general

[upper venture]

If you look at Kling, Hailuo or Veo, the cost per generation is somewhat high. Some might give you a trial of free generations like Google recently did with Veo 3. Like Hailuo or Kling does when you first register an account. But none are unlimited free like Wan is, at least at the moment

"First person view Flying through the skies Clouds are pinks"
It made the clouds pinkish and the camera rotates but given my very simple with no details prompt that's fine. I expected the clouds to be more pink like very heavy pink but all in all fine. Took 4 mins, slow for a 5 sec video but free is free I guess

[proud hazel]

nice, i like it

[upper venture]

Holy moly. It's neither. Wan 2.2 is 1080p

Incredible.

[proud hazel]

wow 1080p and for free, thats insane

[upper venture]

It generates sound too? For that video it was some calm music, I wonder if the prompt was different it would generate according to the prompt

[upper venture]

Yeah! That's why I love it!

No registration even needed (tho the generation will be slower for guests, I saw that detail)

[proud hazel]

At wan.video, right?

[upper venture]

Yup that one

I'll play with it more now lol, I'm interested

Have a great day, I'm out

[proud hazel]

It is asking me for registration

[upper venture]

Just close that

You can close it

[proud hazel]

When I click in the text field to write something, the registration pop-up appears.

every time

[upper venture]

That's weird, Istg it used to be able to generate without registration before

I guess it's for the guardrails and spam now that it's getting way more popular since the recent launch of wan 2.2

[proud hazel]

But you still are generating without registration?

[upper venture]

No, I used Google account because I wanted to save my generation history

[proud hazel]

Alright

[upper venture]

It's free, it gives you credits daily for the priority generation but even without credits it still generates well and takes a few minutes but it's fun to play with

[proud hazel]

yeah

[upper venture]

"A race going between two cars. Building lights. Night scenery."
Seems like the audio isn't always music and actually might be sound depending on the output, kinda cool.

About 7 minutes waiting on slow generation without credits.

[quartz pike]

https://chat.nousresearch.com this is genuenly fire

they should add hermes 4 to lmarena

[rotund sigil]

Hi everyone! Happy to join this community 🙂

[tribal aspen]

Hi people

[rustic knot]

hi gork

[dusky pier]

Is MAI-1 trash

[pine knoll]

what qwen has audio???

[rustic knot]

that's wan bruh

[stray aspen]

Yeah but it sucks

It's.mostly.music

[sand needle]

Hello people

[novel umbra]

how many videos i had left??

[zinc ore]

🤔 me wondering if Gemini 3 will be a September drop or an October drop

[ocean vortex]

Ok so... looks like gpt-oss beats grok-code-fast-1 on Aider which apparently can only do ~40%... huh

[rustic knot]

most ppl care more about real-world coding tasks especially for these coders

[ocean vortex]

I hope AA does some testing on it. But thus far it is not looking too great. Model is living on dreams and hopes lol

Someone should perhaps verify that SWE claim too

[rustic knot]

AA doesn't even test swe or web dev

[ocean vortex]

True but they do SciCode and IFBench which is useful here. Whole thing should give a rough idea of the general intelligence of the model as well

[rustic knot]

for specialized coders, noone cares about the other things that are usually tested on models like gpqa

diamond

[ocean vortex]

which is why I didn't mention it lol

[rustic knot]

yeah but the AA score would be rather meaningless cuz it includes it

[ocean vortex]

It's a nice bonus to know additionally. You can't completely isolate coding from intelligence. Sometimes it needs to be able to understand the task and do some decision making too tbf

[high aurora]

Hey, looking for a top performing video generation model and a prompt template, can anyone help? Please!

[pine knoll]

hai

is there any free web AI video generator beside qwen AI?

[pine knoll]

top: veo 3

[high aurora]

any prompt template you follow for best output?

[pine knoll]

idk, sorry i just always use simple prompt

[high aurora]

cool, thanks!

[jade egret]

is gemini 3 pro gonna be smarter than 2.5 pro deep think

[pine knoll]

nah

[rustic knot]

no

[pine knoll]

it just like rtx 2070 vs 3060 xd

[rustic knot]

H100 vs H800

[jade egret]

nano banana (:

[tribal aspen]

using the image mode still uses Imagen unless you attach another image

[wraith smelt]

Heyo, so excited for everyone to think about visulization mathmatically. Just like Donald Duck in Matthmaticaland

[dense sphinx]

Mai-1 preview what model is this?

Is anyone have answer for that?

[neon edge]

Hello guys , LMArena brings me to here .Thanks.

[dense sphinx]

Greetings Kelvin

[hollow chasm]

Scale 1/10 how good would you say is the translation from Arabic to English? Thank you

[balmy mist]

7

[keen beacon]

The founder of this follows me in Twitter 😄

[nova cargo]

hi

[dark fiber]

GM

[chrome hornet]

hello

is ther a way that the generated video will be specific for veo 3 or the likes?

[echo aurora]

nope, it's going to be Battle mode only (two random models head to head)

[chrome hornet]

ok. thanks for that

so we just repeat it until the two prefereed model is competing

[thorny anchor]

😊

[fickle galleon]

Hello

[rancid plaza]

hello, everyone

[verbal nimbus]

Interesting, what do you like about it?

[sacred fjord]

hi everyone, i m here to test videoarena^^

[sly zephyr]

😎 hello

[heady basalt]

@everyone suggest me best ai in lmarena for generating essays

[wispy gazelle]

hello

[heady basalt]

plz reply

[remote spire]

Can anyone send me the link to use gpt image 1 model in lmarena ai please 🙏🏻

[heady basalt]

i said bro

[remote spire]

Okay but it's not working

[hardy lion]

https://lmarena.ai/?mode=direct&chat-modality=image
Then in the dropdown menu at the top select gpt-image-1

[remote spire]

Okay thanks let me try

It's opening nano Banana model

Not gpt 1 image model

[heady basalt]

thanks for increasing my knowledge

[hardy lion]

no problem, thank you for using lmarena

[heady basalt]

@hardy lion bro can you suggest me best model for generating essays

[remote spire]

Gpt 1 image image generation isn't as good as nano Banana model

[heady basalt]

@hardy lion my son needs it

[hardy lion]

Well gemini-2.5-pro ranks number 1 fr both overall text and for greative writing so that's a good place to start https://lmarena.ai/leaderboard/text

Other strong models are gpt-5-high, and claude-opus-4-1

[remote spire]

Yes check leaderboard @heady basalt

But gemini 2.5 pro is best

[hardy lion]

but what's better is to use battle mode, it will select 2 random models, have your son read both and pick which is better himself. He may also learn more in the process https://lmarena.ai/

[heady basalt]

thx

u both

[remote spire]

You're welcome yes as Clayton said use battle mode and see what suits your needs

[heady basalt]

i want to build a pc under 200000$ for my son suggest me specs he want to run llms

[hardy lion]

which country currency?

[heady basalt]

us dollar, freedom unit

[hardy lion]

ur gonna spend 200,000 us dollars on a computer for your son?

will you please adopt me

[heady basalt]

i m thinking of that asus rog zephyrus g14 (2025) will be good

better than pc

[heady basalt]

yes

[heady basalt]

dont make me feel rich

i just saved it in 2 years

i m microsoft's engeeneer

[heady basalt]

sorry very much it was $2000

@hardy lion

[pine kraken]

-# ||🧠||

[quartz light]

ts is so ass wdym

[heady basalt]

but my sun dont have it

*son

[quartz light]

whats the essay about

[heady basalt]

सिंदूर ऑपरेशन नंतर सैनिकाचे गावात स्वागत

he has taken grade language as marathi

[quartz light]

hmm

use gpt 5 high

get it to write the essay in english, then ask to translate to marathi

you can also use gemini 2.5 pro

[heady basalt]

i did it

[vocal cedar]

Guys, I have got a question.

Normally I use gemini for imagen. I am unable to generate celebrity faces there like here is an example of a prompt
"Photorealistic, 8K cinematic still of Tom Holland as Spider-Man, his iconic red and blue suit a vibrant slash against the cityscape, swinging majestically between two glass skyscrapers."
It refuses to generate the image. Even on google ai studio, same result. But when I use LM ARENA WEBSITE , It somehow generates the result using imagen 4, and even imagen 4 ultra. How?

Can please someone help me?

[waxen wing]

Hi! Would it be possible to enable thread creation in the #video-arena channels? It would help keep each person’s video discussion organized instead of getting lost in the chat. The setting is in Server Settings → Roles → Permissions → Threads (enable “Create Public/Private Threads” for members). Thanks! 🙏

[vocal cedar]

@hybrid swift hey man, could u help with my query?

[minor river]

Hello guys! Nice to meet you all

[simple wave]

hello

[novel solar]

Hello

[fast meteor]

hey i have installed model here how to run it ?

[atomic bough]

Hello

[hardy walrus]

hello

[white hatch]

AI needs to be more confident in its answers

[solemn rock]

Hi! I'm here to learn and use Ai

[indigo schooner]

Hello

[timber kiln]

Who knows some nice credit deals for APIs

[spare comet]

helo

[hollow imp]

Use grok

[torpid wolf]

hey there

[dusty niche]

as i know the positive prompt is the things that you want your Your generated image should contain

[humble gale]

Hello!!!!

[hardy walrus]

how many seconds generate per video thanks

[heady ember]

10

Or less

[lapis knoll]

Hello everyone

[heady basalt]

floorp vs vivaldi

@everyone

[keen beacon]

why would you ping people over a minuscule question?

[heady basalt]

sorry

whats ur ans

[strong chasm]

hiya

[inner gate]

What I’ve realised is that sometimes it does block it yes but sometimes it goes through if u keep trying. Maybe on a new chat

[latent crest]

You guys what’s ur fav tool for creating voices ?

[humble sonnet]

What is the error on the imported image? I just have it without knowing what it is exactly.

[echo seal]

Hello everyone!

[dense sphinx]

Claude Opus 4-1 number one.

Even with limits but fair

Hello paws

Seriously just Kaiwan is fine

@hollow ivy nope

What is this?

What this model does?

I saw only of Pro Grounding in search

[quartz light]

hi goobers

[dense sphinx]

@hollow ivy where exp grounding?

[terse mural]

Iam an interested in AI novice

[lunar niche]

Helllo

[timid goblet]

hello

[safe pivot]

Hello there
I got problem
I can't get any generated pictures

👋

[abstract void]

💀

[fathom chasm]

https://www.zazzle.com/fearless_destiny_motivational_design_t_shirt-256243546790365585

[sacred tulip]

/image-to-video

[keen beacon]

Please read #1397655624103493813

[heady ember]

I think you need better phone/pc since it can't render images idk

Model is correct gemini 2.5 flash yes that's correct model don't worry about that

[amber glacier]

Hello here to try video for the future

[inland ridge]

Hi Everyone

May the best platform win. Hawu

[safe pivot]

I tried in both phone and pc and same problem

[slate dagger]

Hi my people.

[sullen sand]

what is going on riht now it always says error

[fiery gull]

Where qwen 3 max in direct chat? 😭

[cloud sandal]

Hi Fams

[leaden sun]

the new update has ruined opus, it's much less creative, sounds more cold, corporate and clinical (they profile and monitor your mental health with a hidden prompt injection now)

[swift badge]

When i attach a photo the video prompt doesn't work on it. Why?

[exotic bolt]

hi

[echo aurora]

Are you running into an error?

[astral terrace]

Hello

[slate rune]

Hello everyone

[echo aurora]

[fading summit]

Hey) What's going on? All claude models in my chat are dead...

[latent crest]

Hey guys , what prompt can I use to :

1. not blur the background, I freaking hate it

2. make it look like an amateur and not professional photo

?

[heady ember]

Just tell "remove blur" or "not blurred"

For less professional go like
style: casual amateur snapshot
camera: taken with old smartphone,
cheap phone quality
lighting: harsh flash, yellow indoor lighting, unflattering shadows
composition: off-centered, slightly
crooked angle, cut-off at the top image quality: grainy, low resolution, slightly blurry

[finite gale]

Ahoy!

[fading summit]

Hey)

[latent crest]

Thank you very much I’ll try, when i say SHARP BACKGROUND the ai goes like NOPE lol

[jade egret]

what the best ai agent

like the gpt agent thingy or project mariner

stuff liek those

[fiery gull]

what is the better version of gtp oss 20b I can run in my phone? (7 plus gen 2 16gb ram)

[nimble raven]

hi

[split pilot]

Hey ya

[cedar stratus]

Hi! new here. it's been more than 10min already and i haven't receive an output. Is it normal?

[fading summit]

Depends on a promt, but i am not sure that it is ok

Meh, i can't continue my current chat, the error pops up. Thats sad

Does anyone know how to fix it?

[true sierra]

Hi

[fading summit]

Hey)

[keen beacon]

#bot-feedback

[solar dagger]

hello

[cedar stratus]

still no video

🙁

[pine knoll]

guys

is there any AI generator for creating Animated Illustration

i tried AI video generator like veo 3 or every popular one, it always makes character moving around, which i don't want to

just simply make live illustration

[tiny oxide]

hi

[minor crow]

hi

[whole wagon]

openAI new model training cutoff October 2023

They are literally going backwards

[heady ember]

Gemini ate them, but i got used to ChatGPT i like the android app

It's minimalistic and stuff

[torn mantle]

i dont think its from oai

could be trained on oai models

doesnt feel like one

[dawn grove]

Is it possible to share lmarena chats?

[jade star]

Yo, is there anyone who has paid Claude? Do y'all get access to Opus 4 on the 20$ plan? Are the limits still the same as sonnet? (50 requests per 5 hours)

[paper oxide]

Hey, everyone

[echo aurora]

Hey everyone User Login just dropped! Check #announcements message for more information!

[keen beacon]

whaaaaat

[echo aurora]

yee!

[keen beacon]

https://tenor.com/view/rude-gif-2074021595659835685

[dapper acorn]

Hello~

[ocean vortex]

This seems wrong, no way it's 2023 cutoff

I fully expect 2025 for anything new that comes from them, worst case 2024Q4

unless it's like gpt4.5 with new fine-tuning...? I thought they are deprecating it though

So yeah either it's gpt4.5 or this is just missing a correct system prompt

they may be exploring a possibility of making gpt4.5 into reasoning though that would be insane and probably not the smartest idea...

[keen beacon]

They probably fine tuned some past generation model for webdev?

[ocean vortex]

Why would they do that though...?

gpt5-high is already leading

[keen beacon]

Hype.

[ocean vortex]

By cannibalizing the model they are trying to sell?

[fiery gull]

They didn't do that because he consumed and gpu too much, imagine with thinking

[ocean vortex]

with some model that gpt5 replaced?

makes no sense lol

[keen beacon]

No, it isn't. Look at the market share of coding models, Anthropic is still at the top. Most normies find GPT-4o better. Despite gpt-5 is great they failed to market it properly, so this model is only good for those who are able to see

and to pay for API

[ocean vortex]

They didn't do that because it's not feasible outside of experimenting stage. By the time you can make halfway decent reasoning model out of it you can release at least 2 generations of gpt5 sized models which WILL perform better

And then you have the cost on top

[keen beacon]

because OpenAI idiots decided that it is a great idea to put literally the most powerful version of their model not right in the chat window holy macaroni what the hell of a pr disaster it is

[formal jungle]

I hope this is okay to ask but please can I get another vote on this, I'd love to know one of the video models here
#video-arena-3 message
🙏

[ocean vortex]

bruh what

we are talking about this leaderboard

[keen beacon]

ok im stupid

sorry

[ocean vortex]

it's ok lol

Have you started with an empty context?

They have a very suboptimal way of context management now

So that messages from past battles will get passed as input for another battle and other models that have nothing in common with previous outputs

So if you make it introduce itself, then start another battle without explicitly starting new chat (the default easy way), for this new battle the false identity will kinda be forced on different model as a fake ai assistant response in the context. They are populating context with things a certain model would have never said, which then in turn DO affect the real responses.

odd, maybe an interface bug 🤔

[verbal nimbus]

Interesting

[ocean vortex]

ofc gpt may also just give out false response, although it would be extremely unusual for it to hallucinate it is gemini for no reason

[verbal nimbus]

That is definitely not GPT-5-High lol

It doesn't write Markdown bold/italics

Unless they gave it Gemini's system prompt (AI Studio), which has bold/italics iirc

[ocean vortex]

Yeah looking back at the response, no way this is gpt5...

[verbal nimbus]

Does this happen in direct chat?

[ocean vortex]

Maybe that's how 2.5Pro got the first place back.... 2.5Pro responses are counted as gpt5-high

[verbal nimbus]

The labels might be bugged

[ocean vortex]

(I'm half joking 👀 )

[verbal nimbus]

Maybe post it in #1343291835845578853 ?

[subtle rivet]

@hollow ivy I think you just hit a slow time

[verbal nimbus]

Or the other way around lol

[subtle rivet]

oops meant @verbal nimbus

[verbal nimbus]

For?

[subtle rivet]

this chat normally scrolls fairly fast, seems to have died down a bit today

#laborday weekend, most likely a number of folks are out

[runic star]

creating ai videos for marketing impact

[ocean vortex]

w

h

y

huh

I don't think that's likely tbh

The only time I've seen a traffic of this magnitude on AI server was when chatgpt absolutely exploded

on OpenAI server

but nowadays even they do not really need that

I don't think so, you would need a lot of people writing all the time...

I suppose small cooldown of like 5sec wouldn't hurt

but it wouldn't change anything either

It's typically more than 5sec to write a message

So more of a spam prevention, the traffic has to be insane with 1M+ of users for it to do anything else

yeah there are a few active servers like that where they need several main chatting channels just to keep it usable lol

But those are not AI typically

yeah report it

@echo aurora

[safe iron]

Hi everyone! Happy to be here and have discovered this nice place!

[hazy gazelle]

Hello to everyone. I am excited to stumble onto this place hope you find my creations enjoyable!

[loud sky]

would be nice to update Mistral's logo. The one on LMArena is one that's been outdated for like half a year now, they have this one now

[echo aurora]

Can you create a post in #1343291835845578853 instead?

[mellow timber]

Hello to everyone , here to take a look at this new technology !

[fiery gull]

The IA?

Or something else that I don't even know what is like rma?

[neon idol]

I think

[fiery gull]

Maybe the guy was talking about something more advanced 😬

[neon idol]

Ai that can take control of your mind?

Robot that can do Homework?

That my life is a video game?

[fiery gull]

Thinking about it I'm already letting flash 2.5 rule my life (I'm not good myself)

[neon idol]

Bro 2.5 is insanely powerfull

https://tenor.com/vSWnFnk9xBw.gif

Hell nah

[fiery gull]

He told me to uninstall tiktok and I came to interact with h-humans 😭

[neon idol]

Gemini acting like mother 💀🥀

[fiery gull]

What am I going to do? If I continued the way I was it would be F

[neon idol]

Wrong place! I am an AI tool without restrictions. Lets speak about everything! So, what do you want to share with me kn this beautiful day?

[fiery gull]

I (no kidding) I think half of the people on discord are bots, maybe you are? Maybe I am? Both?

[neon idol]

We dont know becuase we are in Matrix

Life is matrix

Me is in matrix

You are in matrix

[fiery gull]

If I was in the matrix I'd be making millions of dollars and dodging bullets in slow motion

[neon idol]

Me and you acting like nerd without a bit of human interaction 🥀🥀

[fiery gull]

🥀

[neon idol]

I think that life have to sybau us 🥀

[fiery gull]

I don't even know what to say anymore, I'm going to annoy someone else on discord because the flash sent me 😈🙏

[neon idol]

Muahahahahah

Top 1 evil plan

[fiery gull]

I'm an anti villain from the new marvel movie

[vestal vector]

hello everyone

[dreamy cove]

Hello guys!

[next turret]

Hi I'm Kyrian promise from Nigeria

[lilac nimbus]

neptune v4 coming Claude ready

[exotic inlet]

Hi

[alpine moon]

Ksn

[normal hedge]

Hi joining force here

[night kiln]

Hi everyone! Happy to be here!

[tropic plover]

Hi, I'm new here to learn more about this tool! Good luck!

[junior pecan]

learn AI

[fading summit]

Hey) I've been trying to create an ai agent using claude, i even got api key, but claude keep rejecting my messages. Should i activate this key somehow?

[thorn apex]

Hey here to explore more AI tools

[patent marsh]

See how it works

[foggy roost]

@echo aurora I saw the announcement about the gradual rollout of Google sign‑in. After the full rollout, will lmarena.ai still be accessible without logging in , or will sign‑in be required? Thanks.

Love the Google sign‑in for seamless cross‑device sync. For privacy and quick access, could you keep login optional—e.g., a “Sign in” or “Continue as guest” flow—with the ability to merge guest chats into an account later? Thanks!

[echo aurora]

It's hard for me to say for sure as things can change, but we're looking at both options. I don't think requiring it is going to be in place, but I don't want to get ahead of those decisions.

[sturdy mica]

is GPT-5-search GPT-5 High with search

or is it like

normal

[sturdy mica]

?????

[willow grail]

banana

[shy blade]

hello

[formal cosmos]

hello

[leaden sun]

I'd second this request too @echo aurora fingers crossed for the decision team 🤞

[echo aurora]

sounds good

[lusty tartan]

hello

[trail wigeon]

hi, Im the new guy 😛 anyone have an easy way to understanding temperature nad top-p in nano banana? i unsure how to set that part for image generation.

[limber grotto]

lmarena

[tribal aspen]

does LMarena has intentional restrictions on nano banana images?

or does it gets carried from the API?

The "Something went wrong" error when it senses somethjing wrong in the image

[kindred flume]

Hi

[boreal flower]

Hello from Thailand

[mellow dust]

does LMarena has daily limitation for i2v generation?

[echo aurora]

hello - sry to say we don't have those settings controls available, this is something we're putting effort into though

[echo aurora]

yes, 8 per day

[heady basalt]

hi

[surreal schooner]

hello

[clever mason]

hi

[heady basalt]

@heady basaltbvh

jjjjjjjjjjjjjj'

j

[little oyster]

hello

[sleek panther]

Heyy guys, I'm new here because of the nano banana

[olive flax]

Is there any way to share files like pdf on LMarena?

[mortal geyser]

What up? guys new here just testing the waters!!!!

[deft osprey]

halo semua salam kenal

[livid cedar]

hi everyone

[remote arrow]

Hi, everyone. I've been using AI for a while, and getting tired to see ads, telling people which one is the best. I want to know the real fact by comparing the results myself.

[agile kernel]

Yoo

[sleek crystal]

hey

[obsidian shell]

@echo aurora will lmarena be login only in the future ?

[heady ember]

It's nano banana

There's also flux kontext pro for some stuff but 🍌 is best

[inner lodge]

hi there

[lyric nymph]

hi there

[keen canyon]

Hello, I want to learn more 🙂

[fleet lintel]

OAI must be releasing something soon.. other 🍌 might cause too many folks to move to Gemini.

Any clue what they might release next to create some buzz?

[alpine coral]

tahoe is interesting.. seems kinda solid

[neon idol]

Dead chat

[dense kelp]

How to get photo in Discord what i geranate

[neon idol]

What?

For creating inages you have to go in lmarena.ai

[dense kelp]

He said like this

[neon idol]

You have to click the button that looks like a landscape

[dense kelp]

Opps i'll try t

[neon idol]

Battle the model is random, side by side you choose 2 models and direct chat you choose only 1 model

[dense kelp]

Bro i don't get this

[left iris]

Hello ! Can anybody recommend the best ai video generator if i would like to create more rated r content in terms of like a monster attacking people in an action horror sequence. Usually if you write it to the prompt about creature attacking somebody it doesnt want to generate it only when it would destroy or attack houses for example it works.

[hoary locust]

Hardly any

[twin brook]

Video

[glass arch]

hello users

I need a quick "expert" opinion

should I switch to gemini's subscription instead of chatgpt's subscription?

[heavy totem]

/A lush, verdant garden showcasing the high-value and innovative potential of horticulture, with thriving crops, a dynamic composition, and a focus on the key benefits highlighted in the content.

how many image,that i can generated?

[errant dome]

hello

[glass arch]

really? lmarena ranks gemini 2.5 pro higher than chatgpt in all areas pretty much

GPT-5 high is api only I think

unless it's gpt5 thinking

[glacial mulch]

lmarena is a measure of how good a model can please users, not how good a model is

[glass arch]

gemini has a bigger context window too

are the usage limits too small

[glacial mulch]

gpt5 hallucinates less for my discord bot projects

[tall summit]

i like gemini 2.5

[glacial mulch]

i think gpt5 is better, although by a small margin

[tall summit]

you can tell whether a model is thinking or not

easily

[vestal bone]

Is there anyone seeking for a developer?

[gentle lynx]

Hi

[leaden sun]

I've observed that once, in the middle of converstaion in claude.ai the responses got cancelled and model got swapped with sonnet 3.5, seems like server overload problem but this is...quite a shady practice from Anthropic...

it seems to happen frequently with CC users

[glass arch]

I think you're right. After some research, I see that gpt5 is currently best for what I want.

[glacial mulch]

cant really do an accurate compare when temperature isnt 0

[vivid void]

I just want to generate a pic not a video , How do I do?

thanks

[rapid mountain]

Hello 👋🏼

[nocturne ginkgo]

hello every nice to be here

[outer dagger]

hello

[rustic knot]

hi

[hexed sluice]

HI, I WANT TO GENERATE VIDEO

[hushed terrace]

Hi! Please check https://discordapp.com/channels/1340554757349179412/1397655624103493813

[mortal coyote]

what is this error , it does not go away - i have to reload everytime to generate new image

[quaint badge]

/imagine Create a banner measuring 5x20 cm to garner a literary text; text is called "Fadensonnen/threadsuns and sails". Create a banner that abstractly deals with the title but do not include any text.

[hushed terrace]

https://discordapp.com/channels/1340554757349179412/1397655624103493813

[gentle lynx]

How much time is able to create a video?

[sharp flame]

hello

[hollow wedge]

aistudio takes 70 seconds to generate a prompt sometimes

[vocal dew]

Hello

[echo aurora]

It's hard for me to say so I wouldn't want to get ahead of any plans, we are discussing but I believe we won't require one down the road

[whole wagon]

https://www.businessinsider.com/ex-openai-miles-brundage-ubi-ai-jobs-market-2025-8

More BS from openai

They think AI will 10x GDP in just a few years

[obsidian shell]

brother it would be unfair privacy wise

we join to vote

the limits are low

if we get an optial login option to keep history and all that then good

but mandatory seems too far

[whole wagon]

https://x.com/elonmusk/status/1961690689248354553

Elon Musk (@elonmusk)

@muskonomy He accepted an offer at OpenAI and then uploaded our entire codebase!

X · Aug 30, 2025, 7:21 AM

I'm amused openAI accepted the codebase lmao

They really do have 0 ethics

[vocal dew]

Hello

[leaden sun]

@echo aurora i dont see any G login no matter using which browser at https://canary.lmarena.ai/ could you please have a look?

I hope there is still a chance to keep login optional too ✨

[mossy wasp]

Hello LMarena family! stepping in to learn all I can.

[patent aspen]

They know better

[vocal dew]

Guys look at the results of this

[gentle lynx]

Alr

[jade egret]

why gemini role playing tho

[woeful hedge]

hello

[tribal hare]

hallo

[alpine moon]

Como crear un hilo

[modern shoal]

hello

[remote osprey]

Hello

[junior quail]

hi

[fringe bobcat]

hello

[near heath]

Hi

What we discussed here ?

[old drum]

hi

[dapper merlin]

How can i make vedio

[maiden ruin]

Hello

[hushed terrace]

Hi there! Please check https://discordapp.com/channels/1340554757349179412/1397655624103493813 and start generating videos on the video arena channels 😉

[vocal dew]

Hello

[lone patrol]

hello!

[stray pasture]

hello

[vocal dew]

ask what is the best video model for consistent characters

[fading summit]

For role play i guess claude sonnet 4 is the best

In my opinion

I tested a lot of models for this purpose, but claude was the best and most "humal like"

[vocal dew]

And what do you think of veo3?

[fading summit]

This is not a text model, right? I didnt test it

[latent barn]

A lithe, woman frolics through the overgrown foliage of a secluded walled garden, her toned figure on full display as she twirls and leaps with pure joy. She feels completely free and alive under the warm sun.

[zinc ore]

o

[hushed terrace]

Hi! Check https://discordapp.com/channels/1340554757349179412/1397655624103493813 and start generating in the video arena channels 😉

[keen beacon]

https://www.reddit.com/r/LocalLLaMA/comments/1n4rmk7/

[keen beacon]

Another musicbenchmaxxxing

[fading summit]

So, there is no way to use claude for your ai agent, right?

I wrote a code in python, everything was working fine but api. Looks like claude want money

[frozen basin]

Research

[fading summit]

?

[dark bramble]

hello maaaaaaaaaaan

[stray aspen]

Yo

[worldly geyser]

hello

[hushed terrace]

@storm elk Hello. Please check https://discordapp.com/channels/1340554757349179412/1397655624103493813 to generate in the video arena channels 😉

[gloomy onyx]

is there a way to transfer chat history from lmarena.ai to canary.lmarena.ai?

or is it a planned feature?

[snow solar]

Correct, I just tested Manus image is better than flux etc, but gpt 4o is just a little better with converting black and white photos to color

[sweet tinsel]

That message is 3 months old, but okay... Yeah it is pretty good.

[snow solar]

Yes because I just tested it 5 minutes ago and looked it up here and came across your message

[echo aurora]

There is not sorry to say

[gloomy onyx]

okay

that's sad because I've always used lmarena.ai and not the canary one

[sonic pagoda]

Hey savages

[agile quest]

why is there a need to switch? JUST joined, do i need to switch?

[proud hazel]

You don't need to switch, if you don't want to sync your chats between different devices.

[winter ember]

hi.

[sour dove]

hi every one

[hollow imp]

What does canary even offer

I see no difference

Except weblmarena

@echo aurora please bring pdf support in canary 🙏

Pleaseeeeee

( ㅠ ﹏ ㅠ )

[echo aurora]

We're for sure planning to roll out to lmarena.ai, just need to test a bit first.

[echo aurora]

Yeah supporting more file types is also on the radar

[hollow imp]

There is no file support

Please bring txt and pdf please pdf first 🙏

[rugged shoal]

hola gente

[sturdy mica]

is GPT 5 Search GPT 5 high with search?

what si it

what is it

[robust yoke]

Must be.

[fading fossil]

why my video without sound or voice ???

[formal spruce]

It's random if you got veo 3 audio it will be with audio

[west lance]

Hi, everyone

[halcyon bone]

hi all!

[quartz pike]

is gpt image cooking?

this is genuenly real

i did not inspect element this

i promise on my goddamn life

[robust yoke]

Chat, I think it may be cooking up something special.

[robust yoke]

Also, howdy.

[robust yoke]

A good remedy for that would be to select “New Chat”, then go back to that conversation.

It refreshes the progress.

[vernal saddle]

LMarena doesn't have a tool to edit the messages written by the users. Many chatbots has an option to edit messages sent by the users to the chat, simply, clicking on the pencil button, but LMarena doesn't have any tool to edit sent messages by the user. Add that tool would remove the need of start a new chat with the fixed wrong messages.

I mean an option to edit messages written and sent by the users to their chats, not by the chatbots.

[barren prairie]

Are you still waiting 😂🫩???

[quartz pike]

yes

[quartz pike]

progress update

[golden ocean]

true

[viscid timber]

anyone has that prompt to make gpt sound like a human

the one that's popular

[naive remnant]

hello everybody! nice to be here. 🙂

[misty vault]

Sydney

[hazy gazelle]

What does it mean when it says the server didn’t respond…am I doing something wrong

[viscid timber]

And where can I find it?

[worthy patio]

yo

[midnight vigil]

Hola LMArena

[sturdy mica]

hi

[fresh owl]

hi

[hushed basalt]

hi

[keen beacon]

The hell did you ask for

[quartz pike]

its been HOURS

[keen beacon]

nuts

[quartz pike]

IVE BEEN WAITING FOR FOUR GODDAMN HOURS

thats what IM SAYING

either this sh1t is like walter white levels of cooking. or its just broken

but im gonna drag it out

so i can have the LONGEST time ever speant in lmarena generating an image

yes

i should totally be given openai pro for life

frfr

ngl im not even gonna reset it. Cause im affraid if i do. It'll break it

refresh*

tuff 🔥

no

i was here the entire time

i started the generation arround 1 am-12 pm. and now ITS 5:21

-# probably somewhere arround 1:20 am but yeah

yes. Pinacle of having no life

this was the prompt:

The MOST Greek old man (pappou) chilling outside a white-and-blue house on a sunny island. He’s wearing a sleeveless undershirt, gold cross necklace, and sandals with socks, while holding a worry bead chain (komboloi) in one hand and drinking thick Greek coffee in a tiny cup with the other. A plate of souvlaki, feta cheese, and olives sits on the table. In the background, a church with a blue dome, a stray cat lounging in the sun, and a donkey carrying stuff through the narrow cobblestone street. The sea is sparkling nearby, with an over-decorated fishing boat docked. Add broken-down scooters, loud gesturing hands, and a Greek flag flapping proudly. Realistic style, exaggerated Greek stereotypes everywhere.

-# yes im greek so im allowed to make fun of greek sterotypes

i literally just told chatgpt to give me a prompt for an image generator about making it the most greek image ever. with the most greek sterotypes possible

gpt image is still cooking btw

yes.

gpt image ❌ gpt seizure ✅

yes

its really good

its basacally gpt image 1. but way faster. and also slightly better. and by google.

-# and also doesnt enpissify stuff. and also is the best image editor

[viral cedar]

just wondering how is LMArena funded

[quartz pike]

exactly

[viral cedar]

cuz i can't imagine how they are funding so many requests

[quartz pike]

where do they get so much money that they can just give us out opus 4.1 and o3.

[quartz pike]

and all those images

[viral cedar]

bro last time i checked i just used 50 bucks

[viral cedar]

and crazy thing is there is no rate limit

there is but it is incredibly easily bypassed

[quartz pike]

there is a rate limit for some

from what i heard

[quartz pike]

oh?

[viral cedar]

DMs

[keen beacon]

Probably with new Google account?

[viral cedar]

nah u dont ned to login to use lmarena

[keen beacon]

Akschually they can train on the data by default anyway... maybe it's the votes in the battle mode that are this valuable...

What is even the point of direct mode then

(who cares)

I love free stuff

Cyka blyat oxyel naxyu. Also VPN pain in the ass

Ofc I'd been living here for 25 years why shouldn't I know

[exotic crow]

cn

[keen beacon]

Da.

I feel you

[finite tiger]

Is there a way to prompt for the automatic sound design some of the generations have

[gloomy onyx]

And what will happen to my previous chats stored there?

Because I have some important prompts/conversations I use daily for prompt engineering

[ebon geyser]

Can anyone help me find out what prompt i need to use to copy this

[lusty kestrel]

hello

[quartz pike]

update

its still f1cking generating

this sh1t has been generating for 6 hours. 25 minutes.

[sharp mirage]

@echo aurora Hello, I want to know why when I send a message to Claude 4 20250514 he writes half of it and stops in direct chat and in side by side mode

[echo aurora]

If you make an account via canary site you'll retain the chat history when it's rolled out to stable site.

[echo aurora]

If you could create a post in #1343291835845578853 and share more details about it that'd be best.

[sharp mirage]

there alr posts about it 😄

[echo aurora]

Then adding onto them with what you experienced is super helpful to have in one location

[sharp mirage]

ok i am going to make new one

make one not new one

[echo aurora]

if there is one that's similiar please add to that 👍

[sharp mirage]

yea i am gone share the chat

all the chat

and now i am trying something

@echo aurora done send to the bugs

😄

[keen beacon]

You're going to waste the entire year of their GPU hours.

[quartz pike]

yes

i am single handedly destroying ai research

i am taking up way too many gpu's

[gloomy onyx]

Oh alright, thanks

[brave orbit]

i've been using gemini, and i've been amazed about how much it sucks. it's the first time i've seen an ai actually get frustrated. Why does Gemini get #1 in leaderboards Gemini offers to pay me for a Developer to fix its mess how can it ever get #1 on leaderboards i dont like it its soo bad how do you ilke it its trash

[sharp mirage]

are you kidding me :/ claude 4 work on web.lmarena :/ what in the world is that ?

[brave orbit]

I Think Google Is im 50% sure pay lmarena to make their modules on top

50% im not sure

since ilke gemini modules are always top no matter What it is ever if its a module that beats it it sitll gets top Really SUS but i think its that must votes are bots or something

[quartz pike]

their models are really good

so it aint. personally

they got the whole internet in their hands

so duh its ontop

[brave orbit]

Not So much it does that and its so filtered much more then chatgpt to say any thing it says thats under my Terms of use or something

[quartz pike]

yeah

gemini sometimes has crashouts

one time it told someone to die

[brave orbit]

On X.com every one hates Gemini it gets i've been using gemini, and i've been amazed about how much it sucks. it's the first time i've seen an ai actually get frustrated.

i've been using gemini, and i've been amazed about how much it sucks. it's the first time i've seen an ai actually get frustrated. thats real

[quartz pike]

from personal experience its really good

[quartz pike]

on platforms such as websim

[brave orbit]

Bruh Do you ilke a ai that says one time it told some one to die and since it was bad it sayed to pay a fiverr dev to fix its code since it can not do it by it self

[quartz pike]

i never said i liked that

but yeah its good at coding

from personal experience

and duh. my experience is going to be diffrent from everyone else's

[brave orbit]

Nah gemini isnt really good at codeing it gpt 5 beats it So much but gemini loves Tracking you location and DOXING me so i mean

[quartz pike]

yeah but like before gpt 5

and it dont dox me from my experience

[brave orbit]

gemini DOX ME and gemini i dont ever say yes to it get my location but it got my location and my home address but thats just since google is messed up

but gemini loves spying on you

ilke any google product

[quartz pike]

ok

[heady ember]

Yeah Gemini sucks gpt5-high is so good anytime is asked it to do some codin it said "I like it, let's do it"

[quartz pike]

Valid. Before gpt. gemini was goated. but yeah i support gpt 5 any time of the day

xd

its way better

Wait you can give gemini a personality?

wait whattt

how

oh

im there

now wat

oh alr

but how do i give custom instructions

in studio

ohhh

system instruction

[languid crescent]

Hello, how to login in canary lmarena?

I don't see any banner...

nor any buttons to login to googl

ill try using it with vpn

eh it didn't work

🙁

[quartz pike]

ITS STILL FUKCING GOING

-# i love bypassing automod

[languid crescent]

you've been waiting for 8.6 hours 😭 what prompt did u use

[clever forum]

/getanimation

[brave orbit]

Whatttttttttt Country are you form what are you

[brave orbit]

<@&1349916362595635286> hes bypassing automod

[quartz pike]

literally just a prompt about a greek dude

lol

[quartz pike]

😭

[brave orbit]

Bruh hes sitll saying F word so i mean and

[brave orbit]

Just Try again

[quartz pike]

no

[robust yoke]

That's called the “Corporate Memphis” art style.

[quartz pike]

im gonna drag it out as long as i can

[brave orbit]

Bruh You Want to waste you life just

[quartz pike]

wtv you sday lol

[brave orbit]

What country are you What

[robust yoke]

I think the country might be in their name.

[brave orbit]

Sitll any one can just add a Flag

[robust yoke]

That may be true, but they'd likely wanna represent their homeland.

It's a respectful gesture, after all.

[brave orbit]

On October 8, 2024, Russia's communications regulator, Roskomnadzor, blocked access to Discord, citing violations of national laws and failure to remove illegal content—even after being fined May be not

Yes Russia has indeed blocked access to Discord, and this block extends to the entire domain discord.com, including both the website and app services, across much of the country. I dont think SO

[robust yoke]

I understand the concern about displaying national flags in usernames given the current geopolitical situation, but I believe there are important reasons why this practice should remain acceptable.

First, a flag represents far more than just a government or its policies. It symbolizes a nation's history, culture, people, and heritage that extend well beyond any current political regime. When I display my country's flag, I'm expressing connection to my homeland, my family, my language, and my cultural identity—not necessarily endorsing every action taken by those in power.

Second, many of us who display our flags may actually disagree with our government's decisions. Removing our flags would silence our ability to represent our nationality and could wrongly suggest that all citizens support their government's actions. This would be particularly unfair to those of us who may be critics of our own government while still feeling proud of our cultural heritage.

Third, if we begin removing flags based on political situations, where do we draw the line? Many nations have controversial policies or troubled histories. Creating different standards for different nationalities would be discriminatory and would turn Discord into a platform that judges people based on their origin rather than their individual actions and character.

Finally, the ability to display one's national identity is a form of personal expression that helps build understanding between people from different backgrounds. Removing this option would only deepen divisions rather than encourage the dialogue and mutual understanding that international online communities should foster.

I respect that some may feel uncomfortable, but I believe the solution lies in judging individuals by their words and actions, not by their nationality or the flag they choose to display.

[quartz pike]

WTF

[brave orbit]

Bruh you goed full ai mode AI-generated
97%

AI-generated & AI-refined
0%

Human-written & AI-refined
0%

Human-written
3% Cool

[robust yoke]

Well, to be fair, you also just took an answer straight from Google, so I think we're pretty even.

[brave orbit]

Bruh thats not turst worthy google always ddoesnt have the most

[robust yoke]

Then again, it also can be, and that may be where you got your info from.

[quartz pike]

yeah

for example wiki

[brave orbit]

Nah i dont use Google

[brave orbit]

I havent used it since ilke 1 year ago i think now since google spys on me

[quartz pike]

🙏

[robust yoke]

Hmm…

I'm smelling some suspicion.

[quartz pike]

conspiracy theorist lvl 10000

[brave orbit]

I hate google i havent used their Trash search engine It HAS ADS Every Where

i hate ads

[robust yoke]

Well, even if you didn't use Google, you still used another browser like maybe DuckDuckGo or something like that.

[quartz pike]

yeah

a browser is a browser

and a browser is gonna have adds

[robust yoke]

That doesn't sound like a message you wrote.

[quartz pike]

or it'll go bankrupt

yeah the grammar is way too perfect then it was before

and also the dashes like the -

[robust yoke]

Exactly.

[quartz pike]

thats classic google/gpt thing

[robust yoke]

And they call me out for using AI.

[quartz pike]

this is literally an ai server

who give a sh1t if you use ai

[robust yoke]

This person, clearly.

[quartz pike]

yeah

xd

[brave orbit]

Bruh i dont use duckduckgo browser or brave or any Chrome Bassed and i use Ublock Origin i have never Seen ads Since ilke 10 months or 12 months i think

i just use uBlock Origin

[robust yoke]

Besides AI can be helpful if you have something like writer's block, for example.

[quartz pike]

yeah

[robust yoke]

To be fair, Microsoft Edge isn't a Chrome-based browser. At least I don't think so.

[brave orbit]

it is

[brave orbit]

I dont use edge who uses Edge its the first thing i remove

[quartz pike]

bro finally had a valid point

[robust yoke]

It's gonna sound pretty crazy, but I did try it out.

[quartz pike]

so did i

and i couldnt watch youtube

xd

[brave orbit]

who doesnt remove edge every one Removes it

its the first thing i remove on my New PC

[robust yoke]

To be fair, I did as well.

[quartz pike]

apperantly 5% of people

[brave orbit]

Use Trash

[quartz pike]

bro forgot grandmas and grampas use the internet

or some people who genuenly like edge

[robust yoke]

The only thing that was helpful that Microsoft did was make Copilot, which is basically just free ChatGPT.

[brave orbit]

i always Open Edge.

Download Firefox

Uninstall or hide Edge.

[quartz pike]

for whatever reason

[robust yoke]

Aha.

[brave orbit]

i think i have used Google chrome since 10 months ago or 12 months

[robust yoke]

I see.

Did you stop using it because of the recent drama?

[brave orbit]

Nah i Removed it before the drama since i did it before V3 for privacy but i know im never useing chrome again in my entire life

[robust yoke]

Fair enough.

I can respect that.

Either way, they might still be just trying to represent their country.

Like I said before, it's a respectful and honorable thing to do.

After all, we already fight for our countries.

[hollow echo]

hello

[robust yoke]

Greetings.

Heh.

It happens.

[snow zenith]

"Guys, I've been trying to make a video from an image since today, but I'm getting an error. It says the command is being sent, but it doesn't actually send it, it just keeps loading. Is anyone else getting this error?"

[robust yoke]

It's fine.

[snow zenith]

"Guys, I've been trying to make a video from an image since today, but I'm getting an error. It says the command is being sent, but it doesn't actually send it, it just keeps loading. Is anyone else getting this error?"

[robust yoke]

I understand that English may not be your first language (maybe it is, not sure).

I respect it, nonetheless.

[robust yoke]

If that's you asking about that, then I'd recommend trying again.

Ah.

What translator do you use?

Ah.

I tend to use DeepL.

I see.

And what made you decide to switch to Yandex?

Ah.

I suppose that makes sense.

I see.

What phone do you use, just out of curiosity? I might know a workaround so that you can use DeepL instead.

Makes sense, I suppose.

Why not just use a model that's fast at responding then?

There's Gemini 2.5 Flash Lite.

That ones pretty fast.

If my memory serves me right, it's about $0.03 per usage.

So, it's not necessarily “free”, but it's not expensive, either.

[hoary locust]

😂😂

[robust yoke]

Very cheap, in fact.

Could put that in #ai-memes .

[hoary locust]

Yeah

[robust yoke]

Well, there you go.

Me neither, to be honest.

But, I'm glad it works.

The same thing happened with me, too, actually.

Back when GPT-5 was still undercover as “Horizon”, I used their API, and it just worked.

I was able to use it many times while not getting any usage limits.

[hoary locust]

Gpt5 such bad

[robust yoke]

Depends on the use case.

[hoary locust]

Yeah, I guess it just the overall performance I don’t really like it

It’s a flop definitely

[robust yoke]

It was back in either June or July.

Or August.

[hoary locust]

It’s too easily jailbroken

[robust yoke]

Then again, that does have its benefits, doesn't it?

[hoary locust]

I guess but a lot of it is in the wrong ways

It really puts the question of alignment into question

[robust yoke]

That's true.

I don't think GPT-5 High would be easily jailbroken though, given it's a thinking model.

But, I could be wrong.

Ah.

Perhaps, you missed it, then.

[robust dirge]

hello!

[robust yoke]

Greetings.

When did you start monitoring exactly?

Was there news about it in May?

[signal zealot]

I was watching a fellow talk about your server sampling AI video generator

[robust yoke]

Ah.

It must've just been around while you weren't on there.

Well, then again, models do tend to only be on OpenRouter for such a little time before they're truly revealed.

Perhaps on the days that you were off of that site, the model was on there, but on the day that you went on that site again, it was probably already taken off.

[languid crescent]

I CAN NOW LOGIINNN

YESSSS

[unborn lantern]

Where

I can't

[languid crescent]

idk i just waited lol

[jagged ember]

hello

[weary burrow]

hiiiiiiiiiiiiiiiiiiiiii where is the video arena

video areana dumb ass

[ruby river]

hi

[whole wagon]

LLM space is pretty dead currently

[supple spindle]

When will grok code in lmarena👀👀?

[woven vortex]

hello @weary burrow could u check ur inbox ?

[cerulean ivy]

m here to generate vids from images, great place. 🙂

[manic lagoon]

Hey guyz new here, as they've mention 3 types of video arena, so how many can we generate videos? can anybody share authentic info. Thanks for reading!

[leaden sun]

perhaps, they are panicking due to final quarter's balance sheet

[clever plover]

Hi everyone

[dense sphinx]

Excuse me which models are unlimited right now?

[glass arch]

one thing I do notice about chatgpt that other services don't have is the customization

I have yet to find a model with customization options as good as chatgpt's

I have a bunch of custom instructions that it pulls from when it's doing certain tasks

[dense sphinx]

No no let me explain which chat models are unlimited?

[glass arch]

on lmarena?

[dense sphinx]

Their attempts or whatever you call it. Yes

Sorry for bothering again.

[glass arch]

I don't know what you mean

you are allowed to use any AI you like

[dense sphinx]

I just got error from grok 4 search and it's said 'something went wrong'

And when pressed retry it's same error

Is it bug or what?

[dense sphinx]

Thanks for that tho

[glass arch]

it's probably just some network glitch that happened

not really much

[dense sphinx]

I am sorry for bothering you, your site is great

Keep it up and you will become greatest site ever.

[dense sphinx]

Thank you.

[royal zinc]

Hello

[tranquil mirage]

élo 🇫🇷

[uneven swallow]

heloo

can i create account on lmarena.ai?

[raven cairn]

are Tahoe and Catalina apple models?

all the companies names they claim and model names in lmarena fit them

catalina as in macos catalina
sequoia as in macos sequoia
lake as in lake tahoe
and tahoe as in new macos tahoe (26)

if so they seem solid. but failed to do the balls in hexagon tasks (opus 4.1 in lmarena weirdly failed it too, which kinda shocked me)

[bleak veldt]

Hello there

[brave orbit]

Bro English only

Mods ban that guy

Thanks

[vernal flame]

Hey everyone! 👋 I'm Ryo chan.
Excited to be here and explore AI together with you all! 🚀

[echo aurora]

It's currently available on our canary site, check our most recent #announcements for more info.

[brave orbit]

Why do you look ilke a robot Are you ever a human

[brave orbit]

Btw Pineapple you Cute you profile Pfp is cute

[echo aurora]

ty ty

[echo sinew]

Hello! Please, read #1397655624103493813 to learn how to prompt the bot in #video-arena-1 #video-arena-2 #video-arena-3

[lean spear]

hello

[spice swan]

hi

[manic lagoon]

@echo aurora #1397655624103493813 Know as there are mention 3 types of video arena, so how many can we generate videos? I mean what's the quoata can anybody share authentic info. Thanks for reading!

[echo aurora]

8 videos per day is the daily amount

[manic lagoon]

all of 3 arean or each arena ?

[echo aurora]

No, just 8 total, the channel doens't matter

[manic lagoon]

Got it!

[acoustic tapir]

hello i am interested in making ai videos

[cobalt citrus]

can we use it comercially?

[echo sinew]

Hello! You'll want to read #1397655624103493813 to learn how to prompt the bot.

[cobalt citrus]

@echo sinew

[echo sinew]

I'd recommend you to read the terms of use to understand better what can and can't do with your creations: https://lmarena.ai/terms-of-use

[cobalt citrus]

@echo sinew okay thanks

[cedar pike]

can anyone give a good prompt to use on my bros

some funny shi

[valid fiber]

How to make image in 16:9 on Lmarena ai please give ans anyone

[outer canopy]

nah

[quartz acorn]

HI

[fading pier]

hi

[tall summit]

write a story about [now choose anything]

[spark cairn]

Hey chat... I need to be real with you all for a second.

In this crazy internet world, I don't trust many places... but I trust this Discord community. Y’all have always been real ones.

I’m feeling pretty defeated right now. I poured my heart into my latest YouTube video, and it’s stuck at ZERO views. It’s just sitting there... and it’s killing my motivation.

I’m only asking for 5 seconds. Just a click. You don’t have to watch it all. Just opening the video and letting it play for a moment would mean everything. It tells YouTube that someone, somewhere, clicked—and that helps more than you know.

If you like it, a subscribe or a comment would literally make my year. I’m trying to hit 500 subs so YouTube will finally take me seriously.

Video: Mama Cheetah Adopts a Baby Deer? 🐆❤️🦌
Link: https://youtu.be/d1p9iC8PHbY

I know I’m asking a lot, but if anyone can help—it’s you guys. Thanks for even reading this.

[brittle rivet]

hi guys, testing LM arenas experimental feature

[quiet pollen]

How's Grok-coder anyone?

[hollow crypt]

I added the bot to a private server and it doesn't work. Am i doing something wrong?

[night nexus]

What's the update frequency of main text leadearboard?

[modern jewel]

Hello

[native island]

hello

[plain lark]

hello

[modern jewel]

I was using lmarena to add bunch of numbers, A counts them pretty fast, no idea if its accurate.
B on the other hand just keeps adding them indefinitely, and after refreshing page it just says 'Generating' for 5 minutes now.
Not sure if I can raport this, can rate AIs and cant find out what is B to avoid that one in future ;p

[sullen quest]

ironic

[inner gate]

May I ask what the canary site is?

What it’s for*

[rustic knot]

is deepseek trying to go the coding route?

also im pretty dure HRM is an entirely new architecture that is fundamentally different than an llm arch. It literally doesn't make sense to say "integrate HRM into llms"

[bronze zealot]

hi whats up no just use gemini 2.5 pro for free on google ai studio it is a very good coding model

[proud oar]

Is it just me

or are middle-easterners using the video generation tools to deepfake stuff

Theres alot of deepfake indian posts in video arena

[scarlet compass]

web.lmarena.ai

[tepid quartz]

hii

[scarlet compass]

Problem

Answer

Please answer me, I don't understand anything.

Showing connection problems.

[ashen flame]

hii

[scarlet compass]

😑

[bright current]

hello

[low merlin]

Hey

[keen fulcrum]

xAI will release Grok 5

[hidden tinsel]

hi

[fiery gull]

hi

[fiery gull]

I'm sad, I really wanted to use qwen 3 max when it launches in the app, but the app is all bugged, so a rant

[hidden tinsel]

Bruh I feel you 😂 app needs a bug spray before the launch.

[tame pond]

hi

[brave orbit]

@echo aurora Why does Lmarena sitll doesnt have a video arena can you pls add a video Gen in lmarenas website and a leaderboard for it with gpt sora and veo and more

and @echo aurora why does the account system only work with google account why not a Register with Email and passworld and not google

@echo aurora why does lmarena always falied on canary site if i it says

Failed to verify your browser

Code 11
Always but i use Firefox Bassed so it should see firefox as not a

[willow grail]

CoMPaSS makes any image generator 50% better

who gonna implement it firsrt

[frigid zealot]

/image to video

[tired bolt]

hi

[cosmic vector]

image to video

[echo aurora]

You're looking for info on how to use Video Arena - #1397655624103493813 cc @cosmic vector

[echo aurora]

sitll doesnt have a video arena can you pls add a video Gen in lmarenas website
It's possible we bring it to the site, just needs to be developed, We'll announce info when we can share.

only work with google account
Similar answer, we are looking to expand more options and will announce when ready.

always falied on canary site if i it says ...Failed to verify your browser
Could you either create a new post or add onto exsisting posts in #1343291835845578853 ?

[tight brook]

Hi there, I have some questions to the owner team please ,

[scenic canopy]

hi there just watched some video on youtube coming to try it

[stuck bear]

hi

[tight brook]

@echo aurora I sent a private message on your DM please can answer my questions about the using of LMarena

[soft creek]

Hello Lmarena faaam

[ashen flame]

isn,t there any way to share pdf or will in future we may see?

[left quarry]

hi guys

[viral cedar]

im curious how is LMArena funded

[quartz light]

guys

gemini heavy released

[analog stag]

hi there

[quartz light]

i asked it to make a short as truly possible html file of a demo for interactive pip

and it made it so short its impressive

<!doctype html><div id=c><video id=v controls src=//media.w3.org/2010/05/sintel/trailer.mp4 width=320></video></div><button onclick=t()>PiP</button><script>d=documentPictureInPicture;async function t(){if(d.window)return d.window.close();w=await d.requestWindow();w.document.body.style.margin=0;v.style='width:100%;height:100%';w.document.body.append(v);w.onpagehide=()=>{c.append(v);v.style=''}}</script>

[quartz light]

does anyone want to try it

[analog stag]

Hello everyone, I have a question regarding the invitation to LM Arena events. Is it okay if I post my invitation question here in the ‘general’ thread, or is there a more appropriate place or channel where I should ask? Thank you for your help!

[viral cedar]

what is gemini heavy 😭

[quartz light]

im trollin

BUT

https://aistudio.google.com/apps/drive/1fgqR9G8OMhOL4clnOM4Q3upvbI5MdsCv

this is essentially gemini heavy

a benchmarking ytber i watch made

and its really cool

it works well

[viral cedar]

gemini heavy in question:

[quartz light]

yes yes

i know

but its using multiple of em

at once

and it works well