#nano-banana
1 messages · Page 3 of 1
I think I’d pay monthly for nano-banana
I already pay $20 a month for ChatGPT
Nano-banana will probably be around the same price (I’m guessing/hoping)
How’d you even find it
Nano banana is baetit's better than chagpt
How do you get nano banana? Is it only by battle?
It's currently available in battle mode here -> https://lmarena.ai/?chat-modality=image You won't be able to specifically select it, but you may get it when served two model's responses
Thanks
Bugged mouth
damn that first looks realistic
Because this flux model only generates Asian girls.
I don't know why it doesn't produce American girls.
How to use nano banana on discord?
Website
hi
Anyone have prompt for clothes swap
is nano-banana support image editing?
Yes, it is reallyyyy good with editing pics
thx
Yeah, the best model by far
Can it rival banana tho?
would you answer her riddle?
don't ask I don't have an answer
you know we'll reach AGI once an AI can successfully create an image of a centaur breakdancing
I simply cannot find this model on the website. Can anyone help me?
Is it only available to subscribers?
what
lmarena have subscriptions?
idk
No we don't
Can only get the model in battle mode here -> https://lmarena.ai/?chat-modality=image
damn tysm bro ❤️
there was not, there is not and I hope there never will be
you can't use it directly
it only shows up on arena battles
since it's a stealth model
So nano-banana failed this prompt
I said to restore it to its former grandeur
This is the original image
This is what nano did
But GPT1 succceeded
Here’s GPT1
lmao, maybe nano needs a more like precise prompt?
I don’t think it’s that
Not every model is perfect
Some models are better at other things than others
true
nano is really good at figurines, look at this PEAK
I mean nano-banana also sucks at spelling a lot
No model is perfect *
I make all my NSFW stuff locally with my Illustrious and NoobAi models
that's smart tbh, downside is having a really good pc in order to run stuff locally
I do 😉
I have a 4080 RTX
I’ll get a 4090 eventually
If you want to change image 1 it's better to send it last instead of first
Ok
I also prefer saying "Image 1" and "Image 2" instead of 2nd or 1st image, and also specify that it's the player from Image 2
but multi-image editing is pretty hit or miss tbh
Alr thanks
That is ridiculous
What was the prompt?
1/7 scale PVC full body anime figure, real photo, on a desk, figure box behind it, PVC material, realistic PVC texture, painted details, nikke goddess of victory inspired room and computer desktop
ahhh neat 😮
Neeeaaaattttt
What company is nano banana from
Russia
very likely google but not 100% confirmed
for all we know its by a foreign company that copied google's homework
Google.. with another image model..
Yandex
Ngl I doubt it it’s more worse than imagen 4
It can edit images so like Gemini 2.5 pro image editing
Maybe
Yeah it can edit images the images it generates are worse
Mosko
Moskow?
Moskow
very good city
I’ve seen pretty good images, you just need to prompt it to be hyper realistic if that’s what you’re looking for
I assume so
you are right
I’ve been learning Russian but I forgot everything
Sephiroth. Not the same prompt because I dislike copying prompts.
Will post Cloud too.
Damn, I can make Cloud an emo
looks dope 💪
damn! what prompt did you use?
I just used image edit and asked for a messy fringe hairstyle. It made him look an emo though.
This was the original pic
Just a Pinterest find
gotcha, i'll try to replicate it cause it looks nice, thanks dude
No problem
Hello
Editing more iconic photos
Similar idea
So one of the problems with nano-banana I’ve noticed is it isn’t always the best with anatomy
Here it did the usual Ai thing of giving Darth Maul 6 fingers on each hand
However this is what Imagen 4.0 gave me
5 fingers on each hand
Hi everyone, how to use the nano banana?
Can it transfer artstyle from a reference image ?
Yes
Or even better, both pose and style
Which combined with a character reference image would make it 3 in total
I wanted a Praying Mantis Demon and this is what nano-banana gave me
And this is what Imagen 4.0 gave me
Imagen did what I wanted
So both models have pros and cons
Like I said before, nano-banana isn’t perfect
It’s amazing at the things it can do, but it has limitations also
Personally, I'd rather this cracked img2img model over a pure txt2img model anyday
It’s not even perfect at Img2Img though
For example
I wanted this picture in the style of The Last of Us
ehh, did ok I guess
better than every other image model atleast
Hmm I don't think the style transfered there
But yeah
That’s what it did
it just made it look like an anime
I noticed it's not particularly good at style transfer
Here’s what other image models gave me for The Last of Us style
Well I'm not sure TLOU's style is even that well trained in its training data for it to know
Wrong boyo
That's not how this works
#anva how to do it?
same for coloring, reference image -> flux kontext -> nano banana
it doesn't preserve the art style
it always changes it atleast slightly
That’s one of those word salad prompts 😩🥴
Prompt Minimalism is king
Been doing that since 2022
And before
When I started with a pixel image thing
See if you prompt simply you can get stuff like this
And none of those are nano-banana
All Imagen
nano-banana sadly kept messing things up 😬
These are nano-banana
Notice the anatomy issues with the Winifred Sanderson picture
The Japanese shrine one I wanted it to be Autumn
Here’s what another model gave me
#nano-banana पुराने समय की मुंबई की तस्वीरें, साधारण घरों और गलियों का फुटेज"
And that’s what I pictured in my head
That’s how I wanted it
So I don’t think there’s a ‘one size fits all’ model
They all have their strengths and weaknesses
Some models are beyond amazing at style transfer and replicating styles
Not yet 
There won’t be for a while
All the models are still disjointed
Next year 
Tbf nano is a massive step up from models that came out just a few days ago
so we never know
a giant leap might happen in the next few months, or the complete opposite and we hit a wall
Is there a way to use it specifically
It’s fun, but it doesn’t excel at every single thing
Everyone is enamored with it because it’s the new, shiny thing
Once you really sit down and kick the tires for hours you start seeing the flaws
Nano banana actually good at tasks no other image gen tools had been done reliably
i mean its still locked in arena, maybe its full potential will be there once its in ais
From my tests I would say it excels at nearly every single thing, but there is a few things that GPT-image-1 understands better and some things where Flux Kontext keeps the original image/idea more in line with the original
but from the literal hundreds of prompts i've tested, it was better for >90%
Not for me
Style transfer sucks
It doesn’t know a ton of characters
Gpt image 1 is ass, good at instruction following but yellow filter, slow, drifting makes it quite unreliable for edits
oh yeah, I tested it, it was bad and so I just gave up and did other things with it
until then, I've still been using gpt image 1
But it doesn’t always keep all the details
most of what I did was image manipulation, so style transfer and knowledge about characters/art styles were not as important
Ah
I mean this is nano-banana
And I love how it turned out
But this is GPT1 and it makes me beyond happy
I only used the prompt:
"Turn this into a stone statue, photorealism, displayed at the park"
nano banana does still require elaborate prompting
Nope
in some tasks
Word salad prompts tend to mess things up
It’s the best at preserving character details.
I've been using this for days and i swear some tasks does require such elaborate prompting
Mmm…I disagree
It doesn’t
and for how long have you been testing nano banana
what have you been using it anyway lol
Unless you mean by style transfer which I don’t play w much. The second best would be flux but it usually botches my characters faces.
I kinda disagree with some of your points, it still bad at some things but some things are actually great compared to other image edit models on the market
It really doesn’t preserve details entirely
It changed the faces completely on some of my stuff
Or outright didn’t listen to the prompt
Yeah it does struggle with faces but it still great for most editing tasks other models used to struggle
I won’t deny that this is not an issue but from my testing it’s better at not botching faces unlike flux.
it also has some things it can do no other image gen models can
for example bounding boxes
You know what model this is? GPT1. It did exactly what I asked.
GPT would be perfect if it had editing and consistent character capabilities.
other models would usually struggle this simple task
how to use Nano banana in Arena
I don’t have a use for bounding boxes 😓
i asked it to create bounding box where is the Omnitrix symbol is
but it does set apart from other models
it does have potential
For what?
it is to be linked with Gemini 2.5 spatial understanding capabilities
And we know that for sure..?
meaning to most extent it can reason in the images in 3d
Probably more accurate to say it can understand depth, I doubt it would reason in 3D would be very compute intensive
Sounds like your issue is more of a you issue, there's no one size fits all model, for me this nano banana model is a game changer
yeah thats what i mean
it does inherit Gemini 2.5 spatial understanding capabilities
nano banana should be treated as image manipulation model
i guess we can see the full potential outside of arena
I’ll just wait for OAI to drop an equivalent model.
What we’ll see is a ton of censorship, guardrails and corporate handholding 🙃🥴
Everyone acts like these models by big corporations don’t do that
they did lower the refusal rate on 2.0 flash preview model compared to experimental
It always happens
basically one you can configure like temp, system prompt, full context
I have to admit that this has been a non issue for me.
It’s a pretty big issue
Sora is sufficiently flexible.
Within reason
Prompts get rejected left and right
Even over arbitrary things
Has never happened to me to be honest. Could you provide examples so that I can understand you better?
Ah
I suppose I can understand though I do not share such interest.
You say that yet you have a Fantastic Four image as your pfp? 🤔
At least open source seems to be converging towards making as good equivalent models to the ones that are closed sourced.
Yes 😊
I’ll always prefer open source
I loved the movie but I don’t make images on anything like that.
Ah
I have a 4080 RTX specifically to make things locally
Because I do NOT trust corporations
I wouldn’t say it’s unreasonable to expect a nano banana os quality model next year. Qwen is setting such a good precedent.
I'm not saying nano-banana isn't bad, not by any means
I've had a lot of fun with it
I can't find the nano banana. Why?
Sad, but it's true
😫
Yep, because the same thing happened to ChatGPT
Because all these corporations don't want people having fun
They want you to make what they want you to make
Which is not how it should be
I actually voted bad on one of my nano-banana outputs
That's definitely not right...or good
However
Here's the better one from earlier
I was able to use nano-banana in battle mode (so you have to try several times) but now it doesn't appear for me. I've been trying for 30 minutes!
Yeah, me too.
Maybe Nano was removed?
Nope
I've been getting it consistently
We are just unlucky then
I just got it 3x in a row
Your chances increase if you're using an uploaded image
Since there aren't a lot of models on LMArena that do image editing
But nano-banana is one of them
So your chances of rolling it are high
wow bad luck
Just rolled it again
Left is the original image, right is my nano-banana one
You can't make images in here
It's only available on the website
I know, by mistake I sent here
but how i use nano banana
there no option to select a specific model
it appears randomly in the LMArena Battle mode. draw lots
hello can anyone tell me how to generate image on this server ?
/image A minimalist cinematic video of a bold white “We Bros” logo in clean sans-serif font, centered on a deep matte black background. The scene starts with a shallow depth of field and the camera slowly moving forward from behind the logo, passing around it in a smooth arc. Soft dramatic spotlight from above, subtle shadow gradient falling onto the background, and gentle ambient light reflecting off the edges of the letters. Slight film grain and light bloom for a high-end branding aesthetic. Ultra-realistic textures, 4K resolution, 16:9 aspect ratio.
You can't
You can only generate images on the actual website
ohh thx
but
when i try to generate image the error coming
this
what is solutuion ?
Hmm let me check
What mode are you using? (battle/side/direct)?
There's nothing wrong with the site
I'm using it as we speak
this model
this also
i also reported that bug
Yeah oth of those models are working for me just fine, it's unclear why you're getting that error.
ohh i dont know why
Lets move this convo out of this channel, I'll tag you.
wait
how to use?
It's a model by Google?
when in lmareana battle the nano banana model is not apearing why
how can I use battle mode
Bro where tf is nano banana
try using image battle... u'll randomly get it
i got it in first chance
And btw how do i change the aspect ratio on gpt image 1 direct chat? Its always 1:1
it should have aspect ratio setting on the site of lm arena itself or you can input the aspect ratio
Where is it on screen
what's the aspect ratio of image that you are uploading?
well... i specified the ratio and it gave me result actually
not the perfect ratio but close
Mine when i say 3:2 it just gives me 1:1
i mean yeah it is rally give me 1:1 for 3:2 , but it's okay for 9:16 or 16:9
yup. its google
Why is banana exactly… nano?
Is it a small model for phones?
we will know today.
It's launching today : https://x.com/OfficialLoganK/status/1957908528925909391
And there is some google event today on Android. So it might very well be optimized for phones
Well in case it is, what would happen if we scale it up for computers, to a size of flux or qwen image? Will it be even better?
A great model
How can enter the battle...where is that option in this
WHose model is this?
That would be google’s
does it work with more than one input image?
battle mode
has anyone been getting nano on their gens?
its random, sometimes every single one is banana and sometimes not one time in 50
nano banana is releasing today
Proof?
Less than 24 hours to go until #MadeByGoogle. Tune in here tomorrow👇
great, really hope the EU won't be excluded
How use nano banana ?
on lmarena.ai with image generation toggled on, battle mode.
can i select this model ?
ok
No, it will be randomly chosen since it is a pre-release model still
Is nano still available?
Yes
Thanks! 👍
I assure you there are more reasons than just random why sometimes nano banana doesn't appear 50 times in a row
so clearly this is not just a diffusion model right?
at the very least it must have received lots of plain text in its training to be able to come up with that
Hi, may i know how to i able to access nano banana on lmarena in mobile?
Nice recreation, nano-banana.
Oh. Thanks. Finally understand your quote.
How can we get access to nano banana?
I don’t see it here
which app is this
It’s LMARENA on mobile
only battle mode
It's available out of the arena
Are you sure ?
If is fake it does the same as the LM arena so is ok xD
isnt this just the google pixel 10 launch?
At first glance that looks like a quick vibe coded site
LOL Ahahah no it's not
This is fake trash
Easily exposed by a little prompt generation testing
looks very gpt-5 coded idk or claude maybe but probably gpt5
Ahh just 5 generations 😭
It creates images xD
Looks like Flux Kontext under the hood
And works . But with just 5 generations... Not useful
Why not just use LMArena
It takes too much time until you get a nano banana
Fake website... Fake
I'm 99% sure this is either fake or reverse engineered lmarena, and it's disgusting that they're trying to charge for resources that others provide for free
Removed as it has just 5 generations.
If it were unlimited 0 issues
monololy🔥
So I have to generate and generate before I hit nano banana?
only battle mode
but there is really high chancce that you will get this frist or second try
bruh was nanobanana Qwen?
yes qern doesnot preserve text details
qwen sucks
also not smart enough to understand some prompts
LMFAO
nanobanana fails when given small text/lots of it
small text is messed up
Can I use nano bana from here ?
Bro
Yes 👍
@stray hearth
Reminder this is an English only server please and thank you! 
Seriously how do you generate pictures in the Battle Mode? I can not find it and there is not an app on app stor
Got to click the little Generate Image button at the bottom of the text box.
THANK YOU MAN!
so no release today except maybe in the android photo editor? 🫤
one of the best models bruhh
is it made by google?
99%
yeah cuz it doesnt have the yellow filter
like chatgpt and chinese models
its distinct
I'm happy when the model is normally available and not just in Battle Mode. At the moment, I'm out of luck and can't get the model in Battle Mode at all.
Use editing image prompts, you might have a higher chance.
will try, thanks
fr battle mode sucks
i love that it has no chatgpt style, the yellow filter
i call it the pee filter
Me too. Too much of those ghibli suckers ruined gpt 😆
Lmao 🤣
yeah
its so obviously flagged as AI image
with the yellow filter on every image
aghhh, I just want nano banana officially released
when did it came out
as secret model
its the Gemini 3 image editor
for sure
they'll probably release it with Gemini 3
at once
why did they tease it on twitter then, one day before this day's google event?
lol
idk, is gemini 3 releasing today?
No
damn
Rumoured to be released on September
see, I mad ethis with nano-banana. its really good at context
yeah google is mogging every other AI models
i think almost a week ago
yeah, like I ahve to give same pprompt multiple times. just so I can get nano-banana
nah, send the link
search on youtube Genie 3, its amazing
I’d find it crazy if this test model is actually just a "Nano" model and the release version ends up being even more extensive and powerful.
i hope so
that would be crazyy
why would they name it nano banana lol
rn, I just trained a gpt model for prompting
so, people ignore it
use Claude it's more creative
oh, I did it with gpt and perplexity.
claude is the best at creative writing
tbh
I will make perplexity use the claude model then
GPT in a nutshell
damn
Fr man, takes eternity to generate
160 second to delivery a crap image
Share the result. Curious
nano banana of course nailed it
90% of the time Nano does e good job
Nano is faaaar superior to GPT
It’s rare I get a result where nano-banana wasn’t the one I picked if it was an option
One thing it’s really good at is following detailed instructions
life has become so bitter and sorrowful not being able to use nano banana whenever you fancy
hello everyone, when will nano banana be available?
We don't know. I hope soon.
Everyone who thought they’d show nano-banana at the Google event today 😓
Me at least
He made a video about nano-banana
In this video, we dive into the mysterious and impressive Nano Banana AI model, available on LM Arena. I'll show you its remarkable image generation and editing capabilities, including realistic cooking scenarios, 3D object masking, and creating highly engaging YouTube thumbnails. We compare Nano Banana with other models like Gemini 2.0 Flash an...
Yeah, I watch him regularly.
banana still hidden?
Yes, no release yet.
Is it still available?
yes in battle mode
Thanks! 👍
this absolute beast of a model will be only available in google photos
Why is nano banana producing such bad results today?
I actually noticed the same thing today. I had better results yesterday.
It was definitely great yesterday, but today I was shocked.
Already nerfed?
Even the GPT image is better now. I think they did this on purpose. They're going to switch to a paid version. Yesterday, Google admitted they did it. They'll probably release it as the new Google version.
They admitted where? Can u send the source please?
This guy is a Google VP, he doesn't tweet for nothing
nano is so good... out of about 100 gens it produces my vote each time
@agile oxide IS coming to the leaderboard ?
Who said this? Stop saying unsubstantiated claims without proof
Where was this said?
Wait, does that mean we didn’t use nano banana?
how do folks know this is a google model for sure. btw, the text consistency on this is crazy
if you tell it to write or draw your creator, it will do google
I gave it cow tools and told it to turn it into a magritte painting
it freakin changed the caption 😄
A) Imput Image B) nano-banana
I can see a lot of consistency and almost zero loss of context in the original image.

It’s the first time I use this model, so maybe it wasn’t such a good prompt, but I just wanted to try it
What was the prompt?
‘The theme of the image should be at night with a moderately starry sky and beautiful colors’
Something too simple, but even with that could do something so good
depends if your image is already too complex for you to add nothing but simple prompts to change major details about it
Covering up thirst trap pics is easy 😉 Try something more elaborate now.
How about this? I made this with only a prompt and my face.
That’s sick.
The result is fabulous, but thinking about it, it looks strange that you don’t see a mirror lol
I mean I could’ve removed the cell phone entirely 🤷🏼♂️
Your face is now property of several different AI labs and lmarena lol
Figure 2 shows the product replacing the product in Figure 1, with the background remaining unchanged.
Ryu vs sagat
banana gone ?
Haven't got it in 50 tries again
There's no way this is about "luck" again
I have tried many times without bananas, yesterday and the day before yesterday were not like this, bananas appeared every time, what's wrong with this?
@agile oxide
How do we know banana isn't already secretly taken down ?
Like we can't independently verify if the model is still up or down and have to look for someone else and ask if they still got banana
I guarantee 100% this can't possibly be just "bad luck"
I have not encounter it after 70+ tries in a row now, if this really was purely luck I could have won lottery jackpot with this odds
I also think it's not a matter of luck, I used it yesterday and the day before yesterday, and I have always had bananas
I guess there's NO way to tell if the model is still up other than waiting for other people to come here and verify
I got it 20 mins ago when I was on the site
I'm using a different device and a different browser, still can't get it yet
Overall when it comes to models with codenames we're not going to be sharing information about them.
So simply just the info of it being still up or down is not also disclosed ?
Is there a phenomenon where other people getting that model (nano-banana) will make others less likely to get it ?
🍊
Like, surely whoever is the company hosting that model can't possibly queue so many requests at once, when more and more people on social medias know and are coming to LMArena to use it
If you think there is an issue with specific models you can flag that to me and I'll let the team know. But overall I'm not going to be sharing info about them.
Sure, as stated earlier I have not been getting any result from nano-banana for 50-70 tries continuously, and this odd behavior doesn't seem to stem from "unlucky". The odds of that happening is ridiculous. There must be other factor at play here but only you know what's going on under the hood
I used bananas for a long time today, but they appeared almost 7 out of 10 times yesterday and the day before yesterday, which is what happened
Like for example I don't know if your site has some sort of hidden limited daily credits system behind the scenes or something
It doesn't feel like a finite point, I haven't used a banana since I used it today. There must be some other reason for this
I wonder what that reason could be
Guess we'll never know since everything is so secretive
After only two days of happiness, such a useful image model actually appeared, and today it suddenly disappeared😩
And why sometimes the output is like this ?
Is this supposed to be nano 🍌 ?
@agile oxide is nano-banana supposed to be on any site other than LMArena? because a rival site called yupp.ai has it as a selectable option, but it does not work when I select it (likely because I think they are scraping the model from LMArena and they probably fixed whatever exploit Yupp is using to get access to nano-banana).
also there is a bunch of similar sites that all claim to have nano-banana, you can find them all easily by googling
like this one
Dat one is fake af
Most of them are running something else under the hood like Flux Kontext
🍌's quality is very visibly superior, can be easily tested
Yes
but i think yupp is the most suspicious because when i tried to use the model on there, it did not work. i think it was trying to start a fake session with lmarena using reverse proxying and it failed because of a captcha
Everyone try it now, it feels normal, and now the banana has appeared again
Nice, it's back 🗣️
do you have any evidence of this or only speculation? that's a pretty serious allegation
this is discussion channel
Bano nanana
Are you sure it's not a fraud model under the hood?
Yeah, i made a lot of tests and that is 100% Nano Banana
变彩色
Nano Bana is the ONLY model that can make perfect convert into Line art
If it has reputation like LMArena, there is a chance the provider of nano banana also reached out to it ?
Don't wanna jump to conclusion that Yupp.ai did something something shady to LMArena
Use in yupp.ai is better bcs we can select the model and get the image in first try, and no need to have luck
it was not working at the same time people said it did not work on lmarena
But if that's really the case then this meme would be funny
then it started working again on lmarena, it came back on yupp
This meme is on point haha
Every time you send a feedback you get some points.
So you can generate a looot of images
AI gacha, best description haha
So that Yupp site makes you sign in with your Google account 🤔 Is that safe?
So I did some research
Yupp.ai also pays you for your input and feedback
You can cash out your Yupp points for real money
That sounds suspicious
I'm just overall suspicious of anything involving money payout
Nothing in life comes for free, especially easy money
There's a saying 'too good to be true'
So yes, it’s a legit site
I did my own research like I said
And also
Nano-banana is selectable on there
Right away
But how ?
No muss, no fuss
I don’t know how
I mean like how did they get access to 🍌
Because they’re also an ai ranking website?
They also provide feedback to all the Ai companies
How is Yupp's reputation comparing to LMArena ?
Gandalf lutando contra Dumbledore e vencendo a luta.
Gandalf fighting Dumbledore and winning the fight.
Magmar
How can i direct use of nana banana model, please let me know!!
You are amazing!
Why’s that 😮
Because this is such a good find.
Thank you 😊 It’s pretty fun actually!
You can send referral codes to earn more credits though it should be enabled as an option in some days.
Do we get rate limited?
I hope not
What determines Yupp credit count per scratch ?
Micro banana
I think the length of the feedback
The more you type I believe
Each test costs x amount of credits, if we run out we can’t do more.
I wrote what equivalently got me 350 before, but now only got 58, when the length is similar
So you can’t buy more credits?
It feels betrayed
I don’t think so.
The thing is Yupp actually pays you real money
That’s why
We just want to play with nano-banana
The max you can earn a month is $50
Hey guys Yupp can draw?
Draw ?
I’m only getting credits.
No there’s the button to cash out
On your profile
There is no cash out button in yupp ai
You should gatekeep the link lol you can send referrals later and get 5k credits per each person signed up.
I just know one thing
I’m not like that, sadly 🙃
I also feel bad
Whenever earning money is involved, things always go to sh*t
It’s not
Yupp is a legitimate site
Things stay pure when no money is involved
I’ll use gpt5 to write reviews 4 me.
Why they named nano banana for the model 🍌🤔
Eh, Polymarket has had millions of dollars change hands based on LMArena results and it hasn’t destroyed the integrity of the site yet.
So credit cost changes based on aspect ratio of image.
Just made 400 credits lfg

What do you mean?
I uploaded a 1:1 ref image and it costed me 200 a 16:9 costed me 250
you use the credits you earned in battle mode for direct chats?
We’re using a different site
Oh also
Imagen is still amazing
yes, I'm asking about the other site
I don't have the nano banana option to select from the available models. Is it country limited?
It's currently only in battle mode.
have you seen my comment about yepp and nano banana yet? also check out #1407814660454879364 @agile oxide
Yeah I did see your earlier comment. Like I mentioned before though I'm not going to discuss private models or share details about them.
i understand, but if it is exclusive to lmarena then you should go after them
after the model is gone
Dude
It’s not exclusive to LMArena
i said "if"
Yupp is a legitimate site, number one
LMArena doesn’t have exclusive rights to any models
i didn't say it was exclusive, i am still leaving open the opportunity that it is tested under the same name on other sites
Well let’s see
but idk if a stealth model has ever appeared on yupp before
Hi can you create a image
Now generate a clear passport size photo
LMArena doesn’t let us select nano-banana in direct chat
So people are gonna use it where they can
true
anobody tried nano with very complex requests? Like many details to get it right
Do u know y video arena is not working?
Create a new gmail
How do you get more credits besides doing that?
Do they reset?
A marketing poster for "SOUNDWAVE" headphones with a clean, gradient blue background. The central headline "Aero Pro" is in a dynamic white font suggesting speed, placed above a yellow rounded rectangle label reading "Open-ear True Wireless Bluetooth Earbuds" and featuring "Freedom Wing". Centrally displayed are two floating off-white open-ear earbuds with rose gold accents and an oval charging case bearing the brand logo, with two yellow spherical accents at the bottom, separated by a thin line from the slogan "Lighter More Comfortable A New Open Audio Experience". The overall aesthetic is coordinated and premium, with a strong 3D product presentation highlighting its lightweight and open-ear design.
by voting in their battle mode right?
By voting and gaining credits.
Then you need to cash out teh credits, but its no working rn
No, cashing out is for literal money
That’s what the Yupp credits are for
You cash them out for actual USD or Crypto or whatever
So the only way is by invites.
But it's not working aswell
so you cannot use the credits you earn directly to generate images?
Well I think so
Each gen costs credits
But those credits can also be turned into cash
ok, then just vote on battles to earn the credits
So I honestly don’t know
the what left
I received 9k credits, but I can’t use them to generate images.
Robert Downey Júnior cantando em português do Brasil “eu vou estar no Brasil yes “
yeah yupp credits on image gen will run out very quickly, especially using using models like gpt-image-1 (500 credits just to have it as an option). if you don't select any models, you should be able to net credits on image gen. nano-banana is in the cheapest credits category. idk if you'll see nano-banana yourself when doing yupp battles, they like gemini 2.0 flash and flux kontext for some reason
read #1397655624103493813
I had 10 credits left
Now I’m back to 200
Just by using Nano Banana and giving feedback lol
I always give feedback
they probably bumped you back to 200 because you are not supposed to run out of credits, they just put you back to 200 so you can do more battles to fill them back up
No show to me /vídeo in arena
Help me
you have to wait for the bot to go back up, i pinged a person who will let lmarena know about it
Using GPT to create cool composition and Nano to fix the colors and light
So I’m really, really liking the results I’m getting with Imagen 4
Nano-banana is doing good but…
Those are the nano-banana ones of the same prompt
Imagen got Chun-Li perfectly
Just from prompting
/image "A professional Gujarati businessman, medium wheatish skin, neatly trimmed beard, short well-groomed hair, wearing a modern tailored business suit with tie. Confident corporate headshot with warm smile, photographed in a clean studio setting with soft natural lighting and blurred office background. Ultra-realistic, high-resolution, photorealistic, professional LinkedIn-style profile photo."
hey @late thorn Hey do u guys know like - I am upload image of my chracter and i need pose and exprssion of chracter in another image. But am not getting
bot down
u can't generate images here bro
Send both images, please
but still bot is down
and i see in #1397655624103493813 i can generate image here
so image1 is my chracter and image2 is the target pose
see
Nano Banana really isn’t managing to transfer the pose.
anything else you can recommend?
Left is the original, right is nano banana
how'd u generated the same pose?
Hello
Good prompting?
Idk, I’ve been doing this stuff for years
well... bro i mean you did... but like can u help me with above
this one
The one you did was an easy task.
Try taking that model and putting it in the same pose as the girl.
I find that quite unlikely.
yea it's like am trying for hours.... but not getting any progress
can u do this ?
Hang on
yea.... it's the right pose , how you get that
I found yours closer to the reference, especially the confident pose and the head.
Because I made a depth map
Of the girl
but the face.... is not right i guess?
yea but like idk if the face is correct
Look guy
The models aren’t perfect
That’s just how it is
Something will always be off
In this case, the best option would be a ControlNet for pose, to capture the eye position and the turned head.
No matter which model you use
ohkay
idk how to do that... i have amd
thanks for ur time
You can use runpod and rent a GPU
Dont generate
and like what would it cost all
/image
It depends on which GPU you’re going to rent, which AI model and configuration you’ll use.
There are many factors that affect the cost.
ohkay then... i can recreate any pose with expression ?
/image tall gaint fat woman
not here bro
draw the new super car on the city road
πudio ghiibre bus
May I ask why I don't have the permission to post and generate videos in the video generation channel? Does anyone know how to solve this?
Servers are down i think
i think this an answer about your Q
Hey everyone - looks like the bot is currently down. Team is looking into it, going to turn off these channels for now.
can you choose which model you use?
hi
In “Chat”, yes.
Also, hi.
not in picture?
You can use image generation models via tapping the little image button within “Chat”.
Video/ Cartoon penguin eating ice cream
Unfortunately, it's not available right now.
That doesn't work for now.
#video-arena-1 car
how can I use Nano Banana
Sad, but its better this way, the competetors have a thing to work now
Its on yupp.ai i think
hi
Thank You Man
Hi
how to use nano banana
hello guys
do we get paid using yupp.ai? KYC needed?
yupp.ai pay you to use?
genrate image of advanced amino acid
yea... but not available in my region
It has yupp credits it asks you often to choose between two models on how good they were, then it gives you a card that you can scratch and win credits from each time
But yes if you ran out of credits i think you should pay idk if they reset daily
Tried a Kerala style ghost called 'Yakshi' in LMArena and it is impressive. It is nano-banana.
how to use nano banana repeatedly.
so, nano banana is the model behind the new photo editing feature in Pixel 10 presented today by Google ?
it's crazy good for a nano/edge model
WTF, can't even imagine how good the full version is
A man is walking and singing in a garden. It is nano-banana
Hello everyone, has anyone here gone to the end of their credits yupp? I still have some left but I can’t find buttons to buy some.
I still have many left over.
A hardworking Yemeni farmer wearing traditional clothing, standing in his terraced farmland in the mountains of Yemen. He is holding a farming tool while tending to his crops under the warm sunlight. The landscape shows green terraces, stone walls, and distant villages, capturing the beauty of Yemeni agriculture and rural life."
Er…
Generate a video ad for the advanced amino acid formula, along with benefits
"Make an 8-second video. Two Egyptian girls enter a large wrestling ring, and the audience around them is cheering, screaming, and clapping. The atmosphere is suspenseful and fun."
What’s up bro ? 😅
/video-generate an old man hunting the deer on the winter
Yupp ai is fake asf
do middle east women wear that usually or am i tripping
the pic on the right
who should we ask
oh gemini
You are absolutely not tripping. Your intuition is spot on.
The way the woman is dressed in the picture on the right (ideogram-v2) is highly unusual and not a typical or culturally accurate representation for women in Saudi Arabia or the broader Middle East.
yea
Cinematic portrait of child surrounded by daisies, curly brown hair moving in wind, ultra detailed skin, Nikon 85mm f/1.4 lens, shallow depth of field, soft golden light, steady close-up
i kinda wanna do jame fridman photohop if he wa jut another photohopper tbh with that bot maybe, idk if ill do it
/carro do futuro com piloto de corrida.
you're saying that it runs directly on the pixel phone?
dunno, just a guess, maybe that's why they named it "nano"
that would be crazy
timelines shortened, by a lot
Meet Pixel 10, Pixel 10 Pro and Pixel 10 Pro XL. Powered by our new Google Tensor G5 chip and our latest Gemini Nano model, they’re our most personalized, proactive and helpful Pixel phones ever.
aren't the weights of the model on the phone? Can't they be extracted?
coloque no bebe uma linda fantasia de um urso panda fofinho, a mulher esta com a roupa amarela
dunno, I'm sure they thought of this
I am a bit annoyed. why this model is not yet out ?
Logan shipping nano-banana this week? super hyped
[Opening Scene]
(Bright, clean, and modern background with soft lighting. A close-up shot of a person smiling confidently.)
Narrator (Voiceover):
"Introducing Dr. Dent Premium Teeth Whitening Strips – your ultimate solution for a brighter, whiter smile!"
[Cut to Product Close-Up]
(A hand holding a pack of Dr. Dent Premium Teeth Whitening Strips, showcasing the sleek and professional packaging.)
Narrator (Voiceover):
"Our advanced formula is designed to deliver professional-grade results from the comfort of your home."
[Cut to Application Scene]
(A person opening the pack and applying the whitening strips to their teeth. Close-up shots of the strips being placed carefully.)
Narrator (Voiceover):
"Easy to use and effective, Dr. Dent Premium Teeth Whitening Strips fit seamlessly into your daily routine. Simply apply, wait, and reveal your dazzling smile."
[Cut to Before and After Comparison]
(Side-by-side comparison of teeth before and after using the whitening strips, highlighting the noticeable difference.)
Narrator (Voiceover):
"See the difference in just a few uses. Our strips are formulated to remove stains and brighten your teeth, giving you the confidence to smile brightly."
[Cut to Lifestyle Shots]
(Scenes of people smiling, laughing, and enjoying social interactions, emphasizing the confidence that comes with a whiter smile.)
Narrator (Voiceover):
"Whether it's for a special occasion or everyday confidence, Dr. Dent Premium Teeth Whitening Strips are your go-to solution for a radiant smile."
[Cut to Call to Action]
(A person holding the pack of whitening strips, smiling at the camera. Text overlay with website and purchase information.)
Narrator (Voiceover):
"Experience the difference with Dr. Dent Premium Teeth Whitening Strips. Visit our website or your nearest retailer to get yours today!"
[Closing Scene]
(Logo of Dr. Dent with a tagline: "Brighten Your Smile, Boost Your Confidence.")
generate a side profile of the person in the image.
nno
|
v
HELLO
source?
#1397655624103493813 generate vedio Pixar style 3d animation 10s Scene 1 (Intro – 0:00–0:08)
📸 Visuals:
Bright playroom classroom with colorful posters, shelves of toys, and a round craft table.
Friendly teacher holding up a plain white paper plate, smiling at the kids.
Children (diverse group, 5–6 years old) sitting eagerly with their own paper plates.
🎶 Action: Teacher starts singing cheerfully, gesturing with the plate.
Pure speculation based on Logan's tweet cadence, but I'm fully on the hype train
Done
fr
google made the internet
is this made with two input images or with a single one with the two base images side by side?
now it's reinventing the internet
Cinematic portrait of child surrounded by daisies, curly brown hair moving in wind, ultra detailed skin, Nikon 85mm f/1.4 lens, shallow depth of field, soft golden light, steady close-up
