#general
1 messages · Page 103 of 1
I used it like 5min ago.
also I didnt get nano in like 50 messages
Are you using Text 2 Image or Image 2 Image?
img 2 img
Strange, image 2 image should trigger it more often. I get it like every 2–3 tries.
hello
I just found free grok 4 on Twitter is pretty useful for finding tweets
I just copied someone's x account link, asked "what he thinks about this topic", and grok 4 reasoned very long, used very smart words to search, founded 200+ tweet and listed
output was long too
its pretty useful
In my opinion grok 4 is absolute cinema
hello
hello\

its a full on circus
but its so damn smart
Is it possible to get other ratios than 1:1 in side by side ?
and with the new corn generator lmao
But it's still not that person

Whose answer do you like best?
4
8
tbf we don't have a baseline
when exa made a twitter mcp, the x team suspended all exa affiliates' accounts (and of course ExaAILabs themselves) until they deleted it
they really don't want any other ais using x
Am I the only one stuck on the cloudflare captcha?
It just keep reloading
Nvm, it was my vpn
Any updates about going back online? Thanks in advance!
hi
#video-arena-1 a beatiful woman walking on streets of miami at clear day, beside a beach
hi new here
Hello im here to creat videos
Oi
Wtf is this
Sorry Sorry 😬
Some things are important though. Battery capacity, screen brightness, screen refresh rate can be a deal breaker too, at least for me... Can't stand 60Hz just feels choppy and laggy af in 2025
hey
Then there are plenty of small details also that can make a difference. Thick bezels while they are irrelevant on the surface level, they will affect your whole experience and perception, will also age the phone much faster
Micro LED
680 HZ
Adobe RGB
5000 NITS
12K
300k usd only
Nah you probably don't need more than 120Hz tbh. Then (AM)OLED is fine too - as long as the contrast and colors are good (good quality panel not just the technology of it)
Thank god VPS's exist 😁
For phone? Yes. For desktop, no.
You do not need anything more then 120hz lol. The difference between 120 and 240, at least visually from a human perspective, is simply neglible.
I know, I'm trolling there is no display in market rn with those specifications
there will benefits up to 1000 hz, (one of the reasons) according to blur busters is because of motion blur iirc
I already mentioned battery capacity. Unless it's a budget phone, are there actually screen brightness issues in 2025? High refresh rate just reduces battery life so I don't care
Increasing refresh rate improves immediacy but reduces the “integration window” and its stability. Look at ISO for camera's for example... there's a reason why every competitive gamer caps their FPS usually below maximum.
i recommend reading this https://blurbusters.com/blur-busters-law-amazing-journey-to-future-1000hz-displays-with-blurfree-sample-and-hold/
A Blur Busters Holiday 2017 Special Feature by Mark D. Rejhon 2020 Update: NVIDIA and ASUS now has a long-term road map to 1000 Hz displays by the 2030s! Contacts at NVIDIA and vendors has confirmed this article is scientifically accurate. Before I Explain Blur Busters Law for Display Motion Blur For those new to
I honestly don't understand people who obsess over phone hardware in 2025. Battery life matters. For pretty much everything else, software moves the needle way more
Heck even photo quality is driven by software
Well if you are only looking at the flagships then nearly all should be plenty bright enough. But there are plenty of budget and midrange phones that are compromising on it... You can live with 60Hz? 🗿
/vídeo Prompt Young handsome man with modern aesthetic beauty standards, trendy haircut, well-groomed beard, talking confidently to camera with natural hand gestures, expressive facial expressions, wearing stylish casual business attire, vertical video format 9:16, good lighting, contemporary background, animated speaking with hand movements, professional video quality, engaging presentation style
More Hz are only beneficial to gaming, right? A normal person don't need more than 120 fps.
I just use the mid range Pixel and never notice anything
Difference in swiping and scrolling is massive 60hz vs 120hz. You actually can see a big difference because you are interacting with it, not just watching a movie etc.
🐑
The people obsessing over this 💩 scroll on their phones all day watching garbage. Same people who get Phone upgrades every year.
Personally, I use my Phone until it's last strings when it's not functional anymore. Even if I can afford to buy another one every week.
But there is no difference after 120 FPS to human eye, right?
there is
Oh
Hard to say, probably depends... But at that point it's diminishing returns and not worth it IMO, certainly for a phone
Yeah, you are right.
why
Because of free access to SOTA models?
gemini 2.5 pro SOTA
kind of and i really like the ui and comparing models without api charges
and gemini 3
🔥
where's gemini 3
There is, but above that it’s about motion clarity, input latency, and stability. The brain can pick up differences in blur and reaction time even past 200 Hz.
not for you
why not?
/video
/video animal cow furry anthro
o hell nah
/prompt
is it only me or deepseek r1 are gone?
@alfadl0607
/video
YO WHAT
WHAT
But my chat history of lmarena still gone 😭
What was my fault man
Yeah, but privacy-wise it's not the same, since IP might not rotate + only 1 person uses it.
A fluffy grey tabby cat with thick fur, holding a small bouquet of flowers, walking shyly down the colorful street. His big expressive eyes show both nervousness and excitement. Pixar-style 3D, cinematic close-up
A fluffy grey tabby cat with thick fur, holding a small bouquet of flowers, walking shyly down the colorful street. His big expressive eyes show both nervousness and excitement. Pixar-style 3D, cinematic close-up
If you want internet “privacy” VPNs aren’t it 👍🏼
Bro see this 😭
Video Arena Bot is Currently Not Working
@wooden cobalt The Video Arena Bot is current not working. The only channels the bot work in are #video-arena-1 #video-arena-2 #video-arena-3 ; however, the bot is currently offline. When it's working again, go to the channels listed above to prompt.
The reason you don't have permission to type in those channels is because the bot is down.
A fluffy grey tabby cat with thick fur, holding a small bouquet of flowers, walking shyly down the colorful street. His big expressive eyes show both nervousness and excitement. Pixar-style 3D, cinematic close-up
I heard it's only available in India, and no thinking mode.
Prompt (Realistic Natural Video – Arabic Speech):
Ultra-realistic short video of an old Iraqi woman in traditional Iraqi clothing (black abaya and headscarf), holding an eggplant. Natural village background. Her face is wrinkled and expressive. She speaks in Arabic (Iraqi dialect), saying:
“اسود سواد اللييل وبعيني تبني
واحد من أهل البيت حسبالك ابني”
Lighting is natural and realistic, with authentic ambience.
⸻
تحب
Also is this real?
Eligible ChatGPT Plus users can generate a limited number of unique referral codes to offer free trials of the Plus plan. An email with a referral link gets sent to anyone you invite.
New users can sign up for an OpenAI account and use the code to activate a free trial of ChatGPT Plus.
A fluffy grey tabby cat with thick fur, holding a small bouquet of flowers, walking shyly down the colorful street. His big expressive eyes show both nervousness and excitement. Pixar-style 3D, cinematic close-up
@wooden cobalt do you not know how to read english?
There was quite a gap in between 1.5 Pro (Feb 2024) and 2.5 Pro (Mar 2024), hmm...
@keen beacon
I see...
Not sure
Oh, you need a referral code?
Yes
WHAT THE ###### ############# ########## #############
Hmm.. I don't see an option
Yupp ai is scam
really?
You can compare their gpt 5 high and lmarena's gpt 5 high
Yes
"A realistic dramatic scene in a cinematic style: an elegant wooden table with a large metallic nail clipper gleaming under strong dramatic lighting at its center. Behind the clipper stands a small wooden ladder, on which a tiny elephant balances unusually on its upper steps, preparing to leap forcefully onto the metallic lever of the clipper. In the foreground, a realistic human thumb appears, with its nail positioned inside the clipper’s opening as if the decisive cutting moment is about to happen. The atmosphere is tense, resembling a movie poster or a commercial shot."
didnt expect deepseek to cook this hard
@stray aspen
whatsup
It's because of tokenization:
Then how does gpt 5 high gets it right
guys, when are we gonna get some new models
when there are new models
/"A realistic dramatic scene in a cinematic style: an elegant wooden table with a large metallic nail clipper gleaming under strong dramatic lighting at its center. Behind the clipper stands a small wooden ladder, on which a tiny elephant balances unusually on its upper steps, preparing to leap forcefully onto the metallic lever of the clipper. In the foreground, a realistic human thumb appears, with its nail positioned inside the clipper’s opening as if the decisive cutting moment is about to happen. The atmosphere is tense, resembling a movie poster or a commercial shot."
We just got them
Video Arena Bot is Currently Not Working
@plucky glade The Video Arena Bot is current not working. The only channels the bot work in are #video-arena-1 #video-arena-2 #video-arena-3 ; however, the bot is currently offline. When it's working again, go to the channels listed above to prompt.
The reason you don't have permission to type in those channels is because the bot is down.
wasn't opus 4.1 a long time ago
Except opus
Qwen was also a long time ago, deepseek was rather recent and gpt 5 in my eyes is long ago as well
Deepseek 3.1 bro
Both OpenAI and Meta tokenizes individual digits I think: https://www.beren.io/2023-02-04-Integer-tokenization-is-insane/
After spending a lot of time with language models, I have come to the conclusion that tokenization in general is insane and it is a miracle that language models learn anything at all. To drill down into one specific example of silliness which has been bothering me recently, let’s look...
Not sure about Anthropic
R2 gonna destroy altman
Gpt very bottom
Gemini 3 and deepseek r2 at the top competing head to head
I'm telling you gpt 5.1 will release a week after gemini 3 or r2
they literally removed R1 for some reason
Actually, I don't think that's the reason, since:
It's available on the deepseek web and apk use it there
no they replaced it with V3.1
V3.1 can reason, right?
Wait yeah
@verbal nimbus
"Thinking"
yeah so on the webui, it's using V3.1 and not R1
@stray aspen did you face any problems with context window or smth?
@rustic knot there is no system prompt setting in deepseek web
Or there is
?
idk, haven't checked recently
how good is deepseek V3.1 at agentic coding would u guys say
Haven't tried
@verbal nimbus
DS claims it has stronger agent skills though
Hello!
Last I used it, it doesn't support custom system prompts
So I have to paste my system prompt with every query?
They probably have their own system prompt, so yeah.
You can use it via a free OpenRouter endpoint (if it's up yet) if you need more control. Free endpoints are logged though.
It's very low temparature
Very rational
DeepSeek 3.1 just feels like a slightly worse version of Gemini 2.5 pro just from doing some text prompts
Still way too much “not x but y” slop for my liking
Why only gemini 2.5 pro can watch yt videos
the ranking numbers are wrong - it shouldnt go 1, 1, 1, 2, 2, 3, 8, 8, 8. it should go 1, 1, 1, 4, 4, 6, 6, 8, 8, 8
the algorithm that determinese the "rank" number is bugged
They are both Google products, and so Gemini probably has access to the internals of some Google services
SWE-bench team are already testing V3.1 on bash-only mode: https://github.com/SWE-agent/mini-swe-agent/issues/458
They said that the thinking variant doesn't support tool calls
so the base without thinking does support tool use while the thinking doesn't
I'm worried they'll make Deepseek into coding model and forgo its other strenghts 😭
I want a model that's cool at everything instead of just coding slop
Otherwise nobody will ever make Sama and OpenAI learn their place
deepseek was never really specialized towards that anyways. It was always anthropic. From R1, it seemed that it would make sense for deepseek to be specialized towards math and science
GPT-5 is the only (!) model that answered my obscure anime music theory knowledge questions correctly, why do wouldn't I want Deepseek to be able to do this too
I was really impressed when it iterated over like a hundred of titles and figured out the correct one
Nobody knows what OpenAI did to it but it was genuinely impressive
Know your place, you fool sam
— gemini 3
1 trick pony
As far as I remember, any training on general knowledge improves LLM's abilities, regardless of the sort of data it trains on. So we need generalist models, not big sloppy fine-tuned ones
Qwen3-Coder is the best open source model for coding so far, scoring around 50% at Brokk power ranking bench and around Sonnet at SWE Bench. It still falls flat before more general models
l
gpt-4-preview
The new deepsek is nice
I asked u smth
3d
I still want the R2 and V4 😕
You teased me to try it 🥹😁
He'll yes
how deepseek compares to gpt5?
LM Arena Prompt for Video Generation
Title: The Lion King: Simba’s Journey
Duration: ~2 minutes
Style: Cinematic, realistic CGI, dramatic lighting, emotional atmosphere
Voiceover: Deep warm male narrator (English)
Music: Orchestral soundtrack, emotional transitions
Aspect Ratio: 16:9
📝 Scenes Breakdown
[0:00 – 0:08]
Scene: Wide shot of the African savannah at sunrise. Animals gather near Pride Rock. Golden light, peaceful atmosphere.
Narration: "In the heart of the savannah, a young lion was born… his name was Simba."
Sound FX: gentle wind, birds chirping
[0:09 – 0:20]
Scene: Rafiki the wise baboon lifts baby Simba high on Pride Rock. All animals bow. Cinematic wide angle.
Narration: "Presented to the kingdom, he was destined to be king."
Sound FX: crowd of animals cheering softly, wind blowing
[0:21 – 0:34]
Scene: Scar lurking in the shadows, green eyes glowing with jealousy. Dramatic close-up.
Narration: "But in the shadows, Scar—his uncle—plotted a darker fate."
Sound FX: low growl, ominous drums
[0:35 – 0:52]
Scene: The wildebeest stampede. Simba clings to a tree branch. Mufasa leaps into danger to save him.
Narration: "Through betrayal and tragedy, Simba lost his father… and his kingdom."
Sound FX: thundering hooves, roaring lions
[0:53 – 1:10]
Scene: Simba wandering alone in the desert, then meeting Timon and Pumbaa in a lush jungle. Lighthearted vibe.
Narration: "Alone and afraid, he found friendship… and a new way of life."
Sound FX: laughter, tropical jungle sounds
[1:11 – 1:28]
Scene: Adult Simba looks at his reflection in the water. Suddenly, Mufasa’s spirit appears in the night sky.
Narration: "But destiny cannot be forgotten. Guided by the memory of his father… Simba chose to return."
Sound FX: mystical wind, echoing deep voice
[1:29 – 1:50]
Scene: Epic battle at Pride Rock. Simba fights Scar under lightning and storm. Dramatic angles, fast cuts.
No
is gpt5 cooked
howdy
nano banana
What’s the coding performance of Deepseek V3.1?
i think he is teasing that the release for this model is on 3 days
what do you mean by major risks?
Whats the différence?
oh interesting. very close as in how many days?
very nice, i hope it's on the 25th, but where is it going ti be available, ai studio?
Hello, how do you get permission to create videos?
can a help me?
Ok
great! thanks for the info, i am really looking forward to use in AI studio next week
اختصر البرومبت
إليك نسخة مختصرة ومباشرة للبرومبت:
Short Video Prompt:
Create a professional animated video showing gleaming solar panels on a lush green farm. A modern, solar-powered water pump efficiently irrigates the crops. The GPS company logo appears elegantly on a panel or in the sky with a subtle glow.
Style: Bright and clean aesthetic with vibrant colors (greens, blues, golds) and optimistic background music, emphasizing the harmony between technology and nature.
i am here to get ai videos
/video
pls help how can i get permission to create ai video
u got to say something positive about me first
/video Calmn opening Generate a
10-second nature intro shot of a calm river in
a natural setting. Slow dolly or steady camera
with a soft focus on the water surface, light
ripples, and surrounding trees. Gentle natural
SOunds (birds, distant water). Subtle title
appears gradually: "American Beaver Life", 4K.
30fps, color graded for natural tones, wide to
medium crop.
Hi everyone! Hope to learn more from everyone
this video here is educational
⬆️
{
"shot": {
"composition": "Tracking medium as she boards; gimbal as she sits; stable OTS at window",
"camera_motion": "handheld→gimbal→locked OTS",
"frame_rate": "24fps"
},
"subject": {
"description": "Calm young niqabi girl",
"wardrobe": "Loose black abaya, long niqab; small handbag"
},
"scene": {
"location": "Public bus interior",
"time_of_day": "late-afternoon golden light"
},
"visual_details": {
"action": "Boards, finds window seat, settles, takes phone, plugs wired headphones",
"props": "phone, wired headphones, handbag"
},
"audio": {
"ambient": "low engine, footsteps, faint chatter, seat creak",
"voice": null
},
"visual_rules": {
"prohibited_elements": ["subtitles","captions","text overlays","logos"],
"policy": "No Captions"
}
}
When video bot I'll be back?
Helo there
helo
How I can generate free videos ?
/image to video
hello
baby in the flowers #video-arena-1
💀
Prompt (for VEO 3, 8 seconds, humorous couple dialogue):
Create a short, 8-second humorous micro-scene for an AI video using VEO 3. Setting: a cozy living room at night after a party; warm, soft lighting; a small sofa and a coffee table with an empty bottle hint. Camera should use a couple of quick, playful cuts to match the quick pace. Characters: MARIA (female, early 30s, stylish, practical, calm but with a spark of wit; shoulder-length dark hair; wearing a neat blazer over a casual top) and JOÃO (male, early 30s, visibly drunk, slightly disheveled hair, wrinkled shirt, slouched posture, holding a cup). The narrative: MARIA ends the relationship with her drunken partner. Tone: light-hearted, witty, playful, not cruel. Language: The dialogue lines must be spoken in Portuguese, and should appear as on-screen dialogue with character names in bold or uppercase. Provide the exact dialogue lines in Portuguese in order, designed to fit the 8-second duration.
Dialogue (Portuguese lines to be included on screen):
MARIA: "João, não aguento mais essa bebedeira."
JOAO: "eu bebo para esquecer meus problemas!"
MARIA: "Quais problemas?."
JOAO: "chifre que você coloca."
shut up
this channel is for yapping, video generator channel and bot is down
.
@echo aurora should make the #video-arena-1 stronger when a new member come
Maybe force direct them to read the #1397655624103493813 then answer it?
😊
make them go thru an iq test
💀
b--bb--b--bbbut its consistently failing to grok right now
grok is even worse 🥀
well, its consistently better at making userscripts
🤓
that's really comforting for Elon
His model is good at userscript
👍
Hello I am here because I like to make videos. I swear I am a human.
bro is a clanker
hello
where is my award winning game asset?
Why is it always "Failed to accept terms of use"??
heres what nanobanana generated
mj 7
yeah nano is more for editing
{
"description": "Cinematic ultra-close-up of a frosty Coca-Cola Zero can on a sleek futuristic pedestal in a high-tech plaza. The logo glows with a red pulse. The tab clicks open in slow motion, streams of sparkling liquid light spiral out, transforming the city. Skyscrapers animate with giant LED screens showing Coca-Cola Zero visuals. A holographic stage emerges mid-air, crowds appear with AR headsets, dancing. The ground becomes a glowing grid in sync with the beat. Drones release confetti and lasers as the city shifts into a Coca-Cola Zero–fueled digital festival. No text.",
"style": "cinematic, dynamic, magical futurism",
"camera": "ultra close on condensation dripping from the can, zooms out and orbits as the cityscape transforms",
"lighting": "daylight fading into vibrant neon reds, blacks, and silvers—cyberpunk festival glow",
"environment": "futuristic plaza transforms into a city-scale holographic party",
"elements": [
"Coca-Cola Zero can (logo glowing, condensation detailed)",
"slow-motion tab opening with light burst",
"liquid light spirals changing environment",
"LED skyscrapers with Coca-Cola visuals",
"holographic stage forming mid-air",
"AR dance crowd moving to the beat",
"glowing grid floor synced to rhythm",
"drones releasing confetti and lasers",
"dynamic screen transitions",
"virtual fireworks lighting sky"
],
"motion": "chain reaction from can opening—liquid energy flows, triggering rapid city transformation",
"ending": "Coca-Cola Zero can foreground, city in full festival mode behind, pulsing with light and music",
"text": "none",
"keywords": [
"Coca-Cola Zero",
"urban festival",
"futuristic party",
"city transforms",
"dynamic animation",
"holographic concert",
"hyper-realistic",
"cinematic",
"no text"
]
}
A, bot doesn’t work?
i think te bots down m8
@limpid schooner another one
hello bro
affirmative, dave. i read you.
yo
Hi
Does anyone know how to put nano-banana in the solo or vs. options? It doesn't appear in any other options
pre-banana and post-banana
You basically roll the dice with getting Nano-Banana to appear
Oh, can't I use it individually? Too bad, it's the one that gives me the best results 🙁
/image-to-video image: [archivo adjunto] prompt:
The same happens with me .
Is grok good at chat?
Yes 👍
/video animate this image
Video Arena Bot is Currently Not Working
The Video Arena Bot is current not working. The only channels the bot work in are #video-arena-1 #video-arena-2 #video-arena-3 ; however, the bot is currently offline. When it's working again, go to the channels listed above to prompt.
The reason you don't have permission to type in those channels is because the bot is down.
Yeah definitely going to look into ways to make eduction on how/where to use bot better. I’m sorry everyone, it’s not great for our text chats.
Maybe the server guide that we see upon first joining should have "Read the Rules" first in the list instead of last, heh.
Yeah good callout. Maybe review #1397655624103493813 instead of generate a video too
hey @echo aurora my chat history is gone.. like all the chats... wtf?
is there anything you can do....
Really sorry for that. Unfortunately there isn’t anything we can do to get those chats back. We are working on a feature that’ll help with this though as it’s really clear to us how frustrating it is to no longer have that history.
whaaat..... aww man i really did work hard on those prompts
what ai app that uses nano banan?
What do I need to do to be able to send prompts on "Video-arena" channels?
It’s currently not working sorry to say. The announcement channel mentions what happened. The team is looking into
ok, thanks.
It feels like flux regressed in quality after the outage
like before it could generate stuff with pretty good accuracy, now it's like it's running off of stable diffusion's default settings
Well, there used to be two different sites: LM Arena and Yupp.ai.
But, for some odd reason, Yupp removed Nano Banana, so now, only LM Arena has it.
What was nice about Yupp is that you could technically always use the model alongside the others, but you'd just have to vote for it in order to gain more credits.
Other than that, though, it was basically free and unlimited.
LM Arena is more of gambling to get it, but the upside is that it's infinite use, and it's free.
I think teach them to successfully do it would be better than just "reading the rules/instructions"
Is this a bot issue or a me issue?
It's coming back online right now, can you try again?
never mind, I think the message was deleted
salveee
Well, seems like the bot is gone again.
I'm running into errors too.
This is why I copy and paste all of my prompts onto a Notepad file so I have them forever
morning! Anybody played with deepseek new version? Didnt have the chance yet and im curious what's the vibe with it
yeah bro... i'll now make sure to save my prompts
Nano-banana is on Yupp again
yep using it rn
I've been saving prompts for the past 3 or 4 years 😉
Oh, neat.
u must have a huge library by now ig uess
Well, yes and no
Notepad files aren't that big
I have tons of prompts but the files themselves take up extremely little space
I've made hundreds of thousands of images locally
Because I have a 4080 RTX
Bought it specifically for AI stuff
nah i meant... u would have a tons of prompts saved... so like numbers
how do u remember .... where to fetch what?
I'm not getting anymore credits
why?
well i used another account and made it public *so it's using halfcredits
They said they can track that and ban you
They can find out if you have multiple accounts
Chat, good news.
You don't actually have to use Yupp anymore.
Seems like I just discovered another free alternative.
nah i meant a new google account that am using and when u make your generations public it uses half of the credits
and what's that bro
I think that site is fake
is it legit?
Yeah but 'pricing'
I'll pay for the official nano-banana
It's launching in 3 days
Hmm…
really.... how do u know that?
Okay looks like Video Arena is back btw 👍
first hi from me
so like where will it be available on...gemini or studio ai?
It's not called “Huzzah”, I just embedded it.
Oh haha
It was a hyperlink, heh heh.
Yeah I think that site is a scam
@potent glacier ...?
I don't know?
I shared the X link
That's what they posted
just a image? what
So I guess in 3 days we'll find out
who's the google employee?
hey i see notebook lm down there ... maybe it'st he presentation made from it
@NotebookLM lead @GoogleLabs + AIFF
🕒 NBLM Audio Overviews, Gemini & PaLM2 post-training, AI Studio, YouTube, Discover, Search, Android.
More LaMDA moments
He's the one
let's hope that your 3 days interpretation is right
Yayyyy
That's what others are saying
👀
It's working for you too, yeah?
Haven't tried it yet.
Will try later
A bit busy rn.
hey what's the best ai for 2d characters and animations
If you can gen locally, one of the Illustrious or NoobAi models
Hello, does anybody know of an AI which I can generate videos like this one?
idk bro... who's the ytber
@potent glacier do u know this
Damn
This isn’t an appropriate question to ping our mods about.
Our terms of service is where you can find information about commercial use.
Where is your term of service can you guide me ?
@pineapple.___.
Yes.
Could you send me it?
Absolutely.
Thank you
I'm new here. Where do I report technical issues? For some reason I am unable to vote on a particular set. I'm still able to vote on others though.
"I still haven't understood, even after reading everything. Just tell me this: can I use these videos and images for commercial use on stock image and video websites like Adobe Stock and Wirestock.io? Please just clarify what I should do so that I can get an answer to my question.
@pineapple.___.
You may do so in #1343291835845578853 .
Also, my pleasure.
Thanks! 😊
No problemo.
Is it just me or is LMArena more rate-limited now?
hlo
@echo aurora ?
Is it just me or the LMArena website is asking for human verification too often after 2-3 min of inactivity?
Is there a way to use nano banana specifically to generate image based on given reference instead of battle?
How to use nano bananas? I am getting results from different models
Hello guys, Legacy Web interface dead now forever?
don't know, first time here. I am apologize if Im doing wrong
Yea
thank u! Sad to know it
Guys I need 1 vote to reveal model name
Please do 1 vote here
Hi~
yes
Thanks guys
Al Qadsiah Club logo, majestic and grand design, 3D style, luxurious golden lighting with a soft glowing effect, dark black background with light mist for a sense of power and legacy, highly detailed, ultra realistic, high quality
❌ Image must be at least 300x300 pixels. Please upload a larger image.
@chilly rock Please stop generating videos to the main chat
@echo aurora Oh no they are flooding #general
They are supposed to be in #video-arena-1 to 4
Where is @echo aurora when we need 🍍
This place is a mess, needs more admins
2 (i think) mods were fine when there were only a few thousand of us, but with the massive influx of new people there should be several more of them
like some of these new people just created a discord account so that they can use the bot here and many speak poor English or no English at all, the new mods should definitely know English well.
why arena site stop drawing, only text answers?
make sure you have selected image generation mode
sometimes the site will reset to text mode
thx
wow, 2.4K folks online. No wonder there is so much craziness here. Need some quality control
A close-up view of a person's heart, which is transforming from stone to a vibrant, pulsating organ filled with light, from the inside out. The transformation is soft and gradual. cinematic, ultra-realistic, 4k.
wrong place, use /video in #video-arena-1 and read #1397655624103493813
those new joiners give the impression of being bots for some reason… @echo aurora Maybe create a new server for older members or disable generating outside video channels?
@drifting fulcrum's Prompt:
A magical, girly autumn forest scene with glowing golden sunlight streaming through tall trees with colorful fall leaves in shades of orange, red, yellow, purple, and pink. The forest floor is scattered with bright autumn leaves, and in the foreground are large, beautifully illustrated pumpkins in orange, teal, and gold tones surrounded by glowing lights and magical sparkles. The environment has an ethereal, fantasy-like glow, with warm sunset lighting illuminating the path through the forest. The words “Autumn Girly” appear in elegant, whimsical lettering at the center, integrated harmoniously into the scene. Soft painterly art style, high-resolution, vibrant color palette, no watermark, no mockup elements.
Generating images. We'll notify you when we're done.
new server is crazy
disable is better
uh hello, do you all know why my images just dont wanna display? sorry if its the wrong channel, but i just dont get why its not working 
🤔
THEO LEAKED THE MODEL HE WAS GIVEN
Er--
idk if its public info
Chat, I think we may have a new model on our hands.
what?
im talking about last month
Ah.
bottom is cursor, top is normal
What a dumbass
Theo
why
Idk maybe it's just another strategy from OpenAI to make hype
nuh uh
actually
hes saying the old model was better
slightly
We do? Kinda?
i just tried to log in on his website he been advertising for so long.. what a great first impression
umm why does it say h2h on top
What even the point of his website, there are dozens of API wrappers and inference provides. Who ever in the world uses it
i think the price is decent idk
like for the first month
1$
OpenAI's prices are decent too, they have the best performing model across a range of domains for the least money ever. For example brokk.ai/power-ranking
I see no point in all these stupid wrappers
Unless I myself am stupid...
thanks for the site
looks super accurate to my experience with the models
Same
no grok 4 though
They were unable to get a reasonable quota o_o
Which means that it is probably even more expensive than new Claude
Does anybody know? Could it be browser issue or maybe a specific model i gotta use?
Hello,!
3 means gemini 3
HELLO!!!
All the channels are ruined lol
Hi I have a question can you generate a photo with a reference image here?
have you found a statistical preference to voting "left is better" or "right is better"? Not that it matters in terms of fairness considering random position, it's just a curiosity.
How to get the best results from video models?
hi all!
Should we still vote if one or more of the generations fails? Does that help find issues in the models, or does it just throw off the ratings?
my guess is that it's fair, because failed generations should be punished
I think some also fail for copyright or censorship reasons, which in my humble opinion should also be punished 😄
Server gained a lot of members recently,it's a mess
Uhh that depends. I would say no. Sometimes errors are out of the control of the labs. Can be related to lmarena server or their front-end issues. Or even your own browser/device.
Yeah after they take video arena and put it here it's a mess
Like I saw guys speaking Arabic or some other languages,dude cool language,but STILL YOU SHOULD SPEAK ENGLISH ITS AN INTERNATIONAL SERVER
Midjourney ended up moving away from Discord so I guess there are maybe some technical issues with it.
Btw what does the star mean? And what if both are good/ it's a tie?
@green plume 😡
ITS NOT FUNNY
bro
im with you too
😡
my whole game gone
all my ideas
for models etcc
Then mr dev bring back our chat histories
im cooked
I'm what
yo
That's for the highlight channel. More info on how it works here: #announcements message
Supp guys
gurt: yo
Hello and good time. Is there any limitation in characters, chats and Submitting files in every day or specific period of time per day for using this website?
This channel #nano-banana for what?
Hi
That channel is for testing out the nano-banana model from LM Arena.
Also, hi.
Can I testing too?
No that channel is for talking about nano-banana @outer estuary
You can only test nano-banana on our site (https://lmarena.ai/?chat-modality=image) in Battle mode
Yes.
Anyone know?
Just like this guy said. ^
is bluedog a new mod?
Certain models allow you to use them infinitely without any usage limit whatsoever.
Yeah, we have new mods starting.
i see i see , congrats
Thanks for answering. So how about chat gpt 5?
ty! I'll be putting out an announcement once mods are settled a bit more to remind folks where to go with issues.
GPT-5 allows you to use it infinitely as well on LM Arena.
From what I've seen, it doesn't seem to have any rate limit.
You new ??? 😁You are old here and maybe older than me
Er-
gng someone take the video bot ping role away 😭
Heh.
pinapple
i got question
what happend with the
chat history
is it gone forever ?
wait what happened to voting in direct chat
like saying "this is good" or "this is bad"
it was in legacy
To answer your question, when it bugs out like that, it deletes your chat history forever. Unfortunately, there isn't currently a way to restore that. But they are working on a way to.
Hey sorry to say if your chat history is gone you won't be able to access it again.
oh
bet
😄
and did you fix it ?
or it can happend somewhere in the future ?
This isn't something that was brough over to the current site. We are exploring other ways to provide feedback in direct chat so it may get that feature again one day.
oh okay
like you could
add gmail
um login
Sorry to say it can still happen, but we are getting a lot closer to fixing this.
ok
thanks atleast for um 😄 answearing
W you
Currently I stuck here
Send a recording of what happens.
No matter how many times I try, this checkbox keeps reappearing
Hmm...
Are you using a VPN? Same problem with a different browser?
Yes. Let me check it with Firefox
image to video not working
I know that a certain issue happened before on LM Arena, but on Chrome.
It's work
But then the person who had the problem switched to Opera, and the website started working just fine. So maybe they can use Opera.
Yeah look into
Actually it's happened on chrome
Yes, exactly. It seems like many issues on LM Arena happen specifically on Chrome, but then when people use a different browser, it suddenly works just fine.
video arena ded again ?
Seems so.
Hey everyone - the Video Arena is down again, we're working on fixing asap.
Oh, well.
Thank you very much!
great
Can we make image to video with video arena?
Yes.
How?
You can do so via the "/image-to-video" command.
However, it only works within #video-arena-1 (or any other video arena channels).
Is there any limit to use it?
Yes.
How much?
I remember seeing something from the bot that said it's limited to 8 video generations per day.
8 video per day
Ohk thnx buddy
My pleasure, Amaan.
Thank you all for providing the info 
Our pleasure, Pineapple.
It's always nice to help out the mods every now and then, too.
Hello, I have a problem why I am still jumping the Something Wrong While Generating The Response. Please Try Again.?
After all, they're also humans just like us.
Try looking in the #1343291835845578853 channel and see if there is already a solution to that.
Not problem ☺️
WHO
Maybe not the bots, but the mods for sure.
True.
Where are you seeing that?
in their profile
Ah.
They probably just decided to change it to that to show that they're an official LM Arena mod.
Unless it's mandatory.
Anyone know?
Would you mind posting about the issue in #1343291835845578853 ?
?!
Sure
Anyone here use LM Studio. I have a question. I installed the application on my D drive, But the app is downloading data to my C drive appdata lmstudio\extensions\backends. How do I force it to use it in portable mode. So that it doesn't use my C drive for storing data?
Why it happens
- LM Studio stores downloaded models and backend extensions inside your user profile directory (typically
C:\Users\<You>\AppData\...). - This behavior is hardcoded—so even if the main executable is on D:, the data path defaults to C:.
- Unfortunately, LM Studio does not (as of now) have a neat "portable mode" checkbox in settings.
Ways around it
-
Set an Environment Variable
- LM Studio respects a variable called
LMSTUDIO_DATA(orLMSTUDIO_HOMEdepending on version). - You can create a folder on D: (say
D:\LMStudioData) and then set a system environment variable pointing there. - On Windows:
- Press Win + R, type
sysdm.cpl, hit Enter. - Go to Advanced > Environment Variables.
- Create a new User or System variable called
LMSTUDIO_DATA. - Set its value to
D:\LMStudioData. - Restart LM Studio and it should start using that new folder for model storage.
- Press Win + R, type
- LM Studio respects a variable called
-
Symbolic Link Trick (if the variable doesn’t work)
- Move the folder from AppData to D:.
- Then create a symbolic link so LM Studio still thinks it’s writing to C: but the data actually lives on D:.
- Open a Command Prompt as Administrator and run:
mklink /D "C:\Users\<You>\AppData\Roaming\lmstudio" "D:\LMStudioData" - This essentially creates a teleportation tunnel: the app writes to C:, but it all gets shunted to D:.
-
Check LM Studio’s Current Docs/Discord
- The developers are constantly tweaking backends; sometimes they add official flags in newer builds.
- They’ve mentioned plans for more flexible data paths, so keeping an eye on release notes is worthwhile.
Quick Recap
- There’s no single “portable mode” toggle yet.
- The environment variable (
LMSTUDIO_DATA) is the cleanest solution. - If all else fails, the symbolic link method always works.
Neat thanks @robust yoke
My pleasure.
Hello
@everyone @oak python <@&1349916362595635286> Why not announce that the lmarena.ai website is unavailable due to the massive cloudflare recaptchas?
why did you try to ping everyone 😭
and the mods
there are so many mods now, it's kind of rude to do that
it's down for you? appears to be working for me 😕
now i am worried about take a missed shot from this one, because it already happened to me twice
too recaptchas.
recaptchas are fine
I recovered the access to the website. 🙂
fortunately they don't go through 
glad to hear it! keep me updated if things stop working.
Hello
working
@echo aurora , why flux kontext max is not in direct chat anymore.
Sorry to say it had to be removed, not sure if I can share details as to why
Hey
hello im new here
Has anyone ever face to this problem? How can I fix it?
Hello
Anyone?
also anyone tried the new sonic model in roo code?
is there a dev specific discord anyone knows about for ai stuff?
copy your last text and refresh your page and paste your last text and if will work
it happened to me too
not now, long ago
It doesn't work
reopen the app Firefox
I did. I closed the browser and reopened it but nothing changed
Uhhh you should not put your key in lmarena. You know that everything you say will get public right?
Get public like how?
They will release all conversations sometimes
Ask the moderator
He will explain better
for your own safety, never post api keys in public, and delete any api keys that have been compromised (like the one you screenshot here)
i've deleted that message for you
Thank you very much
this is sometimes true (if you use assistants/responses api with storage enabled)
the larger problem is that other people will use it to run their inference using your money
weird fact, imagen 3 is more likely to pass the cyclops test than imagen 4 ultra
Heard this exact same problem somewhere. Any reason behind it?
Which do you prefer? Head to how-to for a step-by-step guide on how to generate your own video using the bot.
Sorry to interrupt you. I made a slight mistake. I'm talking about this folder .lmstudio in users
the clothes are so much better tho
look at that insane detail
yeah I agree but I tend to prioritize how close it gets to the prompt over how it looks when I vote
How is mistral medium 2nd without style control
anything new to get more banana models?
Oh and on default leaderboard GPT5 is behind Gemini 2.5 pro lel
And that's high. The version we actually get in web interface is worse
uhm its just lmarena. who cares
Nope, as usual, only editing image prompts increase the odds slightly.
Wut
nope, even plus users don't get gpt-5-high
*pro
oh wow
hold on let me pull up the juice diagram
atleast we can through lmarena 😄 but with absurd wait times
nice
i have been testing chess puzzles, and at simpler ones gpt5 high was doing them well, then i tried middle diffculty and after thinking for a long time it just errored, did this multiple times and always the same
i feel like after it thinks for a certain amount it just errors on lmarena
could be the timeout isn't set up right
would be really nice to know if it can do middle difficulty puzzles and explain it in words too
because that would be a first
it did really well on an easier puzzle, unique response
usually if it is not a puzzle with a check start, the llms cant do them rly
when is biomedical arena coming
its there already
the leaderboard
ah
sorry, i wasn't clear
The researchers have to post many prompts first I guess
Its alright
so no one has said when the leaderboard is coming?
There is a leaderboard on biomedarena that is coming soon
so we will have to wait
So far over 2 days Iv only gotten audio on a video once.. any tips?
hey
Videos don’t generate sounds right??
Award for dumbest poll goes to...
they started focusing on human preference I suppose. And they have always been great at fine-tuning. The main question is where is the new mistral large?
Overdue by some months
I just can not take them seriously again until they do at least that lol. They are simply too far behind
Is this GPT-5 Chat or GPT-5 High? Copilot
bing chat sydney from mcdonalds
Auto router model. Depends, give it specific instructions and it will switch to gpt-5 med/high, or else it will be in gpt 5 chat
Great, I'll tell him/her to think a lot, a lot, a lot, a lot so I can connect to GPT-5 Pro 😆
😆 they are cheapskates, mostly you will connect to gpt 5 medium
what is clippy?
what are some models you guys like to run locally
there are three new models in the arena: clippy, vista, and millenium (i think they are text models and both from microsoft?)
theyre like impossible to get
ive tried like 100 times
and have not gotten them in battle mode
neither is millenium
millenium says it's by Microsoft (MAI), also not that great
Shoot a high-angle aerial drone shot (150ft altitude) of a humpback whale breaching against a vibrant sunset sky. Frame the whale centrally, capturing its full body mid-leap with water droplets frozen in motion. Use golden-hour lighting to highlight the whale’s wet skin and the pink/orange hues of the sky. Ensure 4K resolution (3840x2160) with 60fps for smooth motion. Focus on the drama of the breach—emphasize the whale’s power and the vastness of the ocean.
is gemini 2.5 pro grounding exp available for direct chat?
so which one is good
there is yosemite now as well
none
cant you see my message
they all suck
its not that deep bro
idk, i've only tried two of them
😕
there is four and i have only seen two
theres like 0% chance u get clippy
0% bro
its actually impossible
can someone send me link to dev mode server i need to get back in there
23tN5ZTq8N
please dont scare us like that anymore bro
we need u there
ok, sorry
huh
fck u
#- ..to create a game like this:
-# ..to create a game like this:
(for above game, i used Claude Opus 4.1 Thinking)
is what he meant to say
sorry for the confusion he is causing everyone
lm arena is not the space for this, I'm really not sure what you think this has to do with AI
-# i can read that (small) text perfectly
He has "AI" in his profile bio !
-# He has "AI" in his profile bio !
Is it free ?
lm arena you mean?
Yes
Yeah
its free, just please vote
Lots of shipping this week (with more coming soon 🍌). Here’s the recap:
— We’re making Veo 3 available to everyone in a free, limited-time trial this weekend in the @GeminiApp.
— AI credits are doubling for Google AI Ultra subscribers.
— Over 100M videos have been generated
the banana is coming
Gemini account also tweeted it
https://fxtwitter.com/geminiapp/status/1959035394483503581?s=46
3 free video generations.
︀︀1 weekend only.
︀︀gemini.google.com/veo
︀︀
︀︀Ends Sunday 10pm PT.
is flux kontext pro down or something? regardless of prompt it shows black square for around an hour now (battle mode)
Ah, I see. My apologies.
No, it's been like that.
💀
We can’t modify an image that we generate in lmarena in this same app ?
lol
hola
lol
Hi) Can u please tell more about claude 4.1? Is it better then sonnet 4 for roleplays?
test it out
I just wanted to hear ur opinions about it🥲
fr
Sydney is best for roleplays
No i'm not... i guess...?
REAL
-# im a weirdo and i love writing in this small text
Really? I have never heard about it
@crack
I use claude for my ai dad, because real is still looking for milk
I tested a lot of models, and sonnet 4 is the best in my opiniom
Hmmm, thanks a lot!!!❤️❤️❤️ I definetly will test it rn, but what about claude 4.1?
Why did u deleted it?
I don't get it....
A dramatic scene in the backyard at night, a raccoon aggressively grabbing a small Yorkshire Terrier dog, intense struggle, cinematic lighting, realistic photography, highly detailed fur textures, full body shot, motion blur to emphasize action, tense atmosphere, dramatic shadows
true
sadest sentece ive heard in my life
damn bro sorry
it be like that g
I have, it's pretty good. Seems at least as good as Qwen 3 Coder tbh. Been switching between Roocode using Sonic and Qwen Code CLI.
Mama Cat Finds Hurt Squirrel by the River 🐾🐱🐿️ Brings It Home for Healing Together 🌿❤️✨
https://youtu.be/ooBEZ3bChMo
💖 This is a heartwarming and wholesome story about love, sacrifice, and the simple joys of family. The bond between Mama Cat and Ginger Kitten will melt your heart and bring you comfort.
✨ If you enjoy c...
Nah, it's ok, tnx❤️
What is sydney and where i can test it, btw?
I didn't see sydney on arena
Why??
Same issue here with beta.lmarena.ai (no models popping up)
GPT-5-chat is the non-reasoning version right?
Gpt-5-chat is worse than Gpt 3o
@echo aurora what's the difference between beta lmarena and lmarena.ai?
There is no such thing as "GPT-3o"
Do you mean o3 or 4o?
yesn't
The first gpt model
I'm talking about the 2023 one
how
Why is this happening?
What’s beta arena?
Hmm let me check
It’s working for me. Are you getting this?
yea
Oh right sorry I was thinking of something else, it’s just a beta version
Oh wait I’m getting connection error now too
Kk going to escalate
Thanks
Wait nvm it’s back
Anything special in it that lmarena.ai doesn't have atm
Same thing on a different browser? Is battle working for you?
I'm just checking now
just trying out what it is
beta arena no more?
heyy, anyone here who uses higgsfield for product photography. i would love to have a chat with you:)
lmarena still giving me errors...
Higgs boson
Hyy @all
Not able to use beta.lmarena.ai ....it's saying now model please let me know what I need to do to use it again
Hello, just joined. Never used discord nor lmarena but here I am learning. It looks so cool
Well, welcome to Discord and LM Arena. Nice to meet you.
Thank u !! some models are sooo cool. I have a question; but after u vote will they tell u which model did they use?
First of all, my pleasure. Second of all, yes, that's usually how it works.
For how long have u been using lmarena ?
In Battle mode, for instance, you can type a prompt to give to two anonymous AIs who will then answer your prompt. You are then tasked with voting for which one provided a better answer to you. After you vote, the names are revealed.
I've been using it for quite a while now, actually.
I realized that they have some of the best models on there, and that you can use them completely for free. So I decided to start using LM Arena.
Okay and tell me which models are I fav ? For:
- image generation
-image editing
-video
-image to video
?
Like which ones do u like the most ?
For:
Image Generation: GPT-Image-1, nano-banana, Imagen-4-Ultra
Image Editing: nano-banana (it's the best image editor by far, being able to accurately follow your prompt (sometimes) with no issues whatsoever)
Video: Veo-3-Audio
Image to Video: (Same as Video)
Wow never heard of nano banana ; is it a new model?
Sure is.
It's confirmed to be a new model by Google that is rumored to be released soon.
How do u pick this nano banana ? Or is it just random?
It can change things about an image while keeping the same art style and the character consistent.
Unfortunately, as of right now, the only way you can use it is through Battle mode, and you have to get it by random chance.
Oh I understand like the shiny chance in pokemon
Yeah, exactly.
You said veo - 3 audio
Veo is from google right? But audio because they also add voices?
There were a few other mystery models that I saw while trying to get nano-banana.
That's correct. Veo-3 is from Google, and the reason why I put "Audio" at the end is because it can support audio generation to create voices for people as well as sounds.
No wayyyy jeez that’s insane
Yeah, they're truly an advanced company when it comes to AI.
I wonder what will happen when we will not be able to spot fake ai from reality anymore. I sometimes fall too
For that
Only at that point will society truly be screwed.
still won't let me accept terms of use....
Well, just like any of the mods would say: just report that to the #1343291835845578853 channel.
Either that, or try looking there to see if that issue has already been solved.
You could even provide a screenshot to further help them out.
Do u like chatpgt image tool? I honestly don’t at editing is so bad and about creating images, they will just have a yellow Mexican Netflix filter
Well, it is good for generating long strings of text, though.
So is nano-banana too, I think...
I read on Reddit a question , a dude was saying
“How long until we will be able to generate the last season of got in a way it won’t suck?”
What do u think?
If I'm being honest, I don't know what that show is.
Game of Thrones man, come on it’s crazy famous
Ah.
I know what Game of Thrones is, but your message was typed out in a way that I couldn't tell what you were talking about.
For instance, you could use "GoT" or "GOT".
za wardo
To answer your question, since AI is rapidly evolving just like we've seen with nano-banana and GPT-5, I reckon it won't necessarily be too long until we can make full-fledged seasons of shows or even movies solely with a video generation model.
All the big companies have to do now is just solve the little, minor, inconsistent flaws within image generation models, like:
- The image generator not being able to render very small text
- Sometimes having some distortion on small, minor parts like pupils and eyeballs
And then we're basically screwed.
Oh Lord now I’m just figuring it out that companies and studios will try to earn more and pay less; so they will all eliminate jobs and fire thousands
Unfortunately, at this point in time, companies will do anything just to make a bit of money.
Even if it means laying off a bunch of employees.
Laying off employees often means making money as that’s reduced labor costs
Well, that's my point.
the end goal of capitalism is to maximize value produced by as few laborers as possible for as low wages as possible
not even a conscious choice, just mathematically baked into the incentive structure

Generating images. We'll notify you when we're done.
