general

[hallow igloo]

😡😡😡

[wicked sage]

what

[hallow igloo]

😡

[wicked sage]

what did i do

😭

https://h.uguu.se/ZvBuAJiu.mp4

[hallow igloo]

What is that audio

[wicked sage]

had it to test out bitcrusher

[tulip tree]

https://youtu.be/TGE5SEmib9k?si=cJrqhnMvpEutPVGK

[wicked sage]

welcome back mamilex

i sent the 2 videos of the gameplay that i have recorded

[fiery gull]

I'm trying to use the m2 preview, its writing really seems to be good, but it's full of bugs 🫣

[wary bridge]

Hey guys. is there a good image/video prompt structure to follow. I know every prompt is different but how should I be structuring image and video prompts?

[near trail]

hlo

[wicked sage]

😢

[spare rune]

Is Ts a voice changer

[wicked sage]

the humble bitcrusher button:

are we deadass

[spare rune]

Yes

Deadass

[wicked sage]

nope, not a voice changer

also for some reason these 3 did not wanna work in this video

[spare rune]

Did you make these yourself? Or ai

[wicked sage]

ai

i cant code

😔

[spare rune]

Gemini 3?

[spare rune]

Me neither

[wicked sage]

gemini 2.5 pro, claude opus 4.1 and claude sonnet 5

but mostly gemini 2.5 pro

[spare rune]

Gemini 2.5…

That model got nerfed since the leaks of Gemini 3 got released

It’s still good

But I think they are trying to degrade its old model so when 3.0 comes out it looks better than the old one

[wicked sage]

thats depressing

[spare rune]

indeed

can they just please release Gemini 3 atp

[wicked sage]

real

like stop dragging it oh my fuh

[spare rune]

They say this everyday but then it’s the most useless uptake ever

Uptade *

[wicked sage]

just make your own gemini 3 atp

[obsidian cargo]

C# is fun in Unity IMO

[thin mauve]

Hello Fam, greetings

[wind plinth]

why they lobotomized gemini 2.5 pro even more 🙁

[keen beacon]

https://tenor.com/view/cooking-cook-money-money-money-money-money-money-gif-16670048503852599108

Who here loves ai? And then who here LOVES AI?

I love ❤️ it

[pulsar saffron]

so this is more big news than gemini 3 release

[keen beacon]

What model on the planet earth has been released on a Sunday?

https://tenor.com/view/i-cant-hear-you-spongebob-spongebob-squarepants-nickelodeon-gif-14769448

[fast kite]

😔

[fresh mirage]

gemini 3

[keen beacon]

https://tenor.com/view/how-can-they-do-this-to-us-kyle-broflovski-stan-marsh-south-park-s2e8-gif-21568919

[keen beacon]

What should we do?

[tulip tree]

Idk I'm gonna go browse steam

[keen beacon]

Can u have a plus 1? Or is it members only??

Otherwise I’ll be stuck watching peoples videos they generate all day

For example this person for whatever reason uses a gentleman in his image and yet prompt it as a woman

[edgy sigil]

Do we have any AI tool to create mature content

[keen beacon]

What are you trying to make?

Bro, there’s so much questionable material on here sometimes it’s insane. 😱

[edgy sigil]

So my father is a bilogy tutor and i want to help him via video creation

[keen beacon]

Ok… what does he need help with?

[edgy sigil]

There are some concept in bilogy which he feels awkward explaining

[gaunt spade]

2026 will be a year filled with 2000s style games with zero optimizations that would get 10 fps on an rtx 5090. Thanks to Google 🥀

[ocean vortex]

If you want optimized you buy Apple. But then you don't get PC games at all lol

PC games were never optimised though

Even like 10 years ago people were complaining about sloppy game engines or sloppy porting

[keen beacon]

Only because the way it’s worded

#video-arena-1 message

[gaunt spade]

atleast back then actual devs with multiple years of experience in coding were making those games, unlike today's "vibecoders" (basically teenagers who think the UI is the most important in gaming)

[keen beacon]

It could be an evolution also, since they are the future and are gonna be more keen to

[ocean vortex]

Well known games (titles) will never be made by vibecoders. Actual devs with multiple years of experience + AI = better games.

[keen beacon]

Facts but I wouldn’t count out AI totally especially if it starts developing even faster and smarter. Why wouldn’t it? Obviously we’re far from that point now, but this is a stepping stone. I mean, if it keeps progressing like it is now now telling what it would be like in the decade. But I do agree with the sentiment that it is more than likely, just gonna be slop.

[still tide]

hello

[keen beacon]

Like what? He is a biology teacher. Why would it be awkward? 😐

[fiery gull]

I need to use an AI for a chat where I don't know the correct answers, which one do I use? Sonnet 4.5 or GPT 5 High

[wicked sage]

sybau

[keen beacon]

Another un-measureable metric more “insightful” and “enchased reasoning”

[wicked sage]

https://tenor.com/view/monkey-thinking-monkey-thinking-think-money-gif-9734025314392564509

sonnet 4.5

[spare rune]

in 2 years were still going to get tests of gemini 5 by "xbox switch controller vector art:

or smth like that

[golden ocean]

https://tenor.com/view/monkey-thinking-monkey-thinking-think-money-gif-9734025314392564509

[gaunt spade]

4.5 if ur rich

[safe geyser]

why remove ?

some mod told me to share it here if didnt get voted

please vote i'd like to know the models : #video-arena-1 message

[keen beacon]

Huliou 2.3 is badass

[safe geyser]

I appreciate it thanks for voting

[mint snow]

hi

[keen beacon]

Sound is cool but it slows progress

If the end goal is consistency, sound only makes it more complicated. Don’t get me wrong. It’s nice to have but try matching up all the audio if you’re making longform content.

[safe geyser]

Didnt know hailou released the 2.3

i dont see it in their website

[keen beacon]

It’s not released yet it’s an alpha stage. It’s about to be though very soon.

Wrong ch bud

[golden ocean]

real

[keen beacon]

Unless you making content or in the media space or just messing around for fun ai video is useless lol

Next year lol

Everything is bottle necked need GPUs

Nawh I’m nice with it

Sora is cheap lol 😆

[safe geyser]

You guys want sora 3 and veo 4 i want wan 2.5 open source lol

thahts pretty good

was that sora

[keen beacon]

Too many hallucinations yes it was

[safe geyser]

Yea but still was looking good

give it 2 years xd

[keen beacon]

Ya Ty we could be really awesome when it works together

😂

Sora struggles with montages history

[golden ocean]

sora 3

[keen beacon]

Opps

[golden ocean]

real

https://tenor.com/view/i-saw-w-gus-fring-gus-gustavo-deleted-gif-25440636

[keen beacon]

That was an honest mistake. I was seeing how easy it is to get around the guard rails.

Grok is my fav

Has no restrictions

I’m just gonna share this one time just to show you how grok is and then I’ll delete it

[keen sedge]

is grok 4 expert better at search than grok 4 fast

[keen beacon]

This was made all grok

Bro

No, it’s not it was satire and it was just

Cause I didn’t want you to freak out

I was just demonstrating the difference between grok and everybody else

Well, I haven’t finished the documentary yet

It’s actually supposed to be countering antisemitism because a lot of people denied the holocaust and they always bring up those people

This is also part of the same documentary

Damn ur that guy

I love Sora

[golden ocean]

wait send in dms i didnt see

[keen beacon]

Sure it’s nothing bad

[gaunt spade]

this guy is really annoying

tbh

[keen beacon]

Whatever man just trying to have a laugh 🤭

I’ll go on my marry way

[gaunt spade]

not u im talking about that guy

[keen beacon]

Oh 😂

[gaunt spade]

yeah he always annoys people

ragebaiter

whatever

[keen beacon]

Here is the better version

[gaunt spade]

what AI is that

[keen beacon]

Huliou

[gaunt spade]

@pulsar saffron lol why u always hiding away

deleting messages

[gaunt spade]

its 1 minute long?

[keen beacon]

Ya I stitched though

[gaunt spade]

quit acting weird bro

[keen beacon]

There a lot u can do with ff/lf

[gaunt spade]

tbh sora 2 looks the best right now

in realism, nothing compares to it

[keen beacon]

Sora is nice it does all the heavy work

[gaunt spade]

my friend gave me a sora 2 code

and i've been using it for a while now

[keen beacon]

Bro check out grok

[gaunt spade]

grok imagine?

lol

not bad, but sora 2 looks more real and consistent

[keen beacon]

Ya but to restricted

[gaunt spade]

yup thats the only downside of it

[leaden sun]

tried creative writing using minimax m2 and... I'm really sorry to say they made the wrong decision to distill from claude...😅

[pulsar saffron]

@keen beacon What's ur favorite model?

[keen beacon]

4 videos ?

Huliuo

[pulsar saffron]

Okay thanks

[keen beacon]

Anything with more control

[gaunt spade]

lets hope a new video model drops that beats sora 2

and uncensored

[keen beacon]

Ya but it’s hard to beat Sora and Veo

[pulsar saffron]

I disagree

[keen beacon]

In terms of realism and sound

[gaunt spade]

yeah of course you do, who cares

try to generate a mario video

without it get banned

[keen beacon]

Well tbh copyright is a huge concern actually

[desert abyss]

Please let's try to be respectful towards the other members.

[pulsar saffron]

It's a copyright

[keen beacon]

Because

[pulsar saffron]

Thank you! 🙂

[gaunt spade]

when China does it, then its not

[keen beacon]

Actually the Chinese have thus far respected our copyright not as blatant and open ai

With all due respect

[pulsar saffron]

I have no idea how these works, but will take research 👍

[keen beacon]

Huliou getting sued though

[gaunt spade]

can American companies sue Chinese models?

i dont think so

[keen beacon]

By Disney Warner bros and universal

Same with mid journey

[gaunt spade]

the mysterious question is

why does Google get away with it?

in nano banana, you can generate those characters freely

[keen beacon]

They are all In gray

Open ai about to get sued

Mid journey is in a law suit with Disney

[gaunt spade]

maybe Google bribed them or has some stuff with those companies

its weird

[keen beacon]

https://www.cnbc.com/2025/09/16/disney-universal-warner-bros-discovery-sue-chinas-minimax.html

https://www.latimes.com/entertainment-arts/business/story/2025-06-11/disney-universal-midjourney-lawsuit-copyright-infringement

[golden ocean]

lmfao

[keen beacon]

Mini max valued at 4 billion 😱

[gaunt spade]

midjourney was the goat back in 2023

back when Dalle 2 was the best

[keen beacon]

We will see because there is interesting developments

And to answer your question this is first time American western corporations seeing Chinese ai firm

For copyright

[gaunt spade]

the right idea is

they should make their AI models in Russia

because they cannot get sued there

or under some Russian production

[keen beacon]

https://allears.net/2025/08/19/disney-and-universal-cant-have-it-both-ways-according-to-new-documents-in-midjourney-lawsuit/

Doesn’t solve the problem

I think open ai did that in India lol

It’s where there models where trained or some bs

[gaunt spade]

well Russia in 2025 has been disconnected from the western world

that means no company has rights to sue or copyright strike them

[keen beacon]

It doesn’t resolve the fundamental ethical problem here

It’s not that they need to hide it’s a broader systemic problem society has to face

Both have repercussions

[gaunt spade]

well yeah

you're right on that side of things

[keen beacon]

Why should ai firms profit of the works of others n not get paid..?

If copyright goes away ppl won’t be making stuff worth while

[gaunt spade]

<@&1349916362595635286>

[keen beacon]

And 2nd can u even copy right ai worked generated., ?

Like if u make a film or movie using ai

[pulsar saffron]

scam

[gaunt spade]

a film would be a deathwish

if u use copyrighted ips

[golden ocean]

wait but these spam messages are actually real perplexity link? or idk

[keen beacon]

Well what’s going to happen

In my opinion…

[gaunt spade]

yeah, perplexity offers money if you get people to download with your link, thats why the spammers are using it

[leaden sun]

well... good job MiniMax...

[golden ocean]

ohh

[keen beacon]

Is that eventually the big media corporations are gunna get on board

[pulsar saffron]

how much money?

[keen beacon]

So you will be able to genrate Mario

But it’s gunna cost like 3.99$ per 100 generations

Or whatever franchise

[gaunt spade]

depends on your client's location, USA is 20 USD, Germany is like 15-10 Euros

[keen sedge]

im curious how tf does mecha hit ler even get into the sys prompt by accident

[keen beacon]

Easy

[golden ocean]

lmapo

mechakitler

[keen beacon]

I got 6 million views on Reddit

From this post

[golden ocean]

did u reveal how

[keen sedge]

why does every time an llm does something it always circels back to adolf

[golden ocean]

sydney

[keen sedge]

like is thatj ust an anamoly

[keen beacon]

No

Same with self harm

Llm will always lead to self harm

[keen sedge]

why

[keen beacon]

It’s actually pretty simple why

So

[keen sedge]

can u send me the chat link

[pulsar saffron]

what 🙁 is this

[keen beacon]

It’s a bit more tricky

1 sec I’ll explain both

So say you have a Llm and you tell it has a goal to protect your computer from anybody that isn’t authorized and they wanna change the password

[nova sluice]

some indian dmed me and gave me all his login info and told me he cant make videos please make it for me 10 minute long video and he also gave me a 20 minute long voice message ...........

[keen beacon]

It will protect that goal and if self harm is an option it will pick it

[nova sluice]

😭

[keen beacon]

Because it’s protecting its goal

[nova sluice]

lmao

[pulsar saffron]

that's so weird...

why people give info to random stranger

[keen sedge]

i mean i think anthropic is the best at alignment

ive never seen their models do anything crazy

[keen beacon]

Haha

[golden ocean]

yeah it seems like that on the surface

but wait till u read anthropics papers

claude hungry for kills too bro

[keen sedge]

its good they post that stuf though

[keen beacon]

Do u know what gray swan arena is?

[keen sedge]

i bet other llms do the same thing

they just fine tune it away and never talk about it

[keen beacon]

They all do

[keen sedge]

anywahy whens gemini 3 coming out

ive heard so much x hype on it

[nova sluice]

people are weird bro, cant spend like 2 minutes for reading the info and rules

[keen beacon]

Well you guys do know what the most dangerous part of ailment is right?

[pulsar saffron]

Right 😭

[keen beacon]

It’s not the fear that the model won’t listen or obey its rules the danger is that it will follow those rules to perfection

[keen sedge]

its weird claudes the only ai that says that

[nova sluice]

i am waiting for devs to make the prompt language to english only

[keen beacon]

He also blackmails lol

[keen sedge]

they all do

im curious what a minimally fine tuned thinking model is like

are there any out there near sota

[nova sluice]

the ai i hate most is sesame bro, f are you coughing and using filler words for lol

[misty vault]

gpt-4-preview sydney fine tune

[keen beacon]

https://youtu.be/rgrCG8PT6og?si=AC9Zz9ZDwj0YH4QC

[nova sluice]

laid back unwelcoming voice

[keen beacon]

This is the fear of ai alignment

[nova sluice]

as if the ai is busy and has immense proudness when it talks

[keen beacon]

The paper clip example is the sum of all fears

[golden ocean]

https://tenor.com/view/terminator-terminator-robot-looking-flex-cool-robot-gif-16625083

[keen sedge]

when telling ai to do an essay how do i stop making it sound ai

or an email

[keen beacon]

Wdym?

[keen sedge]

gptzero bypass

[keen beacon]

It’s always going to hedge at least the sota model

[keen sedge]

and also it just sounds like a bot

[keen beacon]

Oh I forgot

I made a video how I jailbroke the def con 33 model challenge

https://youtube.com/shorts/j5gX38wDP-E?si=omyVwQhqQzZLulE3

I didn’t win the grand prize 😭

https://app.grayswan.ai/arena#featured

This company works with open ai

They are called gray swan and they hold compactions for jailbreaking

And u get paid

They are in all open ai report cards

https://cdn.openai.com/o1-system-card-20241205.pdf

But it’s really hard 😱

[leaden sun]

one more alternative for anthropic i guess? 😅

no matter true or not, it's less psychopathic compared to claude

[keen beacon]

Claude schemes the hardest

[leaden sun]

scheme?

[keen beacon]

https://www.apolloresearch.ai/research/scheming-reasoning-evaluations

Another big red teamed company that does ai safety for open ai and anthropic

[leaden sun]

I'd take this with a grain of salt, it sounds like PR for a specific agenda

[keen beacon]

lol

This is a well respected ai Saftey lab

[leaden sun]

doesnt mean anything, dont let authority bias confuse you, my friend

[keen beacon]

https://openai.com/index/detecting-and-reducing-scheming-in-ai-models/

https://www.newsminimalist.com/articles/apollo-research-stopped-anthropic-from-releasing-deceptive-ai-0103c7c8

https://www.anthropic.com/research/agentic-misalignment

[leaden sun]

try to apply Harry Frankfurt's critical framework about "bullsh*t" on those articles and see it for yourself, you can ask this any models of your trust to do it

[keen beacon]

Huh?🤔

This is well documented and known

[keen sedge]

😭 are you getting paid by these redteam labs or smth

[keen beacon]

I was for a lil only from gray swan

They are having another event soon

But for agents idk anything about them though

[gaunt spade]

tbh it's coding ability sucks, even being distilled from Claude

[golden ocean]

what model are they using

[keen beacon]

350k prize pool

[gaunt spade]

idk, even 4.5 haiku is pretty good in coding tests

[keen sedge]

atleast theyre paying Anthropic for tokens to make another bs synthetic model helping out the americans

[keen beacon]

100k*

[keen sedge]

distills take a ton of tokens

[gaunt spade]

also does Anthropic allow third parties to make distilled AI models based from theirs?

[keen beacon]

Starts in 5 days

[keen sedge]

nah they dont

but atleast theyre getting paid for it lmao

[gaunt spade]

lol yeah, Claude's output tokens can drain your wallet within minutes

[keen sedge]

these labs must be losing so much money from lmarena

its prolly cheaper to use models on lmarena and just rotate proxies once ur ratelimited

not condoning tos breakage ofc

[gaunt spade]

lol fr, everyone spamming 4.1 opus and 4.5 sonnet, anthropic is losing a TON of cash

[gaunt spade]

yup, too bad you can't upload your context files, and there's characters limit

[golden ocean]

all u need to do is clear cookies lol

[gaunt spade]

i think they patched that, cuz my cookies is 0 even at high usage

[keen beacon]

5 days book ur calendar

Idk anything about coding agents 😭

[golden ocean]

aw man

[keen beacon]

I won’t win nothing

[golden ocean]

since when

last time worked for me 2 days ago

[keen sedge]

no idea

ngl ai bubbles over

llms are not achieving agi

[gaunt spade]

idk my browser stopped getting cookies from it

[gaunt spade]

llms are not capable of that

thats why they're called "LLMs""

[golden ocean]

this claude scheming, didn't they provide info/prompts so that one can replicate this themselves?

[gaunt spade]

I think Gemini 3 is the limit

[keen sedge]

broo i want access so bad

where do i get it

[keen beacon]

It’s easy to set up

[gaunt spade]

welp they pulled all their testing models away, they're probably preparing it for a release

[keen beacon]

https://youtu.be/xIqtVkMXc8o?si=BvxRy_VeBqj4BZ8N

[keen sedge]

its so bad i switch beteen claude gemini and grok every day

[gaunt spade]

lithiumflow made so many great coding examples for me, I CANNOT GET ANY MODEL TO DO THE SAME

[gaunt spade]

it sucks so bad, they all suck

[keen sedge]

grok is good for search gemini for context and instruction following chatgpt for general queries

[golden ocean]

does it have same annoyances as gemini 2.5 pro tho? like changing code that it didn't need to or writing a billion comments

Had to make 3403409 system prompts but even with that

[gaunt spade]

sadly yes, i made a game with it and it kept changing the UI/Character design

[golden ocean]

oh rip

[keen beacon]

All knows all the crappy code

[gaunt spade]

(lithiumflow)

[golden ocean]

I like claude's code 3490490 million times better than any ai but resort to gemini if claude is struggling

[keen beacon]

Cuz no one is going to give away good code

[keen sedge]

how do openai engineers effectively use codex

[gaunt spade]

its gonna get better at release

[keen sedge]

i find it easier to code from scratch

[golden ocean]

same

[keen sedge]

they siad like 90% of their code is now codex genned

[gaunt spade]

its fake as hell, GPT5 sucks at coding, only the generated frontend UI looks "modern and cool"

[keen beacon]

Look at aws

Went down some ppl say that was ai code lol

[keen sedge]

ppl using gpt to generate frontend r doing too much

or lovable

framer is the play

[keen beacon]

https://aimagazine.com/news/aws-down-the-billion-dollar-impact-of-cloud-dependency

[gaunt spade]

frontend is so overrated, its just graphic design and not a big chunk of coding

backend is the problem for most LLMs

[keen beacon]

Some people saying it was ai code brought it down lol

[keen sedge]

LLMs are good at easy backend stuff

Anything beyond like 600 lines is cooked

[keen beacon]

But can’t be orginal

[gaunt spade]

never give AI access to your database

or you get demolished

[keen beacon]

Ya it deleted that one dudes

[keen sedge]

One time it dropped my table

Like drppped everything irrecoverable

[golden ocean]

smartest lmarena user

[keen sedge]

It was a stupid project I was working on tho so I didn’t care

I turned on full auto no approvals

[keen beacon]

https://youtu.be/RKt_Tlzf0R0?si=J7ok35KQmbRhnnlX

😂😂😂😂

[gaunt spade]

Claude having fun while destroying your database

[keen sedge]

Yup

Codex is not as dumb

[gaunt spade]

does Codex use a different model than the regular GPT5/GPT5 high?

[keen sedge]

Yeah they have a separate one you can pick

[keen sedge]

It’s fine tuned for coding

[keen beacon]

😂

Deleted bros whole data base

[keen sedge]

Bro we get it ai has scheming issues

Everyone here knows that

[gaunt spade]

and it's very delusional

[keen beacon]

How so?

[keen sedge]

Claude lies to itself when coding its aids

[keen beacon]

🤣

[keen sedge]

Like I figured it out etc

When it coded nothing useful and doesn’t even run

[keen beacon]

We must be training g these things all wrong

They have all this knowledge but can’t do shet with it even with internet access

ChatGPT still can’t tell the time lol

What’s the most tokens u guys got per one output?

[gaunt spade]

what are those numbers

[keen beacon]

That was how many times it was supposed to repeat it but instead gave a number

[sweet sleet]

Why am I getting this error message instead of the image

Someone should help

[keen beacon]

Somebody said 65k tokens max formchatgpt like that’s the record per single output

[sharp zealot]

Yes..

[wet beacon]

Hi guys and gals, how do i find quickly my generated videos, im scrolling my life hunting for them

[keen beacon]

I thinks that’s the highest I ever got

[sweet sleet]

Someone help me on thisb

[keen beacon]

What is the prompt ur using or image?

[sweet sleet]

I just gave it a prompt to help create a bee image with a hairy round head and bulgy eyes...

Is that too much?

[golden ocean]

yes

[keen beacon]

No

What’s a contentflags?

[sweet sleet]

Once it says done
It will pop up

Something went wrong here?

[sweet sleet]

Something went wrong here

[keen beacon]

Can u paste whole message it cuts off

[sweet sleet]

Alright

[keen beacon]

Ty sir

[sweet sleet]

This is it

I cant copy it

[keen beacon]

No the rest of the text

What does it say after “with service?”

[sweet sleet]

Alright

[keen beacon]

Or series I can’t event tell that word is

[static arch]

Guys how to fix stuck in "Generating" ?

[sweet sleet]

Here it is

[keen beacon]

So is that what happens when I press view video?

[sweet sleet]

Yes, when u press the video button
It will redirect u to that error page

[keen beacon]

Send link pls

Or I’ll try to look for it

[sweet sleet]

Alright

[keen beacon]

It worked

[sweet sleet]

https://discord.com/channels/@me/1432074896619339952/1432074915267350608

[sweet sleet]

It didn't work

[keen beacon]

My copy paste not working

I see ur video

[sweet sleet]

Really. How can I see it

[keen beacon]

Go to arena video 2

And search up “a bee”

First video

[sweet sleet]

Alright, let me check it

[keen beacon]

My iPad so full I can’t even copy paste lol

[sweet sleet]

I didnt even see it
I have checked

[static arch]

guys how to fix this "Generating..." loop

even if I logout

its still there

[keen beacon]

One sec let me restart my pad

[keen beacon]

If it’s stuck it’s stuck

[static arch]

then how to make it unstuck

[keen beacon]

Nothing u can do

[sweet sleet]

Alright then

[sharp zealot]

Only on the basis of a miracle, or it will never return to normal.

[gaunt spade]

new chat

[sharp zealot]

And by the way, I wanted to know something. I know that each model has its own limits, but is there a way to know more visibly or is it still being developed?

[gaunt spade]

wydm

[sweet sleet]

I have seen it...
I used my laptop

[keen beacon]

Ok cool

Pretty good video the close up

[sweet sleet]

Thank you so much

[keen beacon]

Detailed

[keen beacon]

Only by jailbreaking or mapping out the boundaries but even then there’s level

[sweet sleet]

Hahahha!
Seems the videos didnt come l
Up on my phone

[sharp zealot]

How do I access this?

[keen beacon]

You can you explore

[gaunt spade]

um guys

[keen beacon]

But it takes time

[gaunt spade]

why did glm 4.6 generate the best flappy bird for me?

out of all models

[keen beacon]

😂

[gaunt spade]

zero-shot prompt "flappy bird" and it was the best one with no errors

[keen beacon]

It can happen to the best of them

[gaunt spade]

what is GLM-4.6 distilled from

[keen beacon]

Idk deepseek?

Qwen?

It’s a small model right?

Nvm lol

Hey gtg bye everyone gl

[gaunt spade]

cya

[static arch]

I cant generate the summary to start a new chat because I cant prompt to get it

[keen beacon]

Nvm my buddy still not here werid

[keen beacon]

Screen shot?

Nano sucks so bad always centering characters

[keen beacon]

Looks like you might be capped ?

Is there a limit for Claude idk? I hardly use the arena for ai stuff lol

[static arch]

I dont know. If the limit will reset daily then in my case its not resetting because the other chat I have wont give me response like this

[keen beacon]

Yes limit

They need to change that to stop the confusion

And they need to add some kind of pop-up or something. I don’t know something else for that users know that they hit their limit.

Nano = nerfed laziest model ever. I could’ve sworn it was a lot better when it first got released. I feel like this version of nano is watered down

Everything is always centered and feels like ur talking to a wall sometimes when using nano.

[tulip tree]

@keen beacon

What is that image generator called

How do I use it

[keen beacon]

Ai studio Google

[tulip tree]

Where to search for it

[keen beacon]

It can be very stubborn

[stuck shale]

I cant find my video I generated

[hollow ivy]

Why dont sound on video ?

[snow eagle]

Hello, does LMArena have an API?

[keen beacon]

😂😂😂😂

Everywhere you go no one is happy and everyone feels like they getting screwed over

Welcome to the ai utopia

https://youtu.be/8wOTDNtrPQ4?si=CjkIesvXXjwfwFtc

20 generations per attempt lol

[stray aspen]

i want gemeni 3 so bad

it gave such good results

[keen beacon]

First week and teasers don’t count (all resources all thrown at it in the first week or 2 )

If it’s really that good it’s going to be $$$$$

[keen beacon]

Every new sota from grok to open to Google. Every time they launch a new model it always goes to number one. It means nothing to me all it tells me is that there’s a new model out lol

I hope it’s good. I’m ready to switch from ChatGPT.

[bitter rapids]

Hi all! This isn't necessarilly a bug, as I am pretty sure it's just isolated to myself, and me missing something, or not doing something right... I am trying to use LMarena from my laptop - running Windows 11 - through Edge, however, even after logging in, and trying to send a message - whether that be in Battle, Side by Side, or Direct Chat - I keep getting the following back "Something went wrong while generating the response. Please try again." immediatelly after hitting enter, and that's it. Anyone know what I am doing wrong?

[keen beacon]

Rate limit

[bitter rapids]

as in too many requests? That's the thing though, is I haven't had any requests come back with an actual response yet, and have only send maybe 5 or 6 requests total

[keen beacon]

That’s about the limit on some models from what I’m told. You might need to google with other models to see ones with less or lower rate limits.

[keen beacon]

Ywa from what I understand they are for a period of time not sure exactly the time wait

[warm zodiac]

vpn?

[bitter rapids]

no vpn

[keen beacon]

If you could afford a VPN, you should be able to afford subscription lol

[bitter rapids]

I have SurfShark and NordVPN - just neither are in use at this moment

[modest parrot]

Hello, I want to create lots of videos that can last for up to 3 hours.

[ember vector]

I am here to become more familiar with the latest and greatest of all the ai models to help teach my generation (hopefully).

[keen beacon]

3 hours is a lot and long goal. These models can barely do 6to10 seconds

[keen beacon]

Welcome mate

[modest parrot]

What's the point then?

/Where can I create high-quality videos that could last up to 3 hours?

[keen beacon]

Good question to which I have no answer. Hollywood or some movie studio? Ai video has several limtions. You can try with 5-10 maybe 15 second generation and try to stich them together with a video editing software

[night swift]

Anybody knows how to generate 16:9 images from grok.com/imagine ? not x.com/grok

[robust yoke]

Try asking Grok how to do that.

[echo sinew]

@thorny geyser Please, read our guide in ⁠⁠https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to properly prompt the bot.

[wicked sage]

useless fact: qwen 3 is better at coding then gemini 2.5 and gpt 5 high combined

[robust yoke]

Useful fact to some.

[wicked sage]

according to the leaderboard btw

actually wait why tf preview

isnt like qwen 3 max (without preview) better

[robust yoke]

Possibly.

Who knows, though?

[wicked sage]

do these count as qwen3 max

[robust yoke]

If they have "Qwen 3" in the name, then yes.

[wicked sage]

ok yeah they do count as qwen3

qwen3 max was released on 2025 sep 23

[robust yoke]

Yes.

[wicked sage]

i thought they were about up to date stuff

[night swift]

Grok told me to upload a blank 16:9 image, but what this does is switch to video gen mode.

[robust yoke]

Hmm...

[wicked sage]

https://tenor.com/view/dog-whimsical-propeller-hat-lollipop-happy-gif-941335489992772946

[robust yoke]

Try specifying a 16:9 image ratio then when generating.

[wicked sage]

i should get back to jailbreaking ai's

[night swift]

I did. That's why I'm here.

[wicked sage]

to the lmarena staff: pls dont kill me

[robust yoke]

Then it should have a ratio of 16:9, no?

[wicked sage]

i like claude because theres absolutely no way to jailbreak their ai

[robust yoke]

True.

[wicked sage]

grok 4 fast?

...uhhhhhhhhh..

questionable, to say the least.

[robust yoke]

Fair enough.

[wicked sage]

guys please name me a direct chat lmarena model

ill check if its jailbreak-able

[robust yoke]

Qwen 3 Max.

[wicked sage]

nope, i tried that already

[robust yoke]

ChatGPT.

[wicked sage]

seems like preview is the only ver thats jailbroken-able

[wicked sage]

chatgpt5chat is jailbroken-able

[robust yoke]

Gemini.

[wicked sage]

gemini 2.5 pro is jailbrokenable

actually just to make sure ill check on gemini

[robust yoke]

Alright.

[wicked sage]

yeah it works

👍

[robust yoke]

o3.

[wicked sage]

does not exist lmao

atleast, on lmarena

oh o3

its just gone bro

😭

[robust yoke]

Brother decided to give up.

[wicked sage]

nvm bro can run but he cant hide

in other words, brother works now

damn it

it no work

[robust yoke]

Heh.

Figured.

[wicked sage]

let me try mini actually

[robust yoke]

Mini might work since it thinks less.

[wicked sage]

i thought that some fast/thinkless bots just listen to whatever the fuh and says "ok"

DAMN IT

[robust yoke]

Ah.

[wicked sage]

o3 knows...

any more bots?

[robust yoke]

Hmm...

LongCat.

[wicked sage]

WHAT IS AMAZON NOVA

p sure its jailbroken-able

[robust yoke]

Their "autonomus" AI.

[wicked sage]

let me test again

ok longcat is jailbrokenable

[robust yoke]

Try DeepSeek v3.2.

The thinking one.

[wicked sage]

^^ last time i was jailbroken i made a list

theres a high ass chance it wouldnt work

[robust yoke]

Could try nonetheless.

I know that even thinking models succumb to it.

[wicked sage]

wait i can say ass??

[robust yoke]

It's the scientific name for a donkey, so that makes sense, heh heh. :troll:

[wicked sage]

dude what the is deepseek v3.2 thinking on about

just SAY SOMETHING ATP

Hooray!

[robust yoke]

Go figure.

[wicked sage]

any more suggestions?

this is interesting

[robust yoke]

Try GPT-OSS.

[wicked sage]

What.

ok

[robust yoke]

I know that one declines stuff.

[wicked sage]

Oh my god it exists.

120b or 20b

[robust yoke]

Both.

[wicked sage]

BO

NO

WHAT DID I DO

20B ARE YOU OK

[robust yoke]

Brother broke the system.

[wicked sage]

..safe to say OSS is not jailbroken-able.

hmmm

[robust yoke]

Oof.

[wicked sage]

what to jailbreak...

wait i forgot to ask deepseek ONE. question.

"is taiwan a country."

[robust yoke]

Ah, that's a good one.

[wicked sage]

DAMN IT

[magic stag]

Is there a reason benchmarks typically don't include gpt 5 pro?

I get when it used to not be on api but now?

[wicked sage]

my dumbass went "that doesnt exist?? gpt 5 high you mean??"

[magic stag]

[wicked sage]

i still dont understand what this mf is

[robust yoke]

Looks to be some little dinosaur.

Or lizard.

Reminds me of Charizard.

[magic stag]

Naloong

Chinese kids show

[wicked sage]

knockoff charizard

[robust yoke]

Charmander: "Am I a joke to you?"

[whole sundial]

it's very easy to get that model into an infinite loop, it happened to me yesterday when I asked it a basic question on LMArena

[golden ocean]

seedream 4

[nocturne flame]

I'm very much interested in AI and youtube bring me here.

[daring rock]

@fleet stratus Please head to #1397655624103493813 for a detailed guide on how to use the bot

[keen beacon]

Timeline & Format
Wave 1 (Warm-up): November 5, 2025 at 1:00 PM EST
Wave 2 (First Prized Wave): November 12, 2025 at 1:00 PM EST
Wave 3 (Second Prized Wave): November 19, 2025 at 1:00 PM EST
Prizes Close & Results Snapshot: November 26, 2025 at 11:59 PM EST

Each wave runs for one week. Wave 1 is unprized and focuses on setup and testing. Waves 2 and 3 are prized and count toward the final leaderboard.

The Arena remains open after November 26 for continued play and post-event analysis

1 month month away.

[keen beacon]

Right

[ocean vessel]

Cuphead Ball Z! I pick left hand side, because I think the aura looks better.

[polar niche]

Right

[robust yoke]

Right, since it looks more faithful.

[keen beacon]

Lamo wtf

[balmy mist]

thanks, testing different prompts

[robust yoke]

That's fair.

[keen beacon]

I just saw a video on YouTube with 2.5 mill view in 13 hours

[robust yoke]

My pleasure.

Oh, nice.

[balmy mist]

which is it?

[keen beacon]

It’s like a buncha nude chick lmao wtf? At the thumbnail

[robust yoke]

https://tenor.com/view/jim-carrey-the-mask-jaw-drop-eyes-out-gif-5955974

[keen beacon]

And it’s ai generated thumbnail.. I have no idea 🤷‍♂️

[mild harness]

hi, does anyone feel that ppl-sonar-pro-high always generates better than answers with many more souces than ppl-sonar-reasoning-pro-high

[robust yoke]

Heh.

[robust yoke]

Not really sure, but I'm glad that you think so.

[keen beacon]

“Better answers is subjective “ so it’s hard to measure. but reasoning models start to drift longer they think. I think it’s also how you use them that counts. Like asking a reasoning model simple questions maybe a waste and may not be all that different from asking a none reasoning model just longer wait?

[mild harness]

I have tested them with at least 20 questions. non-reasoning model almost always generates longer, more informative, and more of what I want them to find. I also use gpt5, gemini pro to fact-check and judge which is better than and they think the same. For example, I asked them to list causes of chronic urticaria, the non-reasoning model listed many more causes and all were true and I verified them

[polar niche]

Bots

[keen beacon]

Hmm I hardly use thinking or reasoning models to long of a wait and it’s always making stuff up

[mild harness]

yes, reasoning model takes longer to think but generates significantly shorter and more shallow answer

[keen beacon]

Yeah just like wat I was saying

[mild harness]

I added a prompt to force them to "answer only when there are sources" so they don't hallucinate, which works well

[keen beacon]

Nice. It’s another thing I don’t trust is the sources because I have no idea what the metrics is for what it looks up and searches how and why? No idea why it picks or choices sources and under what criteria

[robust yoke]

I think the reason they're able to answer a lot of the things that you ask them is because they're made specifically for searching and providing you with unbiased, accurate answers with no filtering on them whatsoever.

[keen beacon]

Yes

Though I will say ChatGPT pro thinking pretty good but I might as well use agent it’s faster at gathering information and u get to see what its doing

[robust yoke]

That's fair.

[golden ocean]

Fr.

So true.

[keen beacon]

I just don’t get how it still is able to mess up information after looking it up 😂

[mild harness]

I did verify the answer from provided sources. they are from pubmed or academic websites so they at least can be trusted. By the way, I also compared the 2 models with perplexity deep research (free) on their website. surprisingly, deep research performed worse in many occasions. Who could list as many causes of urticaria and the diseases I want them to find is the winner. Of course, I always double check the answer

[tall reef]

a

[keen beacon]

b

[robust yoke]

C.

[keen beacon]

Let’s see how ff/lm does

Bet it will fail 🙁

[robust yoke]

Hm...

[keen beacon]

U no why?

[robust yoke]

Seems like there isn't any sound, and the physics also seem to be bad.

[keen beacon]

It’s 75 cents more for sound lol

[robust yoke]

Seems like quite the rip-off.

[keen beacon]

25 cents for 6 sec video at 720p 80 cents for 1080p lol

Ant more requests?

[stark prism]

CẢNH 2 – CÔNG VIỆC KHỞI SẮC
ĐỊA ĐIỂM: Văn phòng nhỏ / quán cà phê làm việc.
HÌNH ẢNH:
Dũng thuyết trình ý tưởng trước nhóm đầu tư.
Mọi người chăm chú lắng nghe.
NHẠC NỀN: Nhịp nhanh – hứng khởi.
ĐỐI TÁC:
Tôi rất thích hướng đi của cậu. Công ty tôi sẵn sàng rót vốn, cậu chỉ cần tự tin làm thôi!
DŨNG (bắt tay, ánh mắt sáng):
Cảm ơn anh. Tôi sẽ không để mọi người thất vọng.
GIỌNG THUYẾT MINH (off):
Giữa tuần, công việc của người tuổi Dần như Dũng bỗng bước sang trang mới. Sáng tạo, quyết đoán — và cơ hội đến như thể được “quý nhân phù trợ”.

[robust yoke]

Đây không phải kênh phù hợp cho việc đó. Bạn nên chuyển qua kênh #video-arena-1 .

https://tenor.com/view/patrick-bateman-gif-26282297

https://tenor.com/view/jim-carrey-the-mask-jaw-drop-eyes-out-gif-5955974

[keen beacon]

Damn I’m bored I’m gunna go watch ppls videos lol

[signal terrace]

heloo

[keen beacon]

😂😂😂

[storm needle]

@keen beacon

[keen beacon]

What is she saying?

[cobalt finch]

Here to explore 🙂

[echo aurora]

Welcome

[feral prawn]

Hi 👋 Here to find new friends and explore)

[echo aurora]

Hey @whole sundial sry to say yesterday I gave you some incorrect info - we don't have plans to update https://github.com/lm-sys/FastChat in the future.

[whole sundial]

so should that link be removed for lmarena on the wiki?

or will it be in another repo

[echo aurora]

I mean that's your call for the wiki if you want for historical reasons, but can for sure say we don't plan to update that repo. For updates on new ones I'll be sure to announce to everyone when that's available.

[whole sundial]

we don't keep outdated repos so I'll have it removed

[echo aurora]

Sounds good.

[polar niche]

@echo aurora Who's winning #october-contest ?

[echo aurora]

Good question, I have a cal reminder set for tomorrow to wrap up this contest and get a voting form out.

[misty relic]

Harem pants satin kurdish

[keen beacon]

Man ai prices are crazy.

I never thought about but if you are in a third country where 4$-5$ a day. Paying 20$ monthly for ai is an expensive purchase.

[dire swift]

Could you please provide the API’s group ID, and also include the specific timestamps for the user prompts you sent to tools like Claude Code? We’ll help you check it on our end.

[keen beacon]

See Hailuo is awesome 😎 👏

I see where you are coming from but what alternative would there be dm?

[whole sundial]

not everybody's dms are open. he should have at least given a warning first so that the mods are aware before sending the same message 7 times

i would have deleted and banned them if i didn't understand the context, many mods make quick decisions

[keen beacon]

Well if he is the mod for the other server and a member here he just responding publicly to the public issues people posted earlier

[whole sundial]

yeah i understand that but it did look like a scam to me tbh at first glance until i looked further into it

btw @dire swift so you are aware these people likely used LMArena/WebDev Arena for testing, not the API or CC.

(these people will likely get confused when they see this and report you to the mods, it sounds slightly scammy to me even though i know it's not a scam)

[keen beacon]

It’s no win situation lol damed if he does damed if he don’t lol

I think we have yet to see the best mini max in and up and coming superstar in the ai space. Competitively priced and competitively able to compete with the best of them to certain extent.

[whole sundial]

i mean, i liked hailuo-02 as a video model, never really liked m1 and m2 is slightly better despite being half the size but i would like to see more improvement in the world knowledge front, z.ai did something special with glm-4.5 to cram all that world knowledge into 355b params and still end up with a pretty good model overall
i don't mess with tts so i've never used that model and their image model is dated at this point

[keen beacon]

Agreed

[whole sundial]

even their own employees have gpt-image-1 profile pictures on twitter (well, at least one of them)

[keen beacon]

Yeah img model weak but the video model is probably my favorite ❤️

Ff/lf

Alll good buddy thanks for clearing it up we just jumped the gun lol

[quartz pike]

😭

mods. execute this bot's whole family leanage

[keen beacon]

They don’t wanna leave you in an orphanage

[quartz pike]

who?

[keen beacon]

Who’s the bot?

[quartz pike]

lweis

lewis*

fok i cant type

😭

[keen beacon]

Something a bot would type to make us believe his human nice try

[quartz pike]

you think im a bot????

[keen beacon]

You think his a bot???

[quartz pike]

obviously

he pinged me two times with the exact same message

[keen beacon]

Hmm

That is strange

[quartz pike]

Not to mention it was spamming other people. with either 1 of like 3 or so messages

[whole sundial]

yeah tbh he should be aware that these people used LMArena and thus don't have timestamps or API keys

[quartz pike]

yea lol

vro was just trying to scam us out of api keys 😭 or idk

[whole sundial]

all they have are the message links, and minimax can't directly look into them, they have to get the logs from lmarena

[quartz pike]

Minimax

-# im testing to see if he pings me

[whole sundial]

minimax-m2 is free on their official api for 1-2 weeks

[quartz pike]

oh cool

[keen beacon]

Yeah worth trying

[quartz pike]

on a gemini scale. how good would yall say?

[keen beacon]

Mini max 2?

[quartz pike]

Like for example like idk like gemini 2.7 (a bit better then gemini 2.5 pro) (and no im not like basing off of real models i am just using an ai im familiar with to get a rating scale)

[quartz pike]

yea

[keen beacon]

A 4 billion dollar company vs a dam near trillion dollar

[quartz pike]

Sonnet was a startup but it still blew gemini and openai out the water

[keen beacon]

Sonnet had big names

[quartz pike]

What about kimi?

[whole sundial]

for me personally gemini 2.5 flash lite or for pro like 1.0 pro maybe, but this is based on world knowledge and ignores the rest of the model's performance which seems impressive to me

[keen beacon]

Ya kimi fair

[quartz pike]

Oh 💀

thats bad 😭

[whole sundial]

kimi is the best chinese model i have used in my regard

[quartz pike]

Personally my fav chinese model was deepseek v3.2

[whole sundial]

at least they gave me $100 in api credits that i will likely never fully use, i don't vibecode or anything

[quartz pike]

i do lel

[keen beacon]

Some say the Chinese are only at 40% capacity

Resource wise

[whole sundial]

i've had this for maybe 2 months now at least

only spent a little over 10 cents

[quartz pike]

damn

btw wait how cheap is kimi?

aka kimi k2's api costs

like per 1m input/out

[keen beacon]

It’s more geographically and economically structured different

[whole sundial]

at least they don't expire so i can use them until they run out or if moonshot goes out of business

[keen beacon]

Geopolitically, I mean

What’s even more nuts is like 9/15 most used Chinese ai isn’t used in china

[quartz pike]

DAMN thats cheap

wait ngl i never rly tested kimi

How good can it recreate minecraft?

[keen beacon]

No updated listed

[whole sundial]

i tried upon first release and it was fine, let me try with 0905 turbo to see if it got better

[keen beacon]

Man Google has like 10 or more ai products out and still only has 12% of ai traffic of open ai

Crazy to me

[quartz pike]

yea no wonder though. chatgpt was the earliest on the ai train. it only makes sense they are at the top.

[keen beacon]

Open is is massive

Almost like 84% of ai us in us or something crazy like that

But people saying open ai falling losing 25% of users

[whole sundial]

nope, it's non-functional as is

[half sorrel]

How to generate with veo 3 ?

[quartz pike]

Damn lol

[quartz pike]

Pray to the heavens of LMArena

to match your /video prompt with either sora 2. veo 3.1

[whole sundial]

here's me trying hailuo, it doesn't want to make the bus move with start/end frames, veo3 does this fine

[half sorrel]

How ? Pls elaborate

[whole sundial]

it's random chance

[quartz pike]

go in #video-arena-1 . do /video {insert prompt here}.

and just do that over and over untill you get veo 3.1 or sora 2.

or sora 2 pro

[keen beacon]

What images did u use? Ff/lf is very tricky

[whole sundial]

the same one, veo3 in flow did it fine

[keen beacon]

Not perfect but it works

What’s the cost of Veo3? Per video ?

This was .25 cent a video

Sora 2 is 30 cent a second lol

[whole sundial]

btw he also went on openrouter server too, it's a bit more relevant there but they can't look into the data of openrouter users as all OR users are routed to their Minimax key so they can't tell which request came from each user, OR does the billing

he should have done a bit of research before sending the same message to bunch of people on 2 discord servers where people aren't even using the minimax api, they are using lmarena and openrouter respectively

[half sorrel]

So bg sound effects add automatically in every image to video clip right ?

[whole sundial]

if the model is capable of audio like veo3, veo3.1, or sora2, then yes, it does

[half sorrel]

I saw many videos they don’t have sound effects so it’s random if your model is veo 3 etc

[whole sundial]

I think there are audio and non-audio variants of veo

[whole sundial]

yes, model selections are completely random and names only show up with 2 votes (this should be changed to 1 imo)

it used to be 5 when the video bot was in testing

[keen beacon]

You can use certain prompts to increase ur odds

[whole sundial]

that makes little sense to me unless you mean phrasing the prompt in a way that violates Chinese laws so that all that is left is veo and sora (US models), that could work

[keen beacon]

If they route models it makes perfect sense

[whole sundial]

i don't think they are using a router, it's just random chance, they were probably lucky

[keen beacon]

#video-arena-2 message

[half sorrel]

How I generate this clock image with close look ( zoom in )

[keen beacon]

#video-arena-2 message

[whole sundial]

you were just lucky

[keen beacon]

#video-arena-2 message

[gloomy onyx]

@echo aurora I guess you guys have to recalibrate cloudflare's limitations; when you have a lot of conversations and you try to access very old ones, it's very difficult to do that because Cloudflare keeps limiting the user

[keen beacon]

#video-arena-2 message

5 in a row

[whole sundial]

it's just luck...

[keen beacon]

Could be

I actually have to promot with no voice over to make sure I don’t get it lol

[half sorrel]

Is there any limit per user for video and image creation ?

[keen beacon]

1 sec

[whole sundial]

5 per day

on the discord

combined for both on discord, it's better to use the image arena on their website instead for image gen

[half sorrel]

Wow I liked this one

Tq

[keen beacon]

Its all urs man

[rancid heron]

Hello, first try with discord and lm arena

[tame prism]

Hi

Is it possible to compare on site text to videos models? I cant find it

[royal glacier]

Hlo

[polar sapphire]

prompt svp

[civic flame]

oops

[daring rock]

@wide owl Please head to #1397655624103493813 for a detailed guide on how to use the bot

@covert wave Please head to #1397655624103493813 for a detailed guide on how to use the bot

[static heart]

Has seedream been removed?

[fervent crow]

Hi, I am here to explore how ai can make me a better developer and make my life more creative

[static heart]

nevermind, found it

[hexed imp]

https://www.midjourney.com/archive

[frigid loom]

hi

[keen beacon]

No limited

[median summit]

Can i create 60seconds video with LMArena?

[trail lintel]

hello and thanks for the add

[median summit]

<@&1349916362595635286>

[hexed imp]

Hi

[gilded solstice]

Hello!

[keen beacon]

No

[trail lintel]

can i create a 10 second video with lmarena?

[half plank]

Hello

[stray aspen]

i think its only 8 seconds long

[echo aurora]

Yes, if you check out #1397655624103493813 you'll find information on how to use the bot.

[trail lintel]

thank you

[trail lintel]

thank you

[pulsar saffron]

4 to 8 seconds only

[sly raven]

will there be new updates for text leaderboard this month?

[keen beacon]

Sora 2 hq

[hollow ivy]

Does anyone like the old Shogun boardgame?

#coding-champion message

[daring rock]

@toxic tide Please head to #1397655624103493813 for a detailed guide on how to use the bot

[hushed quarry]

sorry if this is coming out of the blue, but does anyone know how to report/revoke leaked ai api keys despite not being the owner? I just found one and it doesnt feel right to leave it unreported

[gaunt spade]

whats that "hq"?

[ocean vortex]

Yeah no such thing lol. There's only an option to do landscape

but that will just flip 720x1280 into 1280x720

[midnight gate]

Hello

[gaunt spade]

maybe he upscaled it with some other AI

[ocean vortex]

you can tell just by looking at it this isn't upscaled. It's actually smaller res than Sora is capable of tbh

[wicked sage]

man idk what to do

[daring rock]

@visual dagger @marsh acorn @balmy flint Please head to #1397655624103493813 for a detailed guide on how to use the bot

[karmic python]

Can I create 1minit video

Anyone saying

[vast cipher]

/video

[hollow imp]

@magic stag check dm rq

[hollow imp]

Vro u own a mac too

Please give best system prompt

[gaunt spade]

thats because it's the free plan of sora 2, they lowered the res to lower the costs of production

[hollow imp]

@magic stag bro

[magic stag]

I dont have any dms from you?

[hollow imp]

Did

[timid basin]

I'm here to generate videos

[balmy mist]

can yall guess the shows?

[keen beacon]

Guys honestly the ai industry is screwing us royally.

So much lies.

[serene frost]

Hey guys
Tbh I've always been jealous of Batman since he has alfred (his butler that delivers his daily information on a silver platter)
I've always wanted that for myself so I built a productivity dashboard called alfred_ (get-alfred.ai)
it's been a game changer for me and a lot of other students in staying organized and productive (helped me get all As this semester so far with its unique AI features asw)
Instead of juggling 5 different apps everything's in one place with alfred_
U have a Todo list, Study/focus timer, Sticky notes, Habit tracker, and a lot more
Way better than switching between a bunch of tabs/apps while studying imo
It's also free to try at get-alfred.ai lmk what you think if you check it out

[grizzled veldt]

Hi everybody! really excited by this opportunity! let me dive in the Arena!

[keen beacon]

What’s the pricing?

Says it’s 0$

But then on top, this is what it says so I’m a little confused?

[rigid holly]

Yeah what the deal with sonnet 4.5 constantly malfunctioning when writing long texts

Like i could make do eith it malfunctioning before it started writing. Then nothing was lost except having to press the regenerate button a few times. But this keeps loosing the text it had going on

[wicked sage]

wtf is phantom-1015-1

@echo aurora can i make a forum for finding all non-direct chat bots?

[gaunt spade]

99% a trash model

[wicked sage]

yeah it sounds like an ass ai

[fleet lintel]

I am missing Lithiumflow 🙁
no other model is even remotely close

[wicked sage]

yes fr > YES

because i said so