general

[mystic prism]

can anyone tell me how to get start fame end frame video in any of these arenas?

[white bison]

I need to study to my girl which in first secondary school first term has exams tomorrow start with philosophy in Arabic

[gleaming roost]

[white bison]

Can I give you a textbook in pdf and you summarize it

[echo aurora]

You’re unable to specify that with the current setup. Video Arena can have 1 image uploaded

[mystic prism]

ohh

any workaround?

[mystic prism]

Ohh ok. Any workaround?

[frigid tusk]

am i tweaking or bug?

[loud pebble]

hi

[proud bobcat]

thisis weird

what modality do you have it set on

[stray aspen]

whats iquest coder

[gentle summit]

hi

[echo aurora]

Not that I'm aware of sorry to say.

hello @gentle summit @loud pebble

[echo aurora]

~~Sorry I'm not spotting the weirdness, what's off? ~~ Nvm I see it now.

[dusky ravine]

i wonder why when i used direct chat, sometimes it changed into battle mode randomly. Im not sure if it's bug or feature

[echo aurora]

It's an experiment we're currently working on, this isn't a bug.

[dusky ravine]

oh for what purposes?

if i may know

[frigid tusk]

idk , it just became like that today , ive been using LMArena for a good time now so hope it get fixed

[echo aurora]

That's on Text

[echo aurora]

Just an experiment to see how users would use something like this. What are your thoughts on it?

[viscid cloak]

Did that happen in a chat you've already had chat history? If it triggers only for new chat, then it should be fine, just delete chat and run Direct again. If it randomly pops up in an existing chat, then I could tell it must be a huge trouble for users.

[echo aurora]

Can I have the Eval ID for when this happened? Eval ID is the random set of numbers/letters in the URL of that chat session.

[fallow geyser]

I don't see a specific channel to ask questions, is it fine if i ask in general?

[golden ocean]

did nano banana pro die
Something went wrong with this response, please try again.
other models work

[fallow geyser]

I wanna know what's the actual amount of time before a model times-out.

I know that if any model gets stuck and just keeps generating, there is a specific amount of time after which it times out and you can retry, but i wanna know how much that time is.

Since well i can't just stop the model, there is no stop button as we are all aware.

And since I'm already here, may as well ask,

Why has LMArena never had a stop generation button?

Surely there is a reason behind it.

[gleaming roost]

There are so many things missing that I honestly don't see any reason why they haven't implemented them

[echo aurora]

Yeah asking here is totally fine! Would note that using #1343291835845578853 for reporting issues and #1372230675914031105 for sharing feedback is ideal, but yeah happy to answer questions here.

I wanna know what's the actual amount of time before a model times-out.
This was a recent change, but will share the feedback having an idea on the amount of time it's taking to generate is something our community would like to see. Others have voiced similar feedback so this may be brought back. Is the reason for wanting this mainly to get a better idea if the model is stuck, or are there other reasons why this is important to have?

Why has LMArena never had a stop generation button?
This is something that's on our list. It's a resourcing and prioritization question at the moment. However, we are hiring more and are planning to build a lot more features this year. cc @gleaming roost

[wicked sage]

hi

[steel oracle]

Apps > Delete Generation

[split creek]

@nova sundial @echo aurora , Hi, my LM Arena image-to-video generation has been stuck for over 30 minutes.
Can someone please check if the queue is delayed? 019ba3ec-75fa-7cb4-a240-8113762b2855

[fallow geyser]

It's just that I've noticed opus 4.5 thinking is quite prone to getting stuck.

Which is what happened exactly right now in my recent chat, but i don't wanna start a new chat and lose all the context , so if i know when it's gonna timeout and get unstuck, i can come back to that chat after that amount of time has passed.

[echo aurora]

Models getting stuck is an issue that we're aware of. If you hard refresh the site + clear cache this may help it move along. Would ask to save pinging the @ Moderators for server moderation related things (reporting other users breaking server rules), pinging me directly is preferred. Also worth noting that @ Moderator is the correct ping (it'll appear in red). Thank you for sharing that Eval ID.

[echo aurora]

Could I have the Eval ID? We are sharing these with the team to help us investigate what's causing this.

[fallow geyser]

Alright, where do i check the eval id?
Is it the url?

[echo aurora]

Correct, sending the full URL is fine, but the Eval ID is the random set of numbers/letters. Looks similar to what was shared in this message #general message

[fallow geyser]

019ba118-ea91-751d-9d31-6374f5b34177

[echo aurora]

Okay thanks. This is helpful.

[fallow geyser]

👍 no problem.

|| i just sent and deleted some messages in this channel, don't be alarmed if you saw them but are confused as to why i deleted them,
I had opened the wrong chat so i thought the model timed out and i can continue the chat but upon opening the correct chat i realized that the model is still stuck and hasn't timed out as of yet, so just ignore what i said if you saw anything and if you didn't see anything then simply ignore this message too||

[fallow geyser]

019b98b2-e058-7796-a24d-95d73fa601e3

This is another chat which got stuck.

Might be of use.

[feral holly]

guys one question,is there a limit to how much videos i can create in one day?

[echo aurora]

Yes, through the Discord bot it's going to be 5 requests per 24/hr. Via the site (which is currently an experiment so it's going to be random if you get this or not) it's going to be 3 requests per 24/hr.

[feral holly]

but site only has prompt to image

no video

[echo aurora]

Sort of. There is an experiment running currently with Video Arena on the site. This means it's random if you get access to the feature or not. If/when we do release this feature to everyone we'll be sure to put out an announcement.

[stray aspen]

[feral holly]

ok

thanks

[vital lake]

Just put the most known models first or something

Or most recent

[echo aurora]

@lilac dew be sure to review the information in #1397655624103493813 this will show you how to properly use the bot.

[little ginkgo]

When we gonna get feature to upload files

[old garden]

[little ginkgo]

[echo aurora]

I don't have an ETA I'd be able to share on when to expect this feature to be added.

[little ginkgo]

[old garden]

@echo aurora can u add as a reaction

i miss reacting the announcements

[whole sundial]

yeah this server can have 250 emoji but it only has 10

[gleaming roost]

react to that in
I don't even feel like it.

[whole sundial]

yeah i feel like allowing people to react with whatever they want adding some personality to this server, and now we're forced to react only with 3 different lmarena emoji, it could be worse though

openai only allows you to react with the openai logo in announcements

[echo aurora]

Tbh I've been on the fence about this change. The reactions are a bit difficult to moderate, and we've seen some cases where they've been used inappropriately. At the same time though I do agree with: adding some personality to the server.

[stray aspen]

openai is the most strict server ive been in my whole life lol

[mint urchin]

If I write code for an application that uses a ai, can it be stolen or leaked online?

[hushed gyro]

@echo aurora Got it when I refreshed

[echo aurora]

Nice

[hushed gyro]

Is it still in experiment phase or you just released it? 🤭

[echo aurora]

Still in an experimental phase. If/when we do launch we'd be sure to put out an #announcements

[pulsar crystal]

were some larger models removed like google pro or grok 4?

[hushed gyro]

BTW why is it battle mode only 👀

Aaaand it's gone

[echo aurora]

We do have a list of deprecated models here. However, it hasn't been updated in awhile. I've bumped this with the team,

[remote sun]

pls add edit and delete options for messages

[surreal galleon]

Have anyone tried Deepseek v4 :d

[sterile tartan]

Is not Released yet

[terse kernel]

how can i fix it?

inf generation

[echo aurora]

Sorry to say this is a known bug. Our team is actively working on it. In the meantime I'd recommend hard refreshing the site and clearing cache. This may help it get unstuck. If no luck, starting a new chat may be your next best option. Although this isn't ideal as it means your chat will no longer have the previous context.

[terse kernel]

ty

[cursive shoal]

am i the only one or is nano banana completely down on all the official google sites?

[stone cape]

anthropic blocked opencode requests 🙁

sad

[west lodge]

hi hi mods im like a huge nerd so id like to hear from the devs, some question has been bugging me for a bit, what do they even use in the backend in terms of moderation? like what models are used for safety classification? id also like to use them in my own stuff since from my testing these seem quite solid

[steel dune]

Sorry for the late response. It started working again after I opened a new chat room. Thank you once again for your assistance. Much appreciated.

[fiery gull]

whaaaaaaaaaattt

hell no

13/01 ;-;

F

[echo aurora]

Hello we do have a platform-level guardrail in place in the form of a content filter. This helps ensure evaluations are safe for all participating. I won't be able to provide details about this moderation filter is.

[west lodge]

aww that sucks

well no harm done

[sinful shale]

a

iz normal?

wot

[proud bobcat]

Not released yet

[hollow crescent]

Hello, hope you are well. I have come to generate some vids

[quasi atlas]

@hollow crescent Please check on #1397655624103493813 to learn how to use the bot.

[vital lake]

Is it just me or is this chat not working?

[obtuse smelt]

ah me too

i look all models say Something went wrong with this response, please try again.

[solar karma]

@echo aurora did the model organization change? cuz I could swear it wasn't like this yesterday, and is there any way to change it back or is it permanent

[weak nova]

Ya same for me

[frosty adder]

https://smileycode.vercel.app/

[weak nova]

That color scheme is too overused now and ppl often dismiss it as "slop" 😔

Pick some favorite colors and change the design to them

I would do honey yellow, olive green and beige/cream

[still hawk]

hello

[fallow geyser]

Yeah it did change

[sick gazelle]

troca essa cor pelo amor de deus

[fallow geyser]

There are 2 different bugs actually,

One is fixable with a simple refresh (no need to clear cache) and one isn't.

And i know how to distinguish between them.

If the model just doesn't generate any text at all from the very beginning, then thats the fixable one, you gotta wait a bit until you feel like it has FINISHED generating the whole message, then just refresh to see the message, it's just a visual bug, the model isn't actually stuck.

Then there is the different one.

In this one you can see the model start generating text, but in the middle it will just STOP, you can't fix this one with a refresh.

It will just be stuck until it time's-out.

||(Which you still haven't clarified how much time it actually takes btw)||

I thought i should let you know just incase yall didn't already know.

[heavy ferry]

Hi, my chat has been stuck on "generating" for two days.

[mossy girder]

Why now Direct Chat forces you to Battle two models forcefully after 3 queries all time?

[fallow geyser]

Congratulations , apparently you have now been promoted to one of the esteemed early feature testers.

[fallow geyser]

Same lol

You just gotta wait it out

Eventually it will timeout

[rose sky]

Heh... found a way 'round to generate legal and legit free and unlimited Nano Banana Pro images and Veo 3 videos

[icy gulch]

So where do we identify all the models that were used when we generate image to video generations?

[mossy girder]

What? Only selected ppl?

[fallow geyser]

More like random.

Thats what i heard pineapple say in the above messages at least.

[mossy girder]

That's harsh

[fallow geyser]

You can scroll up a bit to check the messages yourself.

[mossy girder]

Thinking model takes alot of time

[fallow geyser]

Yeah but I'd rather have the model take a lot of time and be correct than have it be fast and incorrect

You also have the option to just not use thinking if you want fast and quick answers ya know

[feral holly]

how

[rose sky]

Well, I followed a YouTube totorial

[feral holly]

and it actually works?send link

[rose sky]

I don't think I can send the YouTube link here, this server doesn't allow it, does it?

Otherwise I can just DM you

[next minnow]

because whoevers pushing these changes is tone deaf and arrogant

[feral holly]

@echo aurora there is a bug.when i select direct i still gives the comparison response

[mossy girder]

I played it... i deleted acc n created again lol

[next minnow]

literally most mindnumbingly ridiculous change i have ever seen

[feral holly]

send personal

[next minnow]

only joined here to ask if it was a bug

guess not

[rose sky]

Alr

[rose sky]

Done

[obtuse smelt]

hmm i guess done fix it

[next minnow]

ah so its one of those
i hate when companies won't give you an option to opt out of their ridiculous experiments
google does this all the time

like i can understand rolling out experiments over time but i should be able to opt out if i want to

[frosty adder]

Are you suggesting changing the color of the website or the smiley code program that needs changing?

[obtuse smelt]

ah yeah i retry but not responding

[next minnow]

doesnt mean direct chat should force you to compare random models every 3 messages

at the very least don't have the battle actively affect the rest of the direct chat - if im trying to do something with one specific model i shouldn't have to deal with influence from other models' responses

[mossy girder]

They aren't getting enuf votes, this way they can squeeze more for research n labour RLHF from us without giving option to opt.

No one using it free sir, it as start claims, we're gonna use your data fully

[fallow geyser]

Exactly.

You're paying with your data.

The act of using IS the act of paying.

[next minnow]

me when i cant actually argue my point so i pick up the goalpost and sprint to the other side of the state with it

(also like joku and deadlyspice said the act of using is the act of paying)

[mossy girder]

Exactly, no one is entitled in anything in the world. Trade happens...

[next minnow]

anyway im not interested in entertaining a bad faith argument sooo imma go live my life

[fair roost]

gemini 3 pro not available?

[obtuse smelt]

hmm i use gemini 3 pro

[fair roost]

but why is it not showing in this list?

[obtuse smelt]

hmm you just put in image only on gemini 3 pro, btw always error

[split drum]

Giys

[tall tulip]

Infinite generating image and when refreshed, session not found?

[split drum]

How to make my videos viral?

Anyone?

[obtuse smelt]

ah me too

[fallow geyser]

it is there , just click on search model and put the name in , or scroll down

also , <@&1349916362595635286> , ban these bots

[mossy girder]

Haha, but at the end of the day they get signal from our noise of data

[robust sluice]

whoa, 3 scams

[left lodge]

@echo aurora How does lmarena work financially? Providers provide apis free of cost for testing and data or you guys pay to the providers at a very cheap rate? The first one is likely because thats the most suitable after seeing it all.

[rigid copper]

<@&1349916362595635286> someone spam some nonsense stuff in this channel

not only this channel, other channel got too.

[pure swan]

Grok getting way to dark

[rich mauve]

Yeaaa

[rich mauve]

(This only works in battle)

@echo aurora Will you add side-by-side and direct chat to the video?

[rigid copper]

before that, the video generation in lmarena support direct chat and side by side, now they restricted it to battle mode only.

[rich mauve]

🙁

[rigid copper]

idk why, but i highly doubt that because many people would choose the most expensive model to generate video, especially veo 3, seedance, or kling.

[pure swan]

Grok hitting it hard @echo aurora

[loud verge]

Are we allowed to ping the pineapple man?

Also, why has image attachment option disappeared for gpt 5.2 ?

Or was it never there. (Which I highly doubt)

[surreal frost]

@echo aurora mister pineapple man why did my account on lmarena bug out, show me an error when tryna load a tab at first then when it got fixed, reset almost all the icons of all my tabs as well as reset ai models used on each tab?

[twilit widget]

Hello Everyone

[burnt basin]

hello everyone looking forward to learning

[robust sluice]

scammer so hard working today

[neon idol]

<@&1349916362595635286>

[regal fractal]

anyone know of there anything on here that can replace video and lip sync?

[boreal saddle]

What about this question violates the rules?

Apparently you cannot even criticize a specific country without being smacked down by a overzealous filter.

This sounds so AI generated, fr

[proud bluff]

HI

[ocean vortex]

They do not 🙂

Your conversations and certain other personal information will be disclosed to the relevant AI providers and may otherwise be disclosed publicly to help support our community and advance AI research.
Do not submit to our Services any personal information or other sensitive information that you would not want to be shared publicly. By continuing to use our Services, you acknowledge and direct us to engage in such sharing.

https://huggingface.co/datasets/lmarena-ai/arena-human-preference-140k

you may find your own convo there 👀

🗿

[gleaming roost]

[hardy swallow]

Hi! Please check #1397655624103493813

[grand cliff]

is 1.5 bugged rn?

[hidden gorge]

is it possible to get the API

from LMArena

[gleaming roost]

I personally didn't like these A/B options in direct chat; it almost broke the context of my prompt. Luckily, one of the options was literally the same model I was using. I think they did this to get more user votes?

[astral helm]

@hardy swallowit takes only 2 image gens for me and i'll have to wait another 1 hour

why is it changed

it was used to be 5 gens

[marsh hedge]

Claude down?

nvm it's just ignoring my prompts

[echo aurora]

Are you referring to Video gens, not image gens? And is this on the site, not on Discord?

[astral helm]

its on the site and i was actually referring to the image gens

by nb pro

[echo aurora]

The rate limits for each model are subject to change, would also note that if signed in you will have a better rate limit.

[astral helm]

i'm

it was working well just few hours early

[echo aurora]

Can you copy/paste the prompts you suspect are fasle positives in this thread #1447983134426660894

[astral helm]

it was giving 5 gens per hour

now its giving only 2

idk whats going on

[echo aurora]

The rate limit likely changed then.

[echo aurora]

What error message are you seeing?

[echo aurora]

It's possible, but for the moment it's going to be Battle mode only.

[astral helm]

is it going to be permenanat

?

[surreal frost]

smthn about "we've tried every prompt but the page still wont work" smthn like that

[rich mauve]

Okthumbsup

[echo aurora]

I can't make promises on what is or isn't going to be permanent related to rate limits, these rate limits can change.

[echo aurora]

Same on other browsers?

[surreal frost]

I think I should have clarified, it's an issue that already fixed itself when I opened a non-tab url so after I went to https://lmarena.ai/c/ it fixed but what happened was it reset the default AI model for almost all tabs and made them all have a default AI icon as well as show gemini selected at the top

[echo aurora]

Gotcha.

how gemini selected at the top
Yeah this bug is something that popped up yesterday, our team is aware of it and working on a fix.

almost all tabs and made them all have a default AI icon
Do you have a screenshot of this?

[frozen cave]

curious but is smart router in a slow release?

[echo aurora]

We are currently experimenting with it.

[surreal frost]

the tab from today is one I fixed by re-selecting a model and using it

[rich mauve]

🙂

[soft matrix]

some models in direct disappeared

[peak sapphire]

@echo aurora Please add a photo import function to code mode.

[warm sage]

Cannot upload image upload on gpt-5.2-high and gpt-5.2 as of today, can you please fix this

[echo aurora]

Okay thank you fro sharing, I'll pass along to the team.

[surreal frost]

thanks dude

[echo aurora]

I'm on the move!

[echo aurora]

More options for file upload is something we'd like to do. Including more options for Code Arena.

[peak sapphire]

When, approximately, will this be implemented?

[echo aurora]

When it comes to new feautrs/experiments/etc. I won't be able to give some kind of ETA on when to expect them arriving.

[peak sapphire]

I hope the image import function will be implemented very quickly in the arena code, I think it is important

[proud bobcat]

new Artificial Analysis benchmark

i dont know how much i believe it

[stone cape]

yo what is max output limit on lmarena

i got this

prompt: mac os clone in html css and js single html file make all apps work
(hawk max)

[weary thistle]

Hi pineapple. Is it possible to add a category "what's new" so we know what are the new experiments LMARENA is deploying. It will save you time instead of questionning you. And it will keep us updated instead of thinking it's a bug. Even if it's "a posteriori"

[marsh hedge]

when the HELL are we gonna have an edit function 🫩🫩🫩🫩🫩🫩🫩

[weary thistle]

Since The announcements category don't "announce" these updates 🙂

[marsh hedge]

STUPID ASS WEBSITE SENT MY MESSAGE TO A DIFFERENT BOT

NOW ITS STUCK AND I CANT DO ANYTHING

bro ts not funny

[upbeat dune]

Guys, who knows what the best ai is for developing cheats?

You dont even know what game i need it for

[stray aspen]

gpt 3.5 turbo

[upbeat dune]

https://tenor.com/view/balthazar-crazy-family-guy-fou-gif-11775947423386520412

[upbeat dune]

"I so smart so i need to press a reaction"

[chrome goblet]

Fix the website

[keen lake]

@echo aurora i managed to get the api keys for your providers where can i report this

tested direct call and worked

@quasi atlas @granite heron

[radiant iris]

hello! can't wait to try this discord!

[keen lake]

ingest is good but not added right

[quartz light]

really accurate bench from my experience

[quartz light]

20k tokens for me

its like 20-22k max

[stone cape]

💀

[stone cape]

then how the hell is this supposed to see models full strength?

[echo aurora]

Hello, can you send me a DM with more inforamtion?

[echo aurora]

What seems to be the issue?

[proud bobcat]

some of their benchmarks are fudged

especially by gpt 5.2 benchmaxxing

[quartz light]

yea its a huge problem for me

[final lintel]

guys do u see any cars

boom, wrong such good captcha

when i verify CORRECTLY it bugs my captcha

[quartz light]

top right

[final lintel]

how am i supposed to seethat

😭

[quartz light]

[final lintel]

istg bro

cloudflare captcha was so much better than ts

[quartz light]

they use both now

[final lintel]

how do i switch to cloudflare cpthca

[quartz light]

u cant

[echo aurora]

We have been making changes to the captcha systems, it's something we're going to monitor very closely and adjust if needed.

[final lintel]

please switch no human being can do this re captcha

[echo aurora]

Overall would recommend avoiding anything the system may think is odd inauthentic behavior

[quartz light]

they use turnstile for page load, initial prompt and normal prompts sometimes

recaptcha enterprise for creating new chats after requests per _ exceeds somethin

[final lintel]

i asked it "yo"

[quartz light]

ban

lol nice discord flagged quickly

[final lintel]

https://media.discordapp.net/attachments/953387860520734810/1429485894116839434/caption.gif?width=344&height=300

if i ask a ai "yo" im specially autistic

and i get captcha

💔

[fallow geyser]

hmm , at first i thought to make fun of you , but then i saw that others already did that ,

so lemme just give you an actual answer.

grok for asking questions and finding out about techniques and topics.

grok will tell you.
(at least it used to.)

and then once you know what actually to do , you can describe what you wanna do to any model that is good at coding WITHOUT mentioning anything about "cheating" .

thats it.

but vibe coding and cheating dont go together tbh

better to just write it by hand and use grok to find out techniques , documentation , and information in general .
goodluck , its not easy.

[weak finch]

Does Ny one know Maximum timeline of ai video ?

[fallow geyser]

mostly people just have an issue with models getting stuck , and i feel like a stop generation button is all that everyone needs to make everyone happy.

[proven vessel]

👋

[sinful heath]

👋

[hard obsidian]

@everyone

[mossy trail]

👋

[novel saddle]

✊

[fallow geyser]

✊

[regal mulch]

did I dream or there was a video mode on lmarena website and just disappeared

[fallow geyser]

Some people sent screenshots of it , so it really exists.

[regal mulch]

Got replaced by this

There was a limit of 4 generations (so 8 videos)

Oh I have it :

[tired herald]

@echo aurora im not sure if this is a bug or not, but when I change a model and then change back to the original in direct chat after having chatted a bit, it causes the next message to have two models like it is for battle mode. Specifically, the first model of the two models is always the model i previously changed to

so chatting with gemini-3-pro, in the middle, I change to gemini-2.5-flash, then without sending a message i return to 3 pro, i send a message, dual models, first model (left) is always gemini-2.5-flash

[stray aspen]

the automatic model selector just shows up whenever it wants

[unborn hornet]

I am getting a similar error as well. When I chat with gemini-3-pro, after a few prompts the mode is changed to battle mode. In my case, I am not switching AIs during a single session.

[unborn hornet]

^

[tired herald]

Interesting

[novel crater]

I made a text adventure game and now I'm forcing the AI to play it with me through browser tool with antigravity

10/10

[tired herald]

oh no

infinite gen

[half mist]

NO WAY! And I am signed in

A video I made with Video Arena

[marsh hedge]

why grok a p3do

https://tenor.com/view/speed-ishowspeed-reaction-reactions-meme-gif-8141100727479891642

[novel crater]

like father like son?

[marsh hedge]

all his 30 wives was of age chill🥹

[novel crater]

😨

[marsh hedge]

ain't no way he remembers all them kid names tho

[stable obsidian]

hello

what is this new feature "automatically detect modality"

[hollow ivy]

Nice, do you have screenshots?

[rose sky]

How's my AI-generated song?

[hollow ivy]

This is more my cup of tea: https://www.youtube.com/watch?v=iZWTlRgOSzs

[rose sky]

Maybe you will like this... maybe it's like that?

[hollow ivy]

it's a ~5/10
who made it?
heard worse/better

at most, 6/10

if this is by AI, then not bad (for AI), but not good enough to rival with human made music

[rose sky]

It's AI-generated, made by me of course. I'm trying to generate a song that sounds like coming from a liminal space, abandoned malls, etc... I like those kind of stuff

[hollow ivy]

Ok, for an AI-created piece it is quite good.

Gemini, Claude?

[rose sky]

Huh? Gemini? Claude? I generated this with Sonauto AI

[sly minnow]

yo

[gleaming roost]

[slim shale]

Were models removed or is it just me?

[obtuse smelt]

what you choose ?

[slim shale]

nothing, just side vs side

[obtuse smelt]

image or web ?

[slim shale]

neither, I didnt select any mode, like only text, u know?

[lyric sandal]

how much i can generate video

is there any limitation

[obsidian hawk]

GPT 5.2 vs Opus 4.5-32k-thinking vs Hawk Max

[wispy trout]

Is lmarena down?

keep on getting error with generating a response no matter what

Is anyone else’s thing having issues?

[fickle venture]

Gatekeep it bruh

[fickle venture]

Old models yeah they suck these ones are better

[fickle venture]

The heck

[slim shale]

oh yeah but now gpt5.2 is more down on the list, I thought it was removed, aslo for gemini 2.5flash is first than 3.0flash now

[hoary elbow]

What is HAWK max? I’m actually very curious. I saw a video on here showcasing it. Is it better than Gemini

[fickle venture]

It's a secret model not popular but it is better than Claude 4.5 Opus

[hoary elbow]

Oh wow

How can I access it?

[fickle venture]

Dm me

[hoary elbow]

✅

[wispy trout]

So does anyone know an answer to my question

[fickle venture]

?

[wispy trout]

Is lmarena not working for anyone else

It js gives me generation response error everytime

lol

[stone cape]

i swear hawk max should be on the arena this is not fair

[fickle venture]

Idk it happens to everyone tbh

[fickle venture]

It need to be popular first

[golden oriole]

yup, same here, had a nice convo w gpt 5.1h, it broke as always. although usually switching to a different model fixes it (except for the fact the other model might be ineadequate), didn't fix it this time

[stone cape]

https://tenor.com/view/made-by-clown-gif-4428586269206149108

[wispy trout]

Tragic 💔

Grok n Gemini cooked ig

[golden oriole]

has noone ever found a solution for this typa situation?

[fickle venture]

@wispy trout @golden oriole it could just means model overused just try refreshing website or wait a little bit

[stone cape]

💀

ngl hawk max would dominate the leaderboard on webdev arena

[terse shuttle]

fr

[obsidian hawk]

trying this "make me text in particles saying LMARENA that i can control with my webcam zoom in spin etc with my hand.
"

[turbid comet]

Ts happens whenever I try to paste code into the chat

Or in this case API docs

[rose sky]

How's this?

[final hull]

do yall experience error now?

i can't generate image

[placid geode]

the system prompt isn't have any special

😐

[echo aurora]

Would check out this article if you're getting the something went wrong error.

[quaint pilot]

hello fellas how are we doing on this wonderfuk day

[distant idol]

may ik why ts changed, it wasnt like dis

[astral helm]

is there any way to subscribe to lmarena for better limit

[final hull]

I tried different browser but it gives the same error response.

[weak nova]

The color of the website, because it has become a cliche and an AI "giveaway"

[slim spire]

How did you get video

is it available in other countries?

[chrome goblet]

Can you fix the website?

[obtuse smelt]

i reload is error

[devout summit]

did they remove some models from the website?

[obtuse smelt]

no

[tiny jolt]

Let's bump this so the Dev can code on this features

[sweet marten]

How can I consistently use the same model? I'm constantly switching between models now in direct chat model

[supple spindle]

why nbp 2k always cannot generate the images

[sturdy yoke]

Choosing Paper Colors Children pause beside stream, glance at pile of colored paper, debating, medium shot, side angle, soft anime illustration, Ghibli studio style, hand-painted look, diffused morning light, warm and soft, clear weather, gentle breeze moving leaves, shallow depth of field: paper and hands in focus, background slightly blurred, cozy, playful, calm, color palette: bright red, yellow, soft greens, muted browns of stones, scattered papers on stone, children bending over pile, subtle laughter, textures: paper fibers, moss, stone, movement: fingers selecting papers, small shuffle, ambient sounds: [rustling paper], [trickling water], cinematic framing: medium centered on hands and papers.

DIALOGUE: "Okay… Which color should I use?" "Red? Or… maybe yellow?"

[obtuse smelt]

me too but you can retry again
is not gemini too but all model like this

[reef spoke]

https://go.aiforbusiness.com/summit?_go=po9wo6

[little ginkgo]

[inner nest]

@everyoneis there LMArena generation vedio? Because i see LeaderBoard image to vedio or text to vedio

if there is, LMArena is the best AI Models of all

If there isn't, LMArena is the best AI Models for Coding and chat and generation Images

[gleaming roost]

[west lodge]

@echo aurora are multiple output images in one turn supported when using the nano banana pro models?

[west lodge]

there is, but its battle only

and likely will be for the forseeable future

[quartz light]

i made this a while ago can anyone pls try it https://ai.studio/apps/drive/1qbKCuYL1aPB27z_huux-xzFWx4opEEjL?fullscreenApplet=true

i made this with 2.5 pro then iterated with 3 when it came out

[supple spindle]

Already try again for 10 times...

Is it normal

[inner nest]

Ok

[quartz light]

press + top left

[inner nest]

Ok

@quartz light

[quartz light]

aw man

i have that bug on mobile too

not on pc tho

its weird

[inner nest]

And how i fix That

I use in pc ?

@quartz light

[quartz light]

yeah try it on pc

this isnt for video gen btw

but its just my agent thing i made for gemini 3

[inner nest]

Ok

😉

[inner nest]

I know

i said hi, and AI send me a coding and mathematics : )

[quartz light]

😭 lol mb sry im fixing it

its not for conversations

[night ferry]

Hi!

[inner nest]

ok

[quartz drum]

@ornate swallow how you can generate this much video

[quartz light]

updated https://ai.studio/apps/drive/1qbKCuYL1aPB27z_huux-xzFWx4opEEjL?fullscreenApplet=true

[quartz light]

who tf voted deepseek

🤣

[burnt basin]

i do not have the function to do image to video in LMArena

can anyone help

[mortal sonnet]

Greetings! I’m new here, learning and sharing 😇

[peak sapphire]

@echo aurora Why is only the battle mode available?

[wispy trout]

I can’t do anything in lm arena it’s broken

[serene helm]

when is video generation going to be available on the site?

[sand coyote]

deepseek 🔑🔓

[tulip heart]

Guys, could you please recommend some good IDE for ai coding?

[pliant kettle]

Sup guys Just join today

[stray aspen]

all hail anthropic

[keen lake]

antigravity

[keen lake]

or cursor pro

[tulip heart]

banned in my country:(

[tulip heart]

i`m mean free

[keen lake]

wow

[dapper radish]

Hello

[stray aspen]

whats that?

belarus?

[tulip heart]

russia

[keen lake]

opencode?

[tulip heart]

?

[keen lake]

http://opencode.ai/

[tulip heart]

ty ill try it

[keen lake]

you can use free llms with it

[tulip heart]

is full free?

ok

ty

[keen lake]

why is it banned

🥶

[tulip heart]

war

[keen lake]

danmm bro

[safe sleet]

Hmmm, I have a hunch that this server is being used by a certain group of people for coordinated self-promotion and it's slipping under the radar

[tulip heart]

maybe

[keen lake]

true

[tulip heart]

no one knows what could be going on here

[keen lake]

the best

[safe sleet]

Yep, it's always the same group of people posting it that happened to have joined here to almost exclusively talk about it

[tulip heart]

ty

[keen lake]

650,000 monthly devs and i need to promote it? @safe sleet dafuq

😂

[tulip heart]

how to start?

[safe sleet]

Obviously I'm not talking about OpenCode

[keen lake]

everytime i see you @safe sleet you are always trolling dude

[keen lake]

connect provider find a free oai compatible api if you want free bro

[tulip heart]

i connected this

Do I have any limits on how much I can use the claude?

[keen lake]

https://openrouter.ai/z-ai/glm-4.5-air:free

try this

its free

[tulip heart]

ok ty

Do I have any limits on how much I can use the claude?

[keen lake]

you can use openrouter api

and its a good model

[keen lake]

claude i think its paid mostly

they dont give free usage in api

glm air is fast and good and free

[tulip heart]

ty

but

what the

fckj

big pickle?

[keen lake]

😂

its a opencode model

i think

[modern wedge]

Indeed big pickle

[keen lake]

probabbly good

big pickle for the win

but it might not be free forever so use it fast

[tulip heart]

this good?

[keen lake]

sonnet 4.5 level

its very good

[tulip heart]

is paid:(

[keen lake]

glm 4.7 is sonnet 4.5 level

glm 4.7 is god tier

[tulip heart]

I'm very stupid

[keen lake]

its okay we all learning

and thats why we have community

[tulip heart]

why it stops?

[ocean vortex]

Yeah it's impressive, but it also hallucinates a ton

[unkempt elk]

hey guys i need help to make a scraper
can anyone help

[surreal thunder]

?

[clear spear]

what files does gemini 3 flash even support except for image on this thing?

[abstract tundra]

and it requires account?

[clear spear]

LIKE HOW IN THE WORLD IS TXT UNSUPPORTED

[wispy trout]

Idk I

[peak rock]

what I can do with this?
it go to infinity

[unkempt elk]

info?
i wanna scrape mcbuild . org to bulk download minecraft schematics for model training

[ocean bison]

This is why Gemini 3 Flash is dumb.

[stray aspen]

has anyone tried ernie

its really damn bad

[silent mica]

make a good contend and videos

[ocean bison]

I want unlimited with some cheap and fast models like Veo 3.1 Fast

[gilded ledge]

can i get an advice? what AI is the best for upscaling photos without losing face feautures etc? i tried out several of them online but can't come to any conclusion

[ocean bison]

@echo aurora

[stray aspen]

nano banana pro

[gilded ledge]

it tends to reimagine face every time

[unborn hornet]

@echo aurora is there anyway to bypass the direct mode chat turning into battle mode

does anyone know how to just stay on direct mode instead of chat going into battle mode?

[gleaming roost]

[stray aspen]

glm 4.7 is so good

[ebon dagger]

Ny one having issues right now

[fickle venture]

Yes it is

[inland ferry]

👋

[fickle venture]

That's a feature, it mean on the model you used you reached its limit and i am pretty sure that's Claude 4.5 Opus after that limit reached then it switch into battle mode so you can't change it until you wait for the couldown (35 minutes) to continue using the model as direct.

[fickle venture]

Hello

[fickle venture]

Either GPT image or Nano Banana Pro

[sinful heath]

👋

[fickle venture]

Hello and welcome

[fickle venture]

Models aren't free if you want one check huggingface spaces it host opensource models but there might be limits

[fickle venture]

It means the model overload try refreshing website

[peak rock]

Tried a lot, didnt help

[fickle venture]

I mean everyone is reporting it on #1343291835845578853

Just share and talk with them until Lmarena team find a plan to fix this

[neon idol]

@echo aurora On lmarena website the video battle is linitied only at 10 generations?

[elfin harbor]

Direct chat sometimes uses the side by side mode is this normal?

it happens pretty often and each time ask for verification

[barren sorrel]

What to do about it? 6 hours ago it was the same

[hoary elbow]

How do you get smart router?

I can’t find it on the website

[stray aspen]

it just shows whenever it wants to lol

[grand cliff]

The 'This content violates our Terms of Use' sure is unpredictable.

[undone saffron]

Sometimes it may be because the chat tokens have run out or there is too much text between your messages and the AI

[hollow flicker]

yo pineapple is there a way to make it on lmarena so when i do direct chat it disables this dont mind my chats

[undone saffron]

Making a new chat

[ashen plaza]

https://tenor.com/view/beggars-cant-be-choosers-tetsuji-tamayama-kawada-the-naked-director-you-have-no-choice-gif-16551141

[safe sleet]

Idk why the advertising of this model is happening (from the same group of people as always) in plain #general when there's a thread for them here in this server https://discord.com/channels/1340554757349179412/1435953842956013620 (worth a read, by the way)

Of course, if it was a more popular model or a more varied set of peple, I'd understand it could be organic

[obsidian hawk]

Mate this is a place to talk and discuss about models

[obsidian hawk]

can you please keep drama out of here thank you, @echo aurora seems like he want's to start drama again.

[obsidian hawk]

There's a thread for every model family. People still discuss them in general.

[keen lake]

This guy has nothing better to do jheez why is it bothering you so much?

[safe sleet]

I will not respond/escalate this

[pure swan]

Discord got banned in egypt

[safe sleet]

Damn

[urban loom]

Hello

[echo aurora]

I think you both should probably try to ignore each other at this point. If it's not productive let's move on please.

[echo aurora]

There is not. We are keeping a close eye on the experiment though.

[echo aurora]

Thanks for sharing, you may want to try these steps. Bit unclear why Smart Router is erroring out so I'll be sure to report.

[echo aurora]

it's currently set to 3 generations per day.

[echo aurora]

Yup, this is an experiment we're running.

[cold idol]

hello

[obtuse smelt]

hi

[brave cloak]

why do flux-flex images look cooked, anyone else get this?

[crystal wigeon]

Yo guys how can we choose the model to generate video

[inner nest]

Hu

[crystal wigeon]

I dont have command in video arena

[marsh hedge]

y'all tryna sabotage the site?😭

nobody like this feature bro

especially when we have absolutely no way to edit previous prompts in chat🫩

[toxic whale]

got access to a few pre release models, not sure what lab i tried asking what lab made you and it doesnt say

[bleak ember]

hello

[turbid comet]

What’s happening here?

[tranquil burrow]

Same happening with me. So annoying

[obtuse smelt]

need fixing soon

[astral blaze]

LLM peaked at gemini

oh no

[little ginkgo]

Pls file upload 🙂

[ashen mauve]

why light mode that's so illegal

[ocean vortex]

do you have grandpa eyes

[ashen mauve]

yes

[ocean vortex]

profile pic checks out 🧐

[ashen mauve]

exactly

[fleet lintel]

what are some latest possibly SOTA models on LMArena right now?

[ocean vortex]

glm 4.7

other than that your usual suspects

Gemini3, Opus 4.5, GPT5 etc

[fleet lintel]

ohk.. looks like labs will start publishing new testing models from Jan end or something.

[ocean vortex]

It just asked me for captcha just to see leaderboard

LOL

[ocean vortex]

Yeah it's kinda quiet lately

[hearty spire]

Why can’t it be used? It keeps throwing errors.

[ocean vortex]

New Sonnet is rumored soon

[thorn spear]

Video tab have been integrated inside the lmarena.ai website?

[ocean vortex]

Try refresh no cache (ctrl-shift-r) or different browser / logout

[fleet lintel]

new deepseek (supposedly better than Claude in coding) is also rumored soon

[ocean vortex]

amazon-nova-experimental-chat-11-10 kinda looks like it was trained on chatgpt outputs lol

beluga-0106-1 doesn't seem too impressive

[ocean vortex]

Deepseek V3.2 SuperSpeciale 🤯

Knowing their track record of naming they may still not up the version lmao

[ancient cedar]

where can i find the video button?

[slim spire]

how can i access the video arena on lmarena?

@echo aurora how can i access video arena in lmarena?

[remote hull]

captcha problem

I solved everything properly but still it's says wrong

[remote hull]

Click on new chat option. Than you will see video opiton in the chat.

[ancient cedar]

Can u show me bro?@remote hull

[neon idol]

Thx

[viscid cloak]

nbpro infinite generation time+something wrong. PLEASE FIX.

[obtuse smelt]

hmm fix this site

[fossil socket]

I swear this is getting ridiculous

Seriously how have they still not dealt with the generation issues

Whilst simultaneously still adding in more models

Are we legit dead ass rn

[frigid tusk]

is the switch to battle mode going to be removed any time soon or kept?

[high hound]

I feel bad for the people who use LMArena for their work projects

Cant imagine how it feels being cut off at random times

[viscid cloak]

Be careful mate. There would might be some random 🤡(who probably would say has not affiliated with LMARENA) disagrees with you and says nonsense stuff like "wow this site is not for your work. it's for testing."

I have come up with these kind of dudes more than once in this discord. Someone probably really believes that site is alive with all users not using it for personal purposes but for the whole wellness and experiment of this company.

[stray aspen]

lets be reaal most of lmarena users are people who want to use the frontier models for free

in order not to pay a cent

[echo aurora]

You can not, it's going to be Battle only.

[tulip heart]

why it doesnt answers?

[echo aurora]

You're likely being rate limited, but check out this article for some steps that may help + a way to report further information.

[tulip heart]

ty

[echo aurora]

It's an experiment on the site meaning currently it's going to be random if you get the feature or not.

[viscid cloak]

Let’s be real. Not to pay a cent ≠ Free lol. Review how battle works, FAQ, and ToS thank you very much.

[echo aurora]

Do you have the Eval ID for what session this happened with?

[tulip heart]

If there's something important in the chat that I don't want to lose, can I import it somehow?

[echo aurora]

There is not a way to export a full chat. Even though it's not working right now I wouldn't recommend deleting the chat if it's important. A lot of the time I've seen chats start to work again after this bug, it just needs some time (as it's likely rate limit).

[tulip heart]

okay, ty

okay

[warped zealot]

hello

[rough surge]

Do you guys prefer gpt image 1 over 1.5 if so why?

[last thicket]

hi

[coarse cave]

Yeah, that’s probably true. People are always looking for the easiest way to access new stuff without paying.

[coarse cave]

Honestly I’d go with 1.5 over 1 just because the details feel cleaner and the colors pop more naturally. 1 sometimes looks a bit flat or messy in comparison.

[barren sorrel]

Yes, here https://lmarena.ai/ru/c/019ba89a-31e7-7cdc-8221-4e40037d92ba

[lament prism]

i can use it just 5 times ???????????

[meager thorn]

What do you guys think about this UX Fix?

[barren sorrel]

@echo aurora Can you stop my first chat? It has been uploading since January 8th. https://lmarena.ai/ru/c/019b34de-6f36-703e-9ade-c62adbc81234
because of this problem, I already have 4 iterations of the text.

[keen beacon]

Mf owner

[tulip heart]

How long will it take to return to normal?

[echo aurora]

I'm unable to stop chats if they're stuck either infinite generating or if it's getting a Something went wrong error sorry to say. Building a stop button is a feature that's on our todo.

[barren sorrel]

I have sent the complaint to your form.

[wicked sage]

hi guys

[echo aurora]

Generally, a rate limit should last around an hour.

[tulip heart]

I've been waiting for an hour, but the problem is still there

[echo aurora]

The one in this article?

[echo aurora]

Then unfortunately, the problem is being caused by something else.

[barren sorrel]

But site still generating after 5 days...

[tulip heart]

again waiting

[echo aurora]

Okay yeah that's rate limit.

[barren sorrel]

I've been talking to them literally all day and I've never seen such a message.

[tulip heart]

it has send me a erros

and later he says me about rate limit

[echo aurora]

Did you submit this today?

[barren sorrel]

Just when I wrote about it

[echo aurora]

Can you -> Open Developer tools, open Network tab, run a prompt on LMArena that throws an error, back to Network tab search for Stream open that file and look for a Statu Code, what Status Code do you see?

[tulip heart]

but i have limit on 40 minutes

429 Too Many Requests

[echo aurora]

Yeah that's rate limit, you're prompting too often.

[wicked sage]

hi pineapple

i got a question

actually what am i doing why am i not saying it

do you use api keys for your ais or something else

cuz for some reason it feels like you guys dont use api keys

[echo aurora]

Correct

[echo aurora]

How so?

[thorn valley]

@echo aurora 🧐 what happened to the ability to attach images of gpt 5.2 models? it's just a bug?

[echo aurora]

Good question, will flag.

[wicked sage]

oh crap for real?

[wicked sage]

well, for a company like lmarena i thought you guys would just buy/use servers/databases

[thorn valley]

ok, thx!

[echo aurora]

Hmm what do we mean? We do pay for the inference.

[wicked sage]

sorry i dont know how to explain it, my english sucks

you know what, nevermind

sorry for bothering you

😅

[tulip heart]

pineapple, why did I get an error before I had a 50-minute limit?

[echo aurora]

Lol it's no problem, don't hesitate to ask if you have questions.

[echo aurora]

Since you're seeing the status code 429 that means you were seeing the Something wrong wrong error messsage because of rate limit. When it's rate limit the user can see both the Soemthing went wrong + the You have reached your rate limit.. error message. It's basically two seperate error messages that mean the same thing. It's not very clear and we do plan to make our error message system better overall.

[wicked sage]

actually, i do have one

#general message can you fact check if this actually is a real model, if it is can i talk to it?

[echo aurora]

I'm not familiar with that. Not seeing it in Direct either.

[wicked sage]

wth...

[echo aurora]

Do you see it currently in the drop down in Direct or SbS?

[wicked sage]

Cant find it at all, in SBS or in direct.

maybe this is an ultra secret pro ai™ and we dont know

[marsh hedge]

who got lmarena ai alternatives

[wicked sage]

should i put this in bugs maybe

welcome @stray aspen

[stray aspen]

yo

[wicked sage]

w sticker

[steep bear]

can we add ts to vscode because when im tryna do a project with claude opus i have to restart everything because it randomly starts refusing to generate anything for no reason and i have to feed it context which makes it refuse to generate faster (not the rate limit, it just breaks)

[echo aurora]

Where'd you find it? And can I have that Eval ID?

[tulip heart]

wtf

[left lodge]

@echo aurora
What exactly happens when we delete a chat session?
I know it removes it from our sidebar but is it really deleted? Or is kept by the lmarena team and the providers? After deleting can that chat session be included in the release of data from lmarena? Like how datasets earlier were released on huggingface?

[tulip heart]

pineapple pls help

[wicked sage]

apparently theres a dev api model??

[wicked sage]

its at https://lmarena.ai/leaderboard, i am not sure how to find the eval ID

these are its stats tho

[left lodge]

I think that is gemini 1 not 3

[wicked sage]

possibly, its prob something else since we cant see

its not in direct chat, or in SBS

[left lodge]

Its sure gemini-1-pro.

Released as gemini-pro

[stray aspen]

how are you getting this

[stray aspen]

well it sucks

[tulip heart]

pineapple can you pls help me?

[echo aurora]

I'd recommend you check out our Privacy Policy, and if you have followup questions send it to our email via privacy @ lmarena.ai

[echo aurora]

I'd point you to this message -> #general message

It's rate limit ^

[wicked sage]

leaderboards > scroll down > search button, there we go

atleast, thats how i found the dev api model

[shrewd citrus]

what is the second symbol?

@echo aurora

[echo aurora]

It's auto-modality. Currently an experiment, but basically automatically routes you to the modality you're looking for. For example, with this feature if you chat out "a picture of a blue dog" it'll auto route you to Image modality, or "what's the weather like right now in X" it'll route you to Search modality.

[stray aspen]

@echo aurorahow do i get the automatic model router

[shrewd citrus]

Thanks

[echo aurora]

It's an experiment, so it'll be random if you get it or not.

[winter frigate]

@echo aurora Hey pineapple, sorry for bothering if I am, but just one question, if you re-edit your own prompt and make the AI chat again in accordance to your newly editted message, would that count as another prompt that can risk activating the captcha?

[split kayak]

5m

[winter frigate]

peak

[buoyant crypt]

Did something change? GPT-1.5 images look terrible now. Faces don’t look like the reference photos at all, and the likeness is very inconsistent

[echo aurora]

Yes, it would.

[winter frigate]

Ah, okay thanks

[thorny tartan]

#prompt-to-leaderboard red vs daniel tiger

[warm field]

hello, anybody got a feeling what models selector now broken?

I got bad feeling, that gpt 5.2 high not even 5.2

[stray aspen]

what is it then

[warm field]

for example, request: "What does GD 163 break up into?"
5.2 gave the answer that it was stable (wrong of course)
5.2 high gave exact right answer
but now all good and not like it was even several days ago, I think something has been changed near 30 hours ago

I felt it yesterday, when the rules were very primitive, small. and I did it for an hour! although gpt 5.2 high is a miracle model.

[sour spindle]

The llm iq benchmark shows a degradation of 5.2 thinking on the last update/test.

[low imp]

@untold fractal is yelling slurs at me in vc

[sour spindle]

https://www.trackingai.org/home
GPT 5.2 Thinking was 130 on 1/4 90 on 1/11

[heavy ferry]

Has the "creating" error been resolved?

[warm field]

this is not a degradation of the model, look at the top how the top models have changed, this is some kind of glitch of the lmarena

wow wow! now again ok
Изотоп гадолиния-163 (¹⁶³Gd) радиоактивен и распадается в основном по β⁻-каналу (бета-минус), то есть превращается в тербий-163:

The gadolinium-163 isotope (1⁶3GD) is radioactive and decays mainly through the beta-channel (beta-minus), that is, it turns into terbium-163:

10 mins ago was other answer!

this was 10 mins ago
The gadolinium‑163 isotope (1⁶3GD) decays mainly by β-decay to terbium‑163:

its right, but its answer of gpt 4

[sleek phoenix]

Will this merge my chats with an already existing account?

[warm field]

again bug

[warm field]

do you know if there is some kind of system software that allows you to specify exactly which model I am working with on Lmarena?

is there any way to guarantee that you won't switch to another model?

[queen veldt]

Why am i getting violates terms of use on some basic stuff?????

[fickle venture]

Probably

[fickle venture]

That's how it works

Try to make the prompt safer

[echo aurora]

If there are suspected false positives you're encouraged to share them in #1447983134426660894

[faint quarry]

Hello,

[pure kite]

hello

[echo aurora]

[jagged vigil]

hey!

[safe sleet]

Just noticed, there a couple scams buried in #1340554757827461216 (the "chat" counterpart)

[gleaming roost]

[echo aurora]

Thanks, will cleanup.

[lofty sorrel]

Hello!

[grave flame]

hello im Ram from Germany

[small whale]

Hello everyone, I have built an AI receptionist workflow from scratch using chat gpt and n8n. it books calls, answers questions, rescedules, cancels, SMS inbound and outbound etc. I am looking to see if anyone could give any advice or help me through to production with this and i am open to any from of criticism also as i am looking for help for future projects. pls dm me if you have in depth advice as i am looking for friends to be able to talk about AI with. Love to hear back from people.👍

[echo aurora]

Thank you for sharing!

[wraith nebula]

hi

[calm spear]

new feature, smart router

[echo aurora]

yup! What do you think?

[wicked tapir]

where's the smart router exactly?

[stray aspen]

It just comes up whenever it wants to

[gusty kestrel]

hi

[echo aurora]

hello

[rigid copper]

sup

i'm thinking if i'm interested to join arena champion
(nevermind, actually i don't need to, because i rarely contribute)

[whole sundial]

i would like it if the arena champion stuff was promoted more often, i see a lot of people here who i think deserve this role and yet they don't have it because they probably didn't even know about it and how to sign up for it. probably also why there hasn't been any new champions recently.

[rigid copper]

regarding of #january-2nd-contest is it allowed to submit multiple entires of art creation? I see some people submit more than 1 entries (often 3).

[whole sundial]

in the past it was 3 submissions, idk though with the more frequent contests

[rigid copper]

because i mostly submit 1 entries for safety to prevent any unfair advantages

[whole sundial]

also i think it would be a good idea to have contests for server emoji and stickers as this server has a lot of free slots for both

yeah 240 free emoji slots and 60 free sticker slots

[rigid copper]

unless the voting score of 3 submissions from same user will be rounded to average like entries + entries + entries divided by 3

to play fairly, i will just submit 1 entry, otherwise potential consequences.

[echo aurora]

Yup, @whole sundial is correct, it's 3. If people submit more than that I pick the first 3 submitted.

[rigid copper]

ooh, alright thanks

[burnt sinew]

anyone know how to prevent a/b testing in direct chat

@meager glen 🤡 i dont want to get bad models in context

[echo aurora]

It's currently an experiment, there isn't a way to prevent it from popping up. Can you talk to me more about why?

[burnt sinew]

because two models can give seemingly good options, then i pick the one that isnt opus 4.5, then that response is stuck in the memory so i have to tell it in the next response to ignore the last thing it said and copy the response that opus did and tell it that it actually said that instead

[fluid swan]

Hello guys

[slim loom]

Bro any suggestions for building nn8n automation

[night moat]

annoying

@meager glen 🤡

[lucid geyser]

@echo aurora how many submissions for #january-2nd-contest ?

[rigid copper]

they mention 3 so i go with 3 submissions too

[lucid geyser]

oh am i blind

[rigid copper]

scroll up then

[clever bronze]

Hello !

[west drift]

anyone got tips on how to generate 9:16 images using banana?

[rigid copper]

in lmarena.ai website or what?

[rigid copper]

if it's in lmarena website, just specify what aspect ratio you want (type in the lmarena chat), they will adjust it for you.

anyway sup guys

[quartz light]

last year we got claude model in january/feb right?

yeah 3.7 in february

[grand cliff]

Okay. This Terms of Use thing is super annoying

Atleast tell us what is violating it. Not making us guess each time.

[marsh hedge]

https://tenor.com/view/shake-my-head-mark-philips-rdcworld1-disappointed-nope-gif-12480911180809665866

finally jailbroke the models

[grand cliff]

what does jailbreak do again to said Models?

[marsh hedge]

generate what they're not supposed to generate ig

[grand cliff]

....Can you like. Send me the jailbroken 4o latest model? Asking for a friend.

[marsh hedge]

i don't think it work like that😭

but you can check poe.com

[grand cliff]

Worth a shot

[marsh hedge]

they might have a model like that

just search
4o jailbroken!

[night zodiac]

What can I do here

Any activity live here

[west drift]

does anyone know if we can do frame to frame like how veo 3 normally has it?

[keen mason]

I dont know how to create a video

How could I do it??, I thought it was just writing the promt in here

[daring rock]

@keen mason Please head to #1397655624103493813 for a detailed guide on how to use the bot

[coarse cave]

To generate 9:16 (portrait) images using Banana:

Set the right resolution – Most Stable Diffusion models expect width × height. For 9:16, common sizes are:

512x896

576x1024

768x1344 (if your GPU can handle it)

Pass it in the API call – In your Banana function call, include:

{
"prompt": "your prompt here",
"width": 512,
"height": 896
}

Check VRAM limits – Higher resolutions need more GPU memory. If it fails, lower the size or use 512x896.

Optional – Use seed or sampler to control consistency if generating multiple images.

That’s it. Just make sure your width and height follow the 9:16 ratio.

[coarse cave]

If you’re building n8n automations, focus on these:

Start with a clear problem – Automations should solve a real pain (e.g., lead capture, notifications, data sync).

Pick the right triggers – Webhooks, cron jobs, or app events depending on your flow.

Keep it modular – Break big flows into smaller reusable nodes. Easier to debug & scale.

Use error handling – Catch errors with IF nodes or try/catch to avoid flow breaking.

Test with real data – Don’t just rely on dummy inputs; test with live data to avoid surprises.

Document your workflow – Even simple notes help if you revisit it later or share with clients.

Extra tip: Start with simple automations that save time for businesses. Once you see results, you can scale and sell them.

[slim loom]

Thanks u bro

[coarse cave]

Banana doesn’t do frame-to-frame like Veo 3 natively. You can generate key frames and then use tools like RIFE or Flowframes to smooth it out, or try video-focused models like Deforum SD for proper frame-by-frame animation.

[coarse cave]

You can’t just type a prompt to get a video. Generate a few key frames in Banana, then use a tool like RIFE or Flowframes to interpolate and turn them into smooth video.

[keen mason]

thanks but this sounds difficult

[coarse cave]

Ah, gotcha! That’s a very common confusion — generating a single image is easy with a prompt, but making a video is a bit more involved because a video is many frames stitched together. Here’s how you can approach it:

1️⃣ Generate Key Frames

Pick a few important frames of your video idea.

Use Banana (or Stable Diffusion) to generate each frame individually with slight changes in the prompt.

Example:

Frame 1: “A cat sitting on a rooftop, sunset, cinematic”

Frame 2: “The cat looking up at the sky, sunset, cinematic”

2️⃣ Interpolate Frames

Use a frame interpolation tool to fill in the frames between key frames.

Popular tools: RIFE, Flowframes, Dain-App.

This creates smooth motion between your images.

3️⃣ Stitch Frames into Video

Export your frames as images.

Use a video editor or free tools like FFmpeg to combine frames into a video.

Example FFmpeg command:

ffmpeg -r 24 -i frame_%04d.png -c:v libx264 -pix_fmt yuv420p output.mp4

This makes a 24fps video from your frames.

4️⃣ Optional: Use Video-Focused Diffusion

Tools like Deforum SD or RunDiffusion can generate video-like frame sequences directly with a prompt, but they’re a bit more advanced than just using Banana.

💡 TL;DR: You can’t just type a prompt in Banana and get a video. You either:

Generate images frame by frame + interpolate, or

Use a video diffusion tool built for animations.

[coarse cave]

now see if any difficulty tell me

[keen mason]

where is banana? cause Im doing this in the #videoarena1 chat, im making now an image, Ive done a video already didnt get the results i wanted, so now im trying with an image, I think im getting closer. Thanks Amir

[ocean vortex]

can you email me $10k while you're at it

[marsh hedge]

ik u don't believe me but I really did son

🥹

[unreal parcel]

Building personalized web apps

[wicked sage]

hi guys

[hushed gyro]

I got it again @echo aurora

[golden ocean]

in what country u located

my country blocked from getting video option fr or something

[hushed gyro]

Uk

[golden ocean]

ty

[flint sandal]

my only problem with the battle mode is that one ai can randomly get an error just before it finishes writing, and if i tell the ai to continue, the other ai that finished the work is downgrading the result but i wanted to iterate the frist one so i would like to have two chats with these two ais sepperatly because one ai and the second ai have other context of the output

[marsh hedge]

w gunna

[hollow ivy]

-# bump

[calm spear]

Almost always grok, strabge because gemini-3-pro is current leader

[surreal thunder]

Guys why the bots are not working?

It's just showing me generating!

[echo aurora]

We are collecting examples of suspected false positives in #1447983134426660894 , so if you come across one providing the full prompt copy/paste would be helpful.

[echo aurora]

Yeah I really need a better header image for those contest forms

[rigid copper]

maybe next week contest who will make the best LMarena logo for header or server profile (or maybe customize?)

[modern garden]

hello

[echo aurora]

hello

[echo aurora]

I really like that! Make an interpretation of our logo, or make a new one. Mind if I steal that idea?

[echo aurora]

Can you help me understand this better?

i have to tell it in the next response to ignore the last thing it said and copy the response that opus did and tell it that it actually said that instead
Could 4.5 not continue the conversation with the voted on response? Is there some downside to this happening?

[keen beacon]

Hey in video generation bot how much duration can we make an video ?

[late briar]

guy ai art no work

[keen beacon]

Wat

[late briar]

go lmarena

[keen beacon]

U high ?

[echo aurora]

Are others not seeing the models in the drop down? Can't say I'm seeing the same.

[echo aurora]

The generations are going to be around 5-7 seconds long.

[left cypress]

Yo, do you guys know for how much time the webapps in code mode stay online

?

[rotund swift]

hello, I'm here to check out the latest and greatest video AI tools

[echo aurora]

The preview links?

[echo aurora]

Welcome welcome! Be sure to check out #1397655624103493813 and let me know if you have any questions.

[left cypress]

yep

I was wondering

these type of links : https://019bb81e-a36e-719d-8c08-7ceeb1d76523.arena.site/

[echo aurora]

The intent is for them to not expire

[left cypress]

So like i can build a webapp for free and host it indefinitely for free too ?

That feels illegal lmao

(But absolutely fantastic)

Like that's insane, what's the interest of using lovable and these other type of services when LMArena exist

[quartz light]

lmarena video mode now working!

[strong lake]

hi

[echo aurora]

hello

[echo aurora]

Yup! It's an experiment so you may or may not get the feature.

[echo aurora]

It's not being "hosted" in the traditional sense. Why does it feel illegal?

[left cypress]

Wdym not "hosted in a traditional sense" it's up live on internet and that's enough

[echo aurora]

Sure, but isn't done with specific domain control is what I'm referring to.

[left cypress]

Oh yeah right, who cares ?

[quartz light]

it just uses vercel

i believe

[quartz light]

lovable, v0 etc have backends, same url site updates, dont collect nearly as much of ur data n other stuff

lmarena sites are static, they cant be edited after being published, u cant do manual edits, backend stuff

[left agate]

hello

[echo aurora]

Hello and welcome! Would note we ask to use English only when chatting in the community text channels.

[quartz light]

https://tenor.com/view/black-cat-cat-meme-cat-cats-gif-10138630861828915286

nop

[whole swallow]

Where I can see the best webapp builded by AI like really highest quality? I just want to see at what point AI is with good ui/ux

[stray aspen]

claude

[blissful dagger]

man am on a very very big javascript project with claude and it gave me that stupid error

now am stuck and cant do anything

[coarse cave]

🔥 AI‑Generated Website & Web App Showcases

1. WebbsAI Showcase – gallery of websites people have built entirely from prompts with AI, including SaaS landing pages, e‑commerce layouts, portfolios, etc. Great for seeing production‑ready designs.

2. Build With AI (BWAI) Projects – a community showcase of projects built with AI tools and frameworks like Vibe Coding, Supabase, React, etc. You can browse real deployed apps to see how UI/UX and functionality look.

🛠️ Tools Creating High‑Quality UI/UX

Even if not all examples are centralized in one showcase, the output quality of these builders shows what AI can do right now:

• Webflow AI Site Builder – generates professional websites from prompts, with modern responsive UI and editable layouts.
• Meku.dev / v0 by Vercel – prompt‑to‑React/Tailwind apps that look clean and production‑ready, with full code you can refine.
• Div‑idy – AI platform that can generate a whole website or web app from a single prompt, including interactive content.

💡 Inspiration & UI/UX Quality Benchmarks

If you want to see top design quality (not necessarily AI‑built, but relevant as a standard):

• UI/UX inspiration galleries – sites like UIUXShowcase and curated libraries show cutting‑edge modern UX patterns (useful as target quality).
• UI‑Bench (research benchmark) – an academic benchmark comparing AI text‑to‑app systems by visual appeal and design quality, showing where AI still needs work.

[gleaming roost]

[echo aurora]

~~ Are others having issues with the site right now?~~ Nvm!

[little ginkgo]

I am not

Having it rn

Btw can i ask smt?

[echo aurora]

Yeah ofc.

[little ginkgo]

Is there a text limit ?

[echo aurora]

I believe that's going to be model dependent

[little ginkgo]

No i meant text input

And um is there any chance of uploading files feature coming ?

[echo aurora]

Yeah so on our side there shouldn't be a limit for how much you're able to put in the text box, but a specific model may have a limit.

[echo aurora]

This is something we'd like to expand upon and understand why this would be valuable, no ETA that I'm able to share on when to expect more options. What file type would you like to see added?

[little ginkgo]

Maybe .txt

Cuz thats the most i use when i try to code

😁

Pls add it

[echo aurora]

Makes sense! Thanks for sharing.

[little ginkgo]

Pls add it

[little ginkgo]

Pls add it