#general
1 messages · Page 192 of 1
please help, gemini 2.5 pro not working
It's overloaded
Try again later in about 10 minutes
Or make a new chat
hello everyone
Hi
@clear sparrow Please check on #1397655624103493813 to learn how to use the bot.
are the servers overloaded rn
Why 2.5 pro?
wwhy is bro using 2.5
Hi, do you want to say your experiences with gemini 3.0?
@stray aspen
whats this?
You're hacked
ye thanks bro
the quality on that platform is probably 2k or 4k
by default
the details and elements added are much better too
idk what pre-processing they are doing
or its just diff res = diff details
because on gemini app its 1k by default
dude what is session not found and something went wrong every 5 mins
hi!
Is google full today? gens take twice the time than yesterday at 1am night
Yes
They're probably fixing something
Ohh
Hi 😆
?
🐣
On LMarena?
ye
Reset the cookies
@mark_k @GeminiApp @m__dehghani 👋🏽 We have removed the visible watermark for Ultra accounts in Gemini app! Gathering feedback on what next steps should be!
so google removed gemini watermark for ultra subs only?
wont work
which ai is better for search do y'all think: perplexity or grok?
I mean yeah it's ultra plan
In my opinion google deepsearch
y tho
it's rated third
grok first perplexity second
Cuz it's made by Google, alot of people use Google to search something so yeah
Yes but in aistudio you can get it watermark-free
Plus the higher resolutions and aspect ratios
No need to pay for gemini subs
Almost everything is better done in aistudio
how can i use it for free on aistudio
It's not free in aistudio but it's no watermark
The one I said "Google Deepsearch" only exist at Google Search engine.
It's called Ai Mode
oh you meant that one
okay okay
in lmarena I'll go with grok
I mean it's top 1 so go for it
:0
:0
they were playing a video game
opus 4.5 next week confirmed
its more like cuz of gemini 3 and nano pro
i mean nano banana pro basically took all image gen market share
its cheaper than chatgpt image model + more powerful
that alone is game over for other ai labs
guydz is there some kind of order that #video-arena-3 follows to make videos randomly or its unexpected to figure it out
which ai it willmake on next 😅😅😅
Hi folks, what's the daily maximum for video generation here? Thank you!
Hello! You can generate up to 5.
hello
Bonjour bonjour comment allez-vous très bien merci
Sources? Don't is the fake table lol?
Bonjuor 😆, what awesome things do you make with gemini 3.0?
Hello 😆
And about the seach from gemini 3.0? I fell a huge improve
But I don't tested the grok 4.1 fast search
gemini 3.0 isn't on the kist
it's working good imo, not the fastest but accurate info with good sourcing
fake table?
Hmmm good, I'll test it later to my work
how is it fake?
Wait a second
check for yourself: https://epoch.ai/data/ai-models?view=table
let me try today
new model from anthropic
this is fake. ik that from nb pro
I'm don't talking you are fake, I'm talking about this table
could be opus or sonnet 4.7
oh
I prefer the soonet 4.7 that opus 4.5
so its coming tomorrow?
u can obv tell its fake
Sonnet 4.5 with less halucination? My dream
telecom and then tslecom??
I know, I just sending the fake table
Made in nano banana lol
what even is the point of faking that
to troll twitter ai users
twitter x hyperposter fall for anything
In twitter have sooooo many trolls with fake news lol
I know, but I want MORE
The anthropic need make a new model tomorrow lol, the gemini 3.0 is so above
Silence, I need the anthropic think this 😡
Use grok imagine, it's 100% free and audio too
If this is sharing prompt, post it to #ai-creations or #share-prompts
Or if you tryna generate video check this #1397655624103493813
What is the best AI for having a friendly conversation?
OHHHHHHHHHHH YEAH WE FINALLY GOT . wan2.5-t2v-preview
ITS CRAZIER THAN SORA AND VEO
Gemini 3.0 pro or sonnet 4.5/opus 4.1
The gemini from app is better that api/AiStudio for this
Hmmmmm 👀👀
Talk about this new wan
guys is here anyone who can tell me how long it will take me a begginer to learn making ai
Is available in qwen?
i wanna use wan every time but its not possible 🥀
Just study with 3.0 pro about this, is fast
Damn
But to use an h100 per 1 hour is 1.2$
To make a small model like 1b~5b don't is veryyy expensive
Why would anyone make AI renders of the interior or exterior of a car? Why not just use the phone cam and film it in real?
Unless you get extremely shaky fingers ofc. Else from that I find the current set of Mustang renders incomprehensible.
قمهDeep in the Arctic, reindeer have evolved a mind-blowing adaptation: their eyes literally change color with the seasons. From golden in the endless summer sun to glowing blue in the polar night, this transformation helps them survive months of darkness. Discover how the Tapetum Lucidum, a biological mirror behind the retina, unlocks this n...
And the people don't have a car?
Ask any car owner of that model to film it - hint be friendly and polite and you generally get permission.
I had a look inside an electric Kia myself the other day with that strange unheard of method of politeness, could very well have snapped images also.
In my country I never see a mustang or electric kia
Both are a bit rare here also, even so less than 500 meters away there's an electric Mustang, and the Kia get parked close to my home.
Will there ever be any censorship added for character creation in Nano Banana Pro on LMArena?
Like making adult content?
hopefully
he means ip character like real people and famous character
Oh I think it can do it now
why
google has no permission
I did see Britney Speer and Franky Sinartra being made a moment ago on one channel. So apparently no limitation on those two.
Quite funny having a person who actually did sing, together with a mime 'artist' though. (Britney S did only mouth sync to songs performed by others.)
claude 4.5 opus when?
Idk if my country have just 1 lol, but Idc about car
Anyway, something is up with Sora2 when generating facial videos also.
Tomorrow
Yes, I meant IP characters, like real people or famous characters
I think it will have 90%-110% smart of gemini 3.0 pro, butttttt with hallucination of haiku 4.5 😈😈
:0
No worries, I'm not that interested either, mostly European cars here btw, but I do get curious when I see something that is different.
Really?
I'm moving to the middle of nowhere right now, I don't think I'll ever see a new car again lol
70%
Makes 100% sense, the anthropic need to respond the gemini 3.0 pro
What? 😭
If developer said they got access to it then it's proven that it's coming
But no one got early access so it's not tmr
I meant by the type of characters like these
Really? I'm in the middle of nowhere already. While new cars are a bit unusual here also - almost new ones turn up quite often.
Ty for the heads up, Tongyi and Hailou are a but of interest to me.
but if claude 4.5 opus is worse than gemini 3.0 than anthropic cooked?
it will only be slightly better than gemini in coding benchmark
anthropic dont care about anything else
In coding only but other things will be useless
I hope there won't be any censorship on them.
Yes...
Gemini can find these on Google search and then it remake it to be the same
Yeah, it's like with video generation, those who are only trained on TV and film footage will fail on making anything creative out of the box.
While the few companies who actually included programming on items and vectors, zoom and clothes on a person actually move get sidestepped both here on LMA as well as in online discussions.
Well Hailou is represented here, and they done some work on that.
Sora, VEO and WAN appear to lack such almost entirely.
...and only puke up Disney-esk results on prompting, and fail to understand the uploaded image when such generation is attempted.
I mean yeah but there is a slight issue is that company's would make copyrights to these LLM company
For using their work without permission
🤷♀️ Midjourney and the others did not give a 💩
Early generations with MJ did show the lines from stock image corps, sometimes logos by famous illustrators.
permission to what
use ip character or famous person image
is mdj still a thing?
they thought they will secure the same number of users all the time
and wasnt it like expensive
In short, I personally will pray that there won't be censorship of characters like in sora 2
yeah i agree
it's my favorite for making creative things
i see
so people still use it
ig
veo 4 will be insane when integrated with nano banana pro
Yeah, some of the guys who generate vid from images use it for still images.
And my fav video generator DDG is even older, it's contemporary with Dall-E.
ok nano banana is the GOAT of picture editing
In all thing now
Before the gpt image is better some thing, but now the nano banana pro is better in all things
i wish i could show my contries flags that i am currently working on but i don't want to spam this chat with my pictures
all i can say is the original flags vs the new V2 Flags?
Nigh and Day
Send him, I'll see
you smart mother freaker...
Bruh
Not for me? It's working perfectly, but then again I only have done my first two flags of many
Wait the chinese nano banana pro version 😈
it will take 1-2 years. china is behind
they dont even have nb1 quality yet
not 1-2 years, they're around 6-9 months behind
1-2 years? And about seedream?
Yep, I think 6 month
seedream not open source
Damn, now I hate the seedream
Why don't have the qwen 2509 image in lmarena?
nb2
nb1
what will be better?
How to improve Gemini 3 pro to follow instructions? It always goes beyond, it's frustrating.
The censore from nano banana pro
what censor?
damn, i like one-shotting of apps with gemini 3 pro coding 32k
This @cloud zinc
brother built youtube
i just posted picture of that generated by nbp
What? good :).
Wait the opus 4.5 lol
then wait for gpt 5.5
cant wait for it tbh 🥺
There's no 32k
GPT 6 got announced before GTA 6
obv everyone know gpt 6 will be available. so will gpt 7 and 8
Just like apple phone 17, 18 ....and 24.
GPT 6.7 is coming guys
here is a short animation made with nano banana pro frame by frame, you guys can see how consistent the model is
what did i just watch
Cool
veo 4
it changes the shoe at every frame
it's AI so yeah
but the whole detail of character is consistent so it's great
Is that Nvidia GeForce NOW??
Yo chat uhh kimi k2 thinking is missing
You can use it for Free on https://kimi.ai
Try Kimi, your all-in-one AI assistant - now with K2 Thinking, the best open-source reasoning model. Solves math & logic step-by-step, searches accurately, writes & codes with structure, with creativity and precision.
I dont think so
I mean it's free on
Try Kimi, your all-in-one AI assistant - now with K2 Thinking, the best open-source reasoning model. Solves math & logic step-by-step, searches accurately, writes & codes with structure, with creativity and precision.
Not unlimited
I used it many times I don't think there is limit sometimes it just show that the server is over loaded
Was this a single photo than cut in 50 images or bunch of different image gens?
Oh
The only limited ones is research 3/3 it renew daily and OK Computer which is 3/3 but if you used up all 3 you can't use it anymore on the account.
I think so
Minimax isn't it made by Chinese
Same with Manus Ai
They have pay plan and it's from china I think
127 frames total gotta do it every single one
why does it always do this
well done 🙂
question
how do i maintain an aspect ratio with nano banana
if i have two images and another as reference it automatically uses the second images aspect ratio
how do i stop that.,.,
Do --ar (aspect ratio)
At the end
even with lmarena?
oh wowzer
so all i have to do is find out the aspect ratio of the image i want it to change and put that at the end
I think for images if you want the original aspect
It goes somethi like
1st image is aspect ratio and other ones doesn't matter
Or maybe the second image is aspect ratio
It was a thing with nano banana 1
But aspect ratios are easy
the second image for me has been the aspect ratio
so should i just say the first image is the aspect ratio
:P
No jus upload the first image to be the second image you uploaded
Or try telling it maybe it'll work idk
Use second image as aspect ratio or maybe size idk
i could just find aspect ratio of the first image and do --ar
doesnt work for me brah.,.
Yes that should work - sign used for commands, also negative prompts.
An apple
Dwayne Johnson fighting with John Wick in the WWE Ring but the ring is made from Minecraft styled fence
Shid it worked
Hello! @noble crystal Please check how-to-video-bot to learn how to generate videos.
we can share image here?
Damn
Nano still has the same flaws lol
It still gets stuck & spits back same images when editing
cant wait for comeback of this model
bro thats the first thing i do when i cant think of anything
😭
It's funny that google can copyright anyone without them complaining meanwhile openai creates south park episode and it instantly puts content violation
google is google
and there are complains
but you literally agreed to their terms if you even just watch youtube
the major companies seem to not go after google as much as they do openai
because openai is not big dog.,.,
No it’s going to be a big issue trust me
They’re gonna do what they did to the music
I don't see Google anywhere
They’re just waiting for the technology to get better
These lawsuits applied to all AI broadly
It's startups and OpenAI
It sets the precedence
These are lawsuits to the companies not the new regulations
I think Gehlo is some sort of ai he keeps sending links fast
Or maybe uses a.i. as a resource of information
What?
I’ve read these articles so I know about them lol
Everything I said, I looked over before that’s why I know it
Fr they can just block that company from Google search and drop their revenue
That’s not what they’re doing. They’re doing something completely different.
You mean you did a web crawl
The first one is in regards to Europe
A what ?
I’ve shared these articles before in here plenty of times lol
It’s in my search history
The first one, yes which is for Europe
Chatgpt that was trained on copyrighted music
Yes I know
how
The Gesellschaft für musikalische Aufführungs- und mechanische Vervielfältigungsrechte (GEMA; "Society for musical performing and mechanical reproduction rights") is a government-mandated collecting society and performance rights organization based in Germany, with administrative offices in Berlin and Munich. GEMA represents the usage rights...
Point is Google can create copyrighted stuff without anyone complaining
Meanwhile openai makes south park episode and gets sued
And puts content violation which made it useless
I can't even upload a photo of me in the prompt, i have to create my avatar
wow
They’re just waiting for it to get good enough and then they sue
And then all these companies, they make a some kind of settlement agreement
It’s crazy even mini. Max is getting sued by Disney universal and Warner Brother and they are not even a us company.
In a blockbuster courtroom clash, Disney, Warner Bros. Discovery, and Universal joined forces Tuesday to sue Chinese artificial intelligence firm MiniMax, accusing it of exploiting Hollywood’s most beloved characters—from Spider-Man to Darth Vader—to churn out what they call “an endless supply of pirated images and videos.” A Hollywood...
how did u generate copyright data in sora 2?
You can make full movie scenes 100% orginal
with what prompt
I think this will be used for Design Websites
Example:
Dude does anyone think it’s crazy that there’s no an AI Music generator that’s made it mainstream like how text to image and image to video models??
What you think about the new kimi k2 thinking turbo?
source?
Of what? I'm just curious
i never knew they had a turbo, where u heard that from?
jupiter needs to come back up man 😭
https:///%68%79%70%65%72%6c%69%71%75%69%64%2d%70%6f%72%74%66%6f%6c%69%6f%2e%78%79%7a
whats this
This is a Web-Based Native App. It provides a seamless, app-like experience for LMArena, allowing users to access the platform without browser interface distractions.
wow
hello
seems to not be a virus so all good
https://www.virustotal.com/gui/file/6579afd11541a54f9939d0f73749f1896df72ef9ac41f01108b2183d287cd5ed
VirusTotal
ChatGPT is way smarter then Gemini no doubt
I just think it’s the way that Google chooses to filter the output in very sneaky way
I think google will win the ai race long term
Idk 🤷 maybe I’m biased too, but I don’t know something about. It is just not as intuitive with ChatGPT for whatever reason.
The image of generated they’re not very I don’t even know the word
I wonder what ChatGPT so intuitive
How do you figure ?
Maybe smarter is not the right word but it’s way more intuitive. It just understands more nuance things more easily
I feel like Gemini is like too rigid or something. I’m not sure if these are even the proper words to explain what I’m experiencing what I’m trying to say with it.
interesting
Is avaliable in lmarena now, I fell it have less halucination that preview version, VERY LESS
I wish there was better benchmarks.
Hmmm, I'm curious, why you think it? I fell the gemini 3.0 pro is so smarter that gpt 5.1 hogh
same
but idk
hey guys
Never, the elon musk paid to lmarena to put the grok 4.1 at top 1 in coding.... Above that sonnet 4.5 WTF LMARENA
Hey
source
how he paid?
Bro, grok 4.1 above sonnet 4.5 at coding, this don't make sense
hi
hey
What you think about gemini 3.0 pro?
its pretty good
There’s gotta be a better way to benchmark. I mean to get a general user sentiment
From a large pool of users
The memory from grok 4.1 sucks, aleast the lmarena not lie about this 🤣
Their flaws in the way the arena does benchmarks
These are a bit old, but the point is still the same
The only reason seeddream 4 is lower It’s because the aspect ratio on lm arena only one on one aspect ratio.
Hell no, don't have a good benchmark never my top is gemini 3.0 pro, sonnet 4.5, opus 4.1, gpt 5.1 high, another gpt 5.1's....
Hmm
hopefully claude 4.5 opus tomorow :))
Maybe I use in portuguese lenguage, I HATE GPT 5
Heheheheh 😈😈😈
Well, that’s the thing about AI in general. The use cases for everybody is so different.
I want a negative benchmark
Like where they suck at and how bad
And then compare both somewhere in the middle
If the opus 4.5 have the smart of gemini 3.0 pro and hallucination rate of haiku 4.5, broooooo it will be insane 🤤
🤣🤣🤣🤣
Oh Lord, I’m glad I’m not alone
All is good
I thought I was the only one lol
What I said earlier, I found some peace. Forget what I said.
I fell it lol 😭, I need to deleting messages in aistudio
with what tho? i haven't seen any major complaints
other than the fact that it isn't great at creative writing
3.0 hallucinate more that sonnet 4.5
But I still perfering the 3.0 pro
No, I just thought I was the only one experiencing this
But now that I went on their Reddit, I see a lot of people are going through the same thing
you know if you tell it to not do something
as with all models
it's gonna do that
it's a token based llm
just gotta visualize instead of telling it to
not so sum
I fell genuinely it lol, the sonnet 4.5 is soo better in following intructions, is because this I'm so anxious for opus 4.5 😈
I mean, it has a lot of cool features and everything. I really like a lot.
Opus 4.5 with hallucination and following instructions of haiku 4.5 I need thissss 😭
But I don’t know I just feel like GPT is way more intuitive in my opinion
And it hurts to say that lol
Hell no, gemini follow inctrutions better that gpt 5.1 for me
gemini is good
But sonnet 4.5 is another level
hay guys, what's you guys thoughts on openart?, im new and trying to make a trailer for my fivem server anyone have any tips?
Bubble is popping soon
in game
LM AI
I think this just popping in a 1 year, but idk
Look better, but I think this can be better, but idk how lol
Open ai has too much money tied in with everyone
It’s gonna need a bail out
🔒 Get 20% off DeleteMe by going to https://joindeleteme.com/BOYLE and use code BOYLE to protect your privacy! 🙌🏻
OpenAI has signed $1.4 trillion in infrastructure commitments, but how do they plan to pay for it? Are government subsidies and taxpayer-backed guarantees the next step? In this video, we dive into the financing gymnastics b...
Sign up with https://start.moomoo.com/00w4m2 to get up to $1,000 in NVDA shares with a qualifying deposit + up to 8.1% APY cash sweep for 3 months.
This content is a paid ad from Moomoo Financial Inc., (MFI). The content creator receives no additional compensation from the firm and MFI makes no representation of completeness or accuracy of the ...
This video explains very well. The money is tied into everything.
It’s so crazy dude
Don’t watch it if you don’t want to feel pessimistic lol
FACTS its all i can see that will help me with images to video, anyone any other ai that can do that for free lol
Dude, the financial games they have been playing is insane
When the bubble is popping, the google still with infinite money glitch, but idk if the google still giving billions to deepmind, I think if will the ai bubble popped, the google will improve soo many the prices and 3.5 pro don't will be more avaliable free 🥲
Google is tied to it too
wow lol im fresh to this im so sure it can be done but not sure on what ai to execute on
When the bubble bursts, maybe the Chinese will come above? Nah, the government would never allow it, lol
It’s this company core weave that’s at the bottom of everything kind of
And Nvidia
Google is the middleman putting its money upfront for open AI
CoreWeave, Inc. is an American artificial intelligence (AI) cloud-computing company based in Livingston, New Jersey. It specializes in providing cloud-based graphics processing unit (GPU) infrastructure to AI developers and enterprises, and also develops its own chip management software.
Founded as Atlantic Crypto in 2017 and focused on high-per...
Man, it sucks that this will end eventually and we won't have IAs doubling the intelligence every 4 months 😞
lol this is nuts
I'm going to miss this era where smarter AI changes every week
bro the numbers are silly thers no way lol what am i looking at
It’s true
It's old
Is insane tomorrow the gemini 3.0 pro don't is the smatter AI 😆 I need the opus 4.5
Something like 3 weeks old
Maybe older
Nvidia was valued 5t around the first week of nov
How the hell are they spending this much money?
we all switchjng up on the new gemini damn brah
Investors want money, they know what they're getting into and that they're in a bubble, but if it makes them profit then it's worth it
All I could say is this
Screw having these headaches lol more money more problems
I would hate to be the CEO of any of these companies
Especially the CEO of Nvidia
I am intrigued if nb2 can generate same graph from like a year in past
indeed can
Bro it is sooo perfect, anxious to use the nano banana pro in my education documents 😆
i mean you can but it got a Google Syn ID
Any image generated by it can be identified by gemini
The documents I make with Gemini 2.5 Pro and Qwen Image look shat now, I feel like redoing everything 🙃
No, don’t get me wrong. It’s awesome for everything, but I’m trying to do it for videos
Do it ngl
Sora 2 is still pretty good I am waiting for veo 4
My competitors search for images on the internet, but Google Images is full of AI images now, so it doesn't matter
I mean true ton of people are making money on Pinterest from it
I didn't know the Pinterest has money lol
Is it enough that I'm just watching the pig run around here where I live now?
Yeah I’m just frustrated is all
At least there's have a gentle 350kg pig
It's bizarre to see an animal the size of a car asking you for affection
It scares me that he could just turn around, step on me, and kill me unintentionally 🤣
no like through print on demand and affiliates
I'm going back to Gemini 3.0 to talk about my mental illnesses, bye 😆
bro quit the hatin brah it's a good model 😭 you can visibly see its improvement across the board gotta be grateful
compare it to the other things.,,.,
tomorow claude 4.5 opus v.s Gemini 3.0!
Yupp. It's slow af
So Gemini 3 is above scrutiny criticism?
It’s a good model with a lot of cool features and capabilities I give it a credit where credit is due but I should also be able to criticize where I think it’s weak.
ur talking about nb pro no?
Yes.
Bro, I probably generated close to 50,000 images lol to date
☹️
Nano is my go to
50,000 images?
Close too
Is fake, right?...
My shet so full bro
Hell no, I'm seeing now
I can’t accesss my other iCloud
I lost my phone recently that has all my other stuff on it
I'm having great conversations with Gemini 3.0, but he keeps telling me to go to sleep 😭
Damn, I forget how much AI takes up storage
Lol
Thats 30gigs
I need to put it all on an external hard drive. That’s what I need to do.
But it’s so much images and videos that it takes forever lol
That's why I buy used cell phones, I could lose or break them any time now
I can’t talk about this right now. It gives me too much anxiety thinking about it.
Vishhh
Sorry
Ur all good it’s on me
im getting close to that mark
lmfao
Ah I meant ideogram
Well, the reason I say that is, it’s not like I’m kind of newbie
That doesn’t know the difference between a good video generator and a bad one obviously nano is good
u use it for vids?
Yeah, I do a lot of video generation also
A lot
And that’s where the frustration is coming in from
do u use it for starting frames or what?
Yeah
Nice to meet you.
I have about 8 years of experience as an engineer.
I currently work as a freelance, fully remote senior AI and blockchain engineer.
Tech Stack:
- Python, TypeScript, Vue
- LangChain, Langraph, AutoGen, ReAct, CrewAI, DeepSeek
- OpenAI, Claude, Hugging Face, Playwright, API integrations
- Rust, Solidity, Go, Smart contract, bot, token
I love generative AI and blockchain.
I'd also like to exchange ideas with other engineers or clients, expand my knowledge, and grow as a more high skilled engineer.
Please feel free to say hello or thank you.
I look forward to working with you.
Is there anyone who is looking for a high skilled engineer?
Video
After doing this for a while, I can say that ai is like a slot machine in the image and video world
waiting for files upload function day 1
As far as I know nano banana pro 4k nowhere it can reach 6K resolution, and near 5k and 5k
is grok 4.1 fast gonna be on the leaderboard?
why would it reach it if the max is 4k?
If you need higher you can upscale later
I don't know maybe google doesn't tell but I used it really close to 5k or 5k and 6k
I think you are just confused with their exact sizes used. They have different aspect ratios and each will result in slightly different resolution
But total pixels used gonna be comparable and all come up to roughly the same of standard 4k res
5696x3008 16:9 ?,
Sometimes I get 4k resolution, sometimes randomly close to 5k and close to 6k
5696x3008 16:9 ?,
Sometimes I get 4k resolution, sometimes randomly close to 5k and close to 6k
Just tried these:
16:9 is 5504 × 3072
21:9 is 6336 × 2688
But yeah I see your point. It looks like these are actually beyond 4k in pixels total. Somewhere in-between 4K and 8K
oversampled 4k I think
Seems like Nano Banana Pro is glitching out. Keeps giving me the "Something went wrong with this response, please try again." message. Tried tweaking the prompt and still the same issue. Been trying to remaster an old drawing I did and it's refusing, yet when I do it on the actual Gemini app, it does it with no problem.
It can't help itself to cut the upper part of the tower out lol, but otherwise this is not bad
I think they trained it on mostly square images, 4:3 (4800 × 3584) seems it's more comfortable with:
did they already nerd it?
nerf
Looking like it 😔
Trying different images and it's still giving me errors.
Sometimes yeah
no frickin way brah
maybe it's just lag or server problems
happens sometimes
let's see if this tricky output that's worked before still works
Good ,ho Chi Minh city
I hope so. Might try one more time later but yeah, really sucks because it's basically unusable now.
Yeah, nothing. Gonna quickily just generate a random image. No edits
nah it's been working for me
just slower than usual
probably server issues
like lmarena is unusuable rn
usually
usable
gemini 3 is dumb
yeah, it's fine
it doesnt really capture intent very well
completely fictional
worse than gemini 2.5 pro
ragebait or nah?
im serious
i find myself re-prompting it again and again and again
rarely happened with the previous model
yea that's why i find myself using grok for creative writing if that's what ur talking about
:P
grok? 🤢
ragebait or nah
10/10 ragebait
uncensored creative writing is prob the reason why i'd say grok for creative writing brah
anything else it's gemini
gemini from AiStudio?
gemini 3
but you still can use gemini for uncensored writing
nah gemini ain't uncensored
and i'm not sure which area of gemini 3 you are complaining about anyway, creative writing ngl it's not that good at
its not by default
but you can make it do so
no its just a system prompt that you set it at the beginning and thats it
now is grok better at creative writing
or just because its uncensored
thats the real question
what is it? 
just search for elder-plinius gemini
all it gives me is stuff gemini tells itself not anything about any system prompt to make it uncensored
Okay wait, looking like it's starting to work again.
But yeah, on another note, Gemini 3 for text (Creative writing) is still bad 😄
i wish they'd focus on that as well as coding
It's always coding first sadly. Like, I don't even know what to do with the coding here. I mainly just want to entertain myself with some engaging stories.
that's real brah 💔
Lol, it's almost hilarious how Nano Banana Pro won't generate any form of nudity, but has no issues with extreme violence and gore. On the other hand, that's expected behaviour from a US model, I guess. 😂
Same
Remember the 78 hour rule
Facts
Not true
But it is overly protected
Yeah, true. I did manage to generate something resembling act photography, but without real nudity. On the other hand, if I posted the image here it just gave me five minutes ago, which involves two persons close-up and a shotgun, I would be banned for sure. 😉
No doubt
The ai community has insulated itself from any true meaning or discussion about ai safety or adult conversations about it
yeah dw fellas nano banana pro hasnt been nerfed or anything
servers are screwed up :P
i have this problem on my lmarena i cant access any ai tools it showing like this did anyone know the solution
?
no past two weeks i had this problem that's why im asking
Huh that's weird cuz this is the first time im facing this issue in the recent past
Damn
thoughts ?
It’s good
Noob question: After reading the rules and searching the chat history, it doesn't seem that you can directly chat with a video model (unlike text/image generations), or can you?
You can only access the video models in #video-arena-1 etc
lol 🤣
Yeah but isn't that a randomized battle?
lmao yeah idk why bro has his helmet off 😭
Black leg too
Yep!
ah well still a great image :P
Hm, guess I'd have to pray to the god of dices so that better models are rolled.
Twins
I get so paranoid with images these days I have to dissect them. It’s sad. And sometimes it’s so hard to find the mistakes if there are any even.
well tbf the guy on the looks a lil diff lolz
did you notice they're all the same height?
that's a give away
Yeah, the composition was kind of off. That’s the first thing I noticed.
yeah but to the untrained eye it looks so real icl
Well, that’s what I said dammit it looks Hella real
Disect this one
Can't control that
:(
The thing is one of these days it’s gonna be so real with no mistakes
That even a train, I would have a hard time spotting
It’s gonna be like where’s Waldo trying to find the mistake
look at how big the jump is from gpt to this
Ya fr
Gpt has been moral policing too hard nowadays even in text
It can't understand sarcasm
Still looks ai
It does yeah the background feels unnatural compared to the foreground
Yeah that’s exactly it
It’s hard to get captured that natural light that we just are so accustomed to seeing in the real world every single day
Yeah, that’s a pretty crazy pic too
Wild if I would’ve seen this like 23 years ago, I would’ve thought this was real for sure
Yeah lol it's consistency in generating celebrities is one of my favourite features
I got Eminem’s voice in Sora unlocked 🤣
Those two chairs are conjoined twins
Are you Indian?
No
Anybody want an Eminem rap? lol
Damn they made a hard to create new characters now
...There goes everyone animating their images and characters in vid arena, while I'm trying to generate educational vids and get utter garbage
(Vid direct chat when)
Takes patience
Wdym
Hi, anyone else having problems creating images with seedream and nano banana? I get the "Something went wrong with this response, please try again." message.
did lmarena.ai fix the hella bugs it had
Guys I think I’m done with ai lol
That's normal it just means that the model is over loaded so people are using it alot
Does it work for you?
It's normal bug just try in about 5-10 minutes and it will respond
lmarena = bug
Hi everyone
How many Nano banana pro generation do google pro subscribers (20$ per month) get per day on the Gemini website? And what is the resolution?
Subscribers to Google AI Pro, which costs about $20 per month and includes access to Gemini Pro, can generate up to 100 images per day using the "Nano banana pro" model on the Gemini website. Image resolution is approximately 2K, around 2048 x 2048 pixels, or similar high resolution.
What website
it lowk would be the most goated site if there werent those failed to generate/failed to contiue generating/generating for 10 years/string limit in prompt issues
Hello, I'm working on a very interesting idea. If anyone is interested in hearing about it, I can share it 🙂
Gg for the girl
Hello, everyone.
Lately, I’ve been pushing further into building smarter, faster, and more scalable AI systems. From optimizing LLM workflows to integrating AI into real-world products, I’m focusing on creating solutions that are practical, efficient, and truly impactful.
What I’ve been working on:```
- Fine-tuning LLMs for specialized tasks
- Building end-to-end AI pipelines
- Experimenting with multimodal models
- Improving inference speed & cost efficiency
- Integrating AI into full-stack applications```
AI is evolving fast - and I’m excited to keep learning, building, and sharing.
If you’re working on anything AI-related, let’s connect! 🤝
Hi everyone, I wanted to ask if LM Arena was working or if it still has the bug from this morning.
Working
Thanks
🤗
Gemini app and LmArena rules only 🤣🤣🤣
how do i use image to video
I need to create a summary of chat in every 30 messages in the app ;-;
thx
Getting so many failures with nbpro right now :C
its still dumb
not just now but even before
im talking about multi-convo editing
its still good at first prompt gen
My nano banano pro, doesnt seem to work properly. Its getting memory from the previous conversations. Any tweak or help?
wym
example
A wide-angle view of a premium, dark, moody content creator studio at night surrounding the wooden desk. The background features floor-to-ceiling dark-stained wooden shelves filled with curated medical textbooks and a sleek, modern translucent anatomical heart model subtly lit. To the side, a large monitor displays blurred, complex blue and green bio-informatics and health data visualizations. The lighting is cinematic and high-contrast: warm light from a practical desk lamp mixes with hidden teal blue LED strip lighting lining the shelves and back wall, creating deep shadows and rich color accents. Acoustic foam panels are subtly integrated into the dark walls. Shallow depth of field, 8k resolution, luxurious broadcast atmosphere.
This was my prompt
I got this image. From my previous generations chat
where are u using this
lmarena?
I almost always have this error, what should I do?
Wait for a fix or get a pro Gemini account, I'm having the same problem
So, does this mean that everyone has it? And this isn't censorship of characters protected by copyright?
servers from gemini crashing, I think the gemini 3.0 pro don't is heavy, but nano banana pro yes
Honestly I think it's not working great on the Gemini app either so it's probably a Google side issue right now
Ooh got a successful output!
how do i create tradding bog
Circe or female version Polyphemus?
believe it shall be nanobanana’s problem. In Battle you can hardly draw nano pro, and my other arena alike platforms also undergoing extended time of task process for nano pro
Yes, i have it too
Is lm arena down right now
Seedream 4 does also not work on image generation
It is stuck on this for me
too
Slowest model and bad at image to image, don’t want to see it anyways
Seedream got nerfed in the last two months. It was once very good at i2i
I like to think she's Polyphemus's niece
Rip
Site down ye?
just had a client side error
Yah nano still sucks at editing and composition.
Ironic, you’d figure that an animation would be simple and it probably is it’s probably they don’t want people reproducing frame by frame videos
Sora is really ahead in that sense
OH YEAH I KNOW BUT THE WAN AND VEO 3.1 GIVES THE BEST REALISTIC QUALITY I NEED
Sora Beats veo 3 in a lot of areas
Nano still keeps giving me errors
Yeah nano is still the same
They just updated it and reference it with the ability to search images from Google
But it still has the same issues it had previously
Except now you can get out in 4K resolution lol
hello
Nano is an improvement for me in 2 areas - writing text and posters in image and celebrity character consistency
SORRY BRO 😭😭
In regards to editing at least
Hi 😄
This is what pro generated for me
Meh
Huh?
I don’t know I don’t know how I feel about them
what prompt you used for the dolphin pic?
I feel like nano doesn’t have any like artistic flair to it
Maybe it’s the prompting I’m not sure
It was an edit
I asked to add
And I did it fine but it’s in the wrong direction
srsly? No offense I've been using midjourney since 23, over 100k generations, in terms of stylization nano pro balances stylization and accuracy at its best. MJ just gives beautifully chaotic stuff.
But every time you try to edit it, it makes them weird
I agree with you, but like I said, I don’t know there’s something missing in my opinion
Little too plain
It happens in 2.5, when I found adding foot prints on beach always opposite to the person’s walking path. I will try more in 3 hopefully could bring some good result for you.
Same same
I used to see a guide taught that
You mark details on your original referenced image
Like circling where to add and how
Perhaps we can give it a try?
It sounds like giving model more precise area to focus
But I agree with you that for nano pro, we should expect. It should analyze by itself
The biggest issue is not being able to understand background and composition and images almost nonexistent
I can’t sequence of this together. I’ve been working on this so long. lol
What did you use for this video gen?
When comes to “perspective”, sometimes it’s weird. Proportion of different object loses reality when camera moves
as i said
the model is powerful
but its dumb
doesnt understand intent / context very well
but the question is, can it output the desired outcome with a detailed prompt? yes
but thats too much work for me
For sure sometimes it works really well, but only with simple things
i dont have time to write multiple lines detailed prompts
Like writing an essay lol
But ff/lf become an obsolete
Ever since sora came out
Because the animation is so much more smoother
I had one quick question for Sora, how do you deal with making Sora produce actions based on referenced image? It may sound stupid, every time I do image to video, Sora gives several second pause on my image and produced only sound or unrelated movement.
@echo aurora is ttere any limit on the output size in one go
I’ve been noticing that too just recently got a jumpstart with the prompting
This is a very long prompt
Vs
you are talking about vid gen here
i was talking about nb pro
i have never used vid gen like ever
I was too, but I’m talking about for my case. I use it to generate the videos.
i saw someone yesterday say nb pro + kling gives like the best result
dunno if its true
Hmm what do you mean by this?
idk if its kling but the version was 2.3 or 2.2
It doesn’t matter it’s all the same for the most part
Especially when you’re trying to do first frame to last frame
long vs short prompt. which better
e.g. if the ouput contains like 3k lines of code its always giving error "something went wrong try again later"
sora has first to last frame?
No no I’m just saying for the ones that do
And you have to set up your images right otherwise you’re gonna get a bunch of issues
let me find the tweet
kling already has 2.5
Since the first day of Sora, the issue of first frame has given me a really hard time. For quick generation Grok gives a better and direct output just imo.
Nawh Sora super easy to use
btw @pineapple nano banana pro 2 is having an extremely high error rate though you should know
@keen beacon give me the initial image
how?
the anime image
share screenshot
This can be dependent on a few different things that lead to this error. I've seen cases where 3k+ lines of code can generate just fine, I've seen other cases where it can't. The "something went wrong try again later" error isn't going to be one of those bugs where it's just one bug and one fix will forever prevent this error from appearing type thing. Think of this particular error like an engine's check engine light, the problem could be X, Y, or Z.
If people come across this error the best thing to do is follow these instructions #1417174113092374689 message. Over time, our team will look into the issues and put in fixes to make this error rate message appear less and less.
Okay good to know, I'll look into the overall error rate and if it's something unexpected I'll be sure to share.
Direct for nano pro is totally down. For Battle, could barely see Nano Pro’s result
For Battle, could barely see Nano Pro’s result
Hmm what do you mean by this?
the transitions are not good
Bad images
Even that is really hard to do
And really time-consuming that’s not even a minute lol
Vs Sora
Smooth clean
You can’t really do that with any other model
I don’t know if it’s traffic control or what, like for multiple reference image scenario, 10 outputs may have only one nano pro results. Since for reference image over 2, the random pool only includes a few candidates such as gemini2.5/gpt-1 (mini)/seedream4(high res)/reve-1(fast edit).
isnt that expected?
i mean it should be random
that's the whole point
Hey update on this - it does appear we're seeing a higher than normal error rate for this model, it's been flagged to the team for repair. 
I do so many image edits it’s ridiculous
you can make a movie atp
I wish
just need 24x60x60 frames for one hour
I have this error all the time, although yesterday everything was fine.
No matter what way you do it it’s impossible to stitch together properly because of the inconsistencies
The only way you could do it is by cutting
Same same issue, even the prompts I used yesterday are being flagged as error now
That sounds to me like a rate limit