.11 has better error messages

[covert lake]

just download the youtube video urself and use it as an input in weights.com

either pay for youtube premium, or use cobalt, or yt dlp, or literally google "free youtube video downloader site"

[chilly lake]

and get a virus 🙂

I think I used this https://github.com/StefanLobbenmeier/youtube-dl-gui

[elder willow]

Would anyone help me find a simple male voice model? No celeb, no anime, no weird voice. Just high quality normal speaker

[river adder]

You ate with that

Fucking slayed

[covert lake]

Yt dlp and cobalt are safe

[covert lake]

You can search rvc ai voice models at:

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

if there isnt one, you can:

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

[hidden grotto]

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

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

[covert lake]

AI has to be trained on something

[chilly lake]

i mean googling something 'free' and end ending up on a download page with 10 fake 'Download' buttons with virus links

[covert lake]

Smh

[elder willow]

I know where and how to search, I am just not able to find a simple male speaker with not overused voice.

[bold yoke]

/chirp

/create

[covert lake]

There's no Suno AI bot

Promos ain't allowed and will be deleted

[orchid pasture]

aveti vre un model cu calin georgescu?

[fresh folio]

i see

[fathom flax]

whats the best vocal isloation today that is free?

[gray rover]

uvr / mvsep

https://docs.google.com/document/d/17fjNvJzj8ZGSer7c7OFe_CNfUKbAxEh_OBv94ZdRG5c/edit?tab=t.0

This should be helpful

Overviews, info on models used in uvr / mvsep and much more. Generally a 101 guide

Other than that, there's " audio separation " discord server ( google it ) where uvr and mvsep devs are, helpful and informative community

But I personally recommend using gabox's voc fv4 model for vocals / voice

[fathom flax]

thanks!
and right now whats the best free way to train a model, havent been that for months

[gray rover]

you mean rvc models?

[fathom flax]

yes

[gray rover]

Well, if not locally then colab or kaggle really

[fathom flax]

ive done alreay the one with google

there is something new?

[gray rover]

But new in what way? What are your expectations?

If you're asking quite literally about something better than rvc / applio itself? then no

[fathom flax]

i mean something whre i shouldnt work too much

like

drop a clean 30 minutes file

and then wait

🥲

[gray rover]

Unfortunately no, training good models requires a bit of work

[fathom flax]

there is an easy guide?

[gray rover]

check ai hub's docs

or research channels on this discord

[fathom flax]

its updated?

[gray rover]

I think so ye

In any case, you can leave a msg here and some helpers ( hopefully ) could help you out with stuff

Alternatively, ask on #✨│ai-help

[fathom flax]

thanks!

@gray rover do you know if youtube support FLAC?

[gray rover]

as in?

you mean in terms of ripping the audio?

[fathom flax]

like if the audio from youtube is already compressed by them and then i only downloading a large file but with and m4a quality

[gray rover]

oh

the thing with youtube is

any audio people upload, ends up getting dynamically compressed ( volume dynamics ) and undergoes general compression ( codec wise )

all of that is either opus or aac

( it's why people shouldn't use stuff like yt to mp3, because you'd further compress the opus or such to mp3 )

[fathom flax]

so whats the best way?

can i use spotify? apple music?

[gray rover]

yt-dlp imo

cli tool for downloading

the command would be:

yt-dlp.exe -x URL

-x argument tells the program to fetch on the best available quality from their servers

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[gray rover]

mostly it's .opus or .webm ( which will contain opus )

rarely aac

then you'd use ffmpeg to convert opus to wave

[fathom flax]

yo wait

too much

i aint following

[gray rover]

Here, you download the .exe
https://github.com/yt-dlp/yt-dlp

You run it like so

cmd can be opened in the address bar

Here

the url is your youtube video's link

then ( as long you have ffmpeg installed and properly added to path / configured )

[fathom flax]

oh ok

[gray rover]

In the same command line window

ffmpeg -i input.opus output.wav

[fathom flax]

i think i download in the past ffmepg

[gray rover]

input is whatever yt-dlp downloaded and output yea, converted thing to wave as output ( just name it whatever you want and have it .wav at the end

[fathom flax]

so how the whole line should be?

[gray rover]

yt-dlp.exe -x URL
for downloading stuff

ffmpeg -i input_from_yt.opus output_from_ffmpeg.wav
For conversion of opus files to wave

You'll then get a 44.1khz wave files
( and keep them that end-to-end. If you work on those files or process / denoise or whatever, always export them as 44.1khz wave. Those will go to rvc )

And that's pretty much all there is to it

Nothing too crazy

[fathom flax]

hold on a second

i first need to download the file

then i open again cmd?

[gray rover]

As you can see in the folder there's the yt-dlp.exe file

it's the one from github

[fathom flax]

yes

[gray rover]

It downloads the yt audio to that folder

[fathom flax]

i downloaded right now the first clip

[gray rover]

yes, check it's properties

[fathom flax]

opus

[gray rover]

yes, in that case:

ffmpeg -i yourfile.opus yourfile.wav

-i is an argument for input

I name my stuff as songWAV.wav ( the output from ffmpeg

to avoid confusion

[fathom flax]

so i need to first change the file name?

[gray rover]

nope

[fathom flax]

beacuse its too long

[gray rover]

just add wav suffix

before extension

gonna help you keep it clean

if you download a lot of stuff ( and keep opus copies ?

tho ye, you can rename stuff ofc

for the output the name doesn't matter

[fathom flax]

lets say i download via yt dlp a file that his name is: blabla.opus

[gray rover]

ye, then that's for input, output you can name it whatever

[fathom flax]

whats the line will be?

[gray rover]

ffmpeg -i blabla.opus blablaamazing123.wav

[hollow dagger]

Hello

[fathom flax]

oh i see

[gray rover]

yup, pretty simple

[fathom flax]

let me try it

where can i upload images?

or i can dm you

[hollow dagger]

I need some sample data to practise creating a chat agent, such as a business, and I need a lot of them so I can create a lot of chatbots. Could you guys please help me with this?

[warm sapphire]

hey guys what is the best model for realistic female voice ?

[night lake]

this: https://discord.com/channels/1159260121998827560/1341216399372062823

[polar flax]

https://tenor.com/view/jeff-land-shark-marvel-rivals-video-game-gif-10022472975975249

[lethal flax]

is there a locally running app of any sort that lets me inpaint/remove items from videos?

[solar torrent]

I don't think there's any AI tool that can edit videos in bulk. There are websites that generate video after a frame of a video or an image.

[polar flax]

only those that can swap faces

[deft surge]

Bandidu não dança dança

Bandidu ginga e balança 🔥 🔥 🥶 ☝️ ☝️ 😂 😂 😂

[ionic pumice]

fr whatever that means

[grand breach]

What is our thoughts on Grok 3?

[solar torrent]

Never used Grok.

[grand breach]

what LLM do you use

Claude 3.7 sonnet thinking is really good at code

[solar torrent]

Microsoft Copilot

[grand breach]

aintnoway

LOL

[polar flax]

https://www.tomshardware.com/video-games/xbox/xbox-announces-copilot-for-gaming-ai-assistant-early-access-coming-to-xbox-mobile-app-more-details-to-come-at-gdc-2025

[dusk geyser]

hot take: gabox fv4 is the best model

[queen kernel]

What I have to do after this ?

I have installed kokoro

[chilly lake]

activate the virtual environment if you made any, then use a script

each line comes out as a separate file, but they can be merged into one

[queen kernel]

I was waiting for your response 😭

Thank you

[chilly lake]

why, the github has an example script

[queen kernel]

Got an error

Espeak is not installed

That's why I was waiting for your reply

What to do

[chilly lake]

read the github page

[queen kernel]

I have installed espeak

[chilly lake]

environment variable

[queen kernel]

I just created environment variables but with different names. My bad

[chilly lake]

use a new terminal window after that

[queen kernel]

I restarted my PC.. lemme see if it works.

I was asking questions in kokoro discord server and they was very rude to me 🥺

[queen kernel]

It said failed to load voice "hi"

[stuck ridge]

i have rx 7600 and a rode mic

how do i make this work chat

[queen kernel]

What?

[icy pendant]

You should specify what "this" is but i assume you want realtime voice changer

https://rentry.co/ForkVoiceChangerGuide

Download AMD version, virtual cable, read audio setup, model upload etc.

For questions ask in #🔍│help-w-okada

[stuck ridge]

ok! thanks

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[chilly lake]

you need to actually use 'h' as lang code, not 'hi'

and the voice name from the list

[real iron]

I want a very good female voice model

[queen kernel]

I did the same. But getting error failed to load voice "hi"

[chilly lake]

you should've provided the full error, not jhust the last line

[queen kernel]

Oops sorry.. lemme send you

[chilly lake]

is they key, it is e-speak problem

so hindi language goes thru some weird phonemizer and you need that installed and I have no idea what

you need to pos an issue in kokoro's github

someone may answer

[solar torrent]

No, thanks.

I do things that don't really get me bored.

[tough beacon]

jakkari jakkari

[knotty meteor]

is there an ai site or something like that that can help me modify some text on a photo

?

[bitter wagon]

could someone suggest me a available voice model that sounds more like a matured man with deep voice like a man

[supple stream]

you mean image genarator

[topaz niche]

Hi

Where’s the ai

[covert lake]

elaborate

[gentle cipher]

I need help

Are you an admin?

[covert lake]

Elaborate the issue in specific help channels based on the software you need help with

[covert lake]

Junior admin

[gentle cipher]

Ok

[mental pelican]

somebody have juan gabriel link of hugging face

[sterile stream]

Hello. I am new to using ai tools and was wondering if i could be given any tips on how go get into new ais and the idea of using other AI than chat gpt. The only(almost only other uses were major)tool i have used so far is chat gpt for coding.

And also have gotten advice and information from it

I would love advice on AI for coding. i don't know much coding so I would love advice on AI that help you understand. I am willing to also purchase with money premium versions of AI at a price of 20$ a month

[polar flax]

claude 3.7, if you want the project feature, check here
https://tactiq.io/learn/chatgpt-project-vs-claude-project

[woven magnet]

Hi. I am also new to using ai tools. I want to learn about generative ai to build a tiktok channel using ai to make video , but i dont know where to start . Can anyone give me some advices. ( btw i dont have any background on Ai,)

[gray rover]

Tone it down

🤔 now that's some crashing out

counter strike might be stressful m8 but you should chill

[simple gate]

is Ilaria RVC on huggingface.co down ? been a week i cant convert

[floral cairn]

yall know a free alternative to Krea AI Training? Like a model that you feed it images and it creates images like those

[polar flax]

comfyui, fluxgym, etc

[sour temple]

a

[floral cairn]

any online one? I see both of those are like code or something

[solar torrent]

Weights.com can train a Flux.1 image model there.

[solemn wren]

yo guys do you know any website i can turn MIDI files into mp3 vocals?

preferably free

[strange wraith]

A convert?

Send to me I got you g

[ionic pumice]

synthv, vocaloid, or cevio

[solar torrent]

You can use a soundfont full of spoken vocals, and convert that MIDI file into mp3.

[strange wraith]

He might not know how to use them. But yeah those work as well boss

[void dock]

hey....
hey ! i am new here . i am software engineer . i am working on model training projects ..

[strange wraith]

Awesome welcome !

[lyric rapids]

guys whats the diffrience between w ocada coice changer and rvc???

[strange wraith]

Rvc is for music and vocal training and Inferencing music

Ocada is a real time changer to sound like the desired person on the spot

Via game discord etc

[lyric rapids]

okk

[strange wraith]

TTS is spoken word to voice

🙂

If you need any help getting started feel free to reach out

[solar torrent]

The correct name for the realtime voice changer program is W-Okada, not W Ocada.

[strange wraith]

My man

Sup namari

[lyric rapids]

so i shoukld use w okada to changemy voice on discord calls right

[strange wraith]

Yes

[lyric rapids]

plus is there a easy softwear other then okada

like catfish or i forgot the name?

[strange wraith]

Doesn’t work as well imo

If you need help with setting it up I can get you

[solar torrent]

Don't use W-Okada for catfishing someone.

[strange wraith]

He’s asking if the other program

Is named that

No it’s not

[lyric rapids]

oh ok

[strange wraith]

It’s more for like

Having fun

[lyric rapids]

when ill setup and of i need help ill ping u

[strange wraith]

Not Pretending to be someone and scaring one

That’s illegal

[solar torrent]

Don't use Voice.ai. It is a scam site that trying to eat your PC more than W-Okada.

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

[lyric rapids]

wdym more then w okada?

[strange wraith]

It uses more pc

Than needed

[lyric rapids]

https://youtu.be/FzU-H_-zOc4

[solar torrent]

Performance. What do I mean?

[pine acorn]

Staff Applications Open

---

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

Click here to apply!

[lyric rapids]

what abt htis one?

[lyric rapids]

ok if i opress stoip in there

[solar torrent]

Don't send a YouTube link without the context here.

[lyric rapids]

then my pc wont be used up right?

[strange wraith]

Just use okada

It will be better

[lyric rapids]

ok

but

it uses alot of PC recources

[strange wraith]

Apply still isn’t working just letting you know

[lyric rapids]

right?

[strange wraith]

And performance isn’t as good

🙂

[lyric rapids]

what apply?

[strange wraith]

Oh not just taking about something else

Your fine!

Thanks for being soo kind

I’ll be here for any needs and so will other kind members

Like namari

👍

[lyric rapids]

ok

[solar torrent]

Shit. For installing and such about W-Okada, go to #🔍│help-w-okada. The website for mod/helper application for this server is broken right now.

[strange wraith]

I know

Just re iterating it apologies

[void dock]

@strange wraith are u a chat bot? i think u know every thing 😅

[strange wraith]

Nah bro I just been in so for a min just trying to help

I use to struggle soo much with this shit

[solar torrent]

I'd just take it as an insult if you call me like that.

[strange wraith]

Just like helping out

It’s okay ahah. Take it how you want it just wanna treat others how I would like to be treated

🙂

[void dock]

no no ,, i am realy sorry if feel like that ..

[strange wraith]

Nooo I took no disrespect