#nano-banana
1 messages · Page 5 of 1
How to get acces to nano bnana
go on https://lmarena.ai and select the image button, put in your prompt and see if you get nano-banana
you may have to do it a few times before it shows up, especially if you are not uploading an input image and are just doing text-to-image
it is a stealth model, so it can not be manually selected
?
i can't see that
Can anyone tell how to genrate. Video in 9:16. With veo 3 model
I refreshed the page and it disappeared, strange. Could that be the final model’s name? Musa is the scientific name for banana.
you gotta cut the center out of a 16x9 video
If i get one genration so. How i can keep genraying with nano banana
Not like that i nees to genrate
I have student pro model
So, Google is behind Nano Banana 🤌
Can anyone tell
Plz tell how to keep genrating with nano bnana
Just keep trying it's luck
There is no way to guarantee it
It's designed that way
Then try
But nano bnana not availabe on single mode
had to feed this back to nano banana for an edit cuz it generated a weird stone head at first
its creepy tho
yea kinda normal bullying right... just.... tossing someone's head
shes a dullahan her head ain't attached
just for fun
cool... ig
woo image quality is insane
the architecture doesnt quite make sense though ehe
LOL
it's designed like that on purpose, it's not a glitch
A spooky-cute digital illustration of an adorable fluffy black kitten with oversized, neon-glowing emerald green eyes. The kitten is depicted with small, sharp claws, red devil horns that curve playfully, delicate bat wings, and a vibrant red tail tipped with a black arrow. It sits perched upon a cracked hellstone platform, adorned with a small patch of red, armor-like fur on its chest, while an ethereal, swirling purple mist fills the space around it. The entire scene is isolated on a deep black background, ensuring transparency and ideal suitability for a T-shirt design, emphasizing the phrase "Cuteness is Hell".
gravy boat
not here bro...
is it true.. 3 sep
the day before silksong
yea was thinking the same
anything new to get more banana models?
All the fun I've had, have been with nano banana image editing
The model makes it look so easy
this is interesting - this confirms that it's a gemini derivative, not just a one off for pixel phones or whatever
Hey, I have been trying to reach out to you to report an issue with the banana model on LMArena, could you help me ?
it's always better to state your problem to everyone than to ask a random person for help
I know, but you're a Moderator, not some random, that's why I specificially tried to contact you. So normally for other models, if the output has an issue, they would still appear and show error, like in the screenshot attached below. But for nano-banana, this model would simply not appear at all. This gives off the confusion that the model is taken down or disappeared from the site, when it could have just showed up and gave us error indication like this.
So I was wondering is this behavior an unexpected bug, or intentional design ?
i don't think any of the moderators here actually work on the site, and i have the weakest connection to lm arena since i wasn't hired for this
Oh, sorry, I just assumed the Moderators and the Developers are same, because normally I only see just Mods and normal Online users in the right tab
So who can I contact to report this issue ?
are you sure it's an issue
"if the model isn't available it isn't chosen in battles" sounds rather normal to me
But the model is available, it's just that when it should have showed error output, it simply does not even show up at all, unlike all the other models that actually showed error
I have tested this many times to be sure it's not luck, but a repeated pattern/ behavior
This has confused many users into thinking the model is "down"
is that to say that you chose it and it doesn't show (not sure if that's even possible), or that it was randomly chosen while battling then shows then disappears?
Not just me
If you have that model you can build whole story not even story make full documenty animate movie
This isn't direct chat, this is in Battle, specifically the nano-banana model
okay, so the model shows then disappears?
No, what I mean is that:
Normally, for all other models, when the output encounters an issue, it should still show up, whether on the left or right, to display warning like in the screenshot below, right ?
But, for nano-banana, the model will simply never show up at all, if the output supposedly encounters an error. Instead, it will be always other models appearing. This is how me and some other users here, sometimes didn't get nano-banana like 50 to 70 times in a row, never encountered once.
It's not luck, it's a pattern/ behavior that I only noticed after many tries.
how would it "encounter an error" if it didn't show up
it sounds like nano banana just isn't chosen in battles, and that doesn't sound like a bug in itself
But it happens when for example the input contains keywords that 🍌 model filtered, its censorship is stricter.
When I use a certain input, the banana would not appear for like 50 tries in a row.
I switch to a safe testing prompt, and it would appear at frequent rate again.
Then I switch back to the prompt earlier that may trigger the model's censorship, and the model would permanently never show up again. It's a pattern
imagine nano banana replying with text content to the input prompt lol
ok that sounds like good behavior
no need to invalidate a battle, prevent a vote, upset users, etc by showing "ruh roh this model can't handle this" instead of just choosing models that will comply ahead of time
Good behavior is still showing up and display error like this to make everything clear, since different models have different limits to what they accept
no?
some models will accept this input, but not other input, and vice versa
Why not ?
why should a model that doesn't want to take the input be forced to take the input and error out
but how would it know it doesn't want to take the input, to make the decision to NOT appear to begin with ?
The decision must have been made regardless
do you mean "i know"?
But you're saving another model's computation to generate the output
I'm referring to the 🍌 model
I can build whole documentry
Nano banana?
In order to show error, or to simply NOT show up, the model must "know" whether it accepts the input or filter it, right ?
bro how to only nano model for image
Any one know how access quicky
on the site
i don't know about the internals and whether google's providing a "can nano banana be added to the model pool for this battle" api or whether lm arena just did a keywords.some(k => prompt.includes(k)) on their side - either way i don't think models should show up and just error out, the arena should minimize errors not maximize the pleasure of nano banana users imo
By choosing NOT to show up, instead give it to other model, the site is wasting computation of another model, when it could've saved energy by showing that error. This clears the confusion for the user, so they won't keep re-trying to get that model in vain/ false hope.
"wasting computation" ...are you not voting?
The user is falsely lead into thinking it's unlucky that they didn't get the 🍌 and will keep retrying to get it, wasting computation
just tell them face-on that the input has error, so try to change it
can you answer the question: are you voting?
Isn't honesty better ?
Ultra-realistic portrait of a Cristiano Ronaldo lookalike with slightly exaggerated features for comedy, lean athletic build, sharp jawline, slick short black hair, wearing a red football jersey with golden accents and muddy shorts. He is standing dramatically in front of a small Indian golgappa street stall, holding a shiny copper lota like a trophy. Golden sunset lighting, cinematic mood, busy street background.
Ofcourse, how else would it let me continue ?
which one is nano banana
how to use nano bnan?
On the contrary, the more you retry, the more votes lmarena gets. The arena likes it when you retry
It's not available in direct
sketch to painted illustration
1 -> 2: "Extract the character onto a white background"
2 -> 3: "Paint her with blonde hair, white skin and dark blue skirt"
ok, even i think this same
But the point is that the user is trying to get a certain model to vote for it, and it would never appear, and the site is not bein honest up-front about it
nano
then i think everything's working as intended, you're ranking how different models (from the pool that are eligible) do for your prompt, which is the point of lm arena
that said i now understand how this experience could be infuriating if you just want to get to nano banana... i guess there are a few ways to fix this besides making it show up and error, maybe there could be a toast like "you're battling the subset of models that will take your prompt" or a direct image generation or battle between fixed and random option for nano banana
If there is a problem with the input, just say it so we can know to change, instead of silently making the 🍌 model never appear, confusing users into thinking the model is taken down. Isn't honesty better ?
Why does it show error for all other models, but except specifically this one ? The inconsistency makes no sense
You cannot choose, only works in battle mode.
"Why does it show error for all other models" probably because it's a genuine error from the api
It's probably the same for 🍌 because I tested it in yupp.ai and it did show error and tell us upfront about it
odd
it told us there's a problem so it switched to other model, it didn't silently use another model like LMArena
I'm really impressed with this! I uploaded 4 images in an image prompt, covers from a particular comic series, and asked for the {title} {issue no}. It replicated the issue number perfectly, even arguably improving on it. I won't add the others to save a little server space, but I added three others of the same series. I would have been happy with a generation of a random one just 'inspired by' these covers, but it nailed it. 🔥
Ignore the extra leg, that's just some artistic license
Assalamualaikum
Wa Alaikum Assalam my brother
@humble verge I want some help
What is it ?
Ai video make for my product advertising
Im new in marketing
You can try this section
mm ok
But I never generated videos before so I couldn't help further, sorry
buddy ehich ai model?
Good model. My only wish is to respect original character designs and stop making them look like generic A1 slop
maybe a prompt can solve this?
I hope so
Let's try
Usually good models respect the og artstyle
I hate prompt engineering, ton of unnecessary babysitting instead of just "hey do this". Ends up making me all the thinking instead of the model, what even is the point of ai then.
We may not have reached this level yet
I think this will be possible once the models start learning our preferences based on our previous requests
I literally ask GPT-5 about obscure anime similar to Madoka Magica and hated for reasons that are found in any other major anime ever, and after 5-10 minutes of thinking it correctly identifies it among dozens of other shows.
When I ask Qwen, it annoys me with its politically correct commentary about how UhM aLl FlAwS aRe SuBjEcTiVe and then proceeds to tell me how Eva does not qualify because its flaws are "outweighted by the quality of the work", making up new rules, completely misunderstanding my prompt and just never finishing the task successfully
/crie uma imagem de um baby sorrindo
What kind of prompt engineering should I do here?
I'm not good at prompt engineering either
"No, Qwen, identify shows just criticized for being boring regardless how great they are. No, do not come up with your pc nonsense. No, don't make up rules. No, stop cheating and count each flaw honestly"
And it still will not finish the task
Meanwhile GPT-5
Hey ChatGPT, which anime similar to Madoka Magica are hated for being boring confusing nonsense?
Sure buddy, here's the list: Neon Genesis Evangelion...
Prompt engineering sucks. Machines should ask us questions when they're unsure, not making up garbage confidently.
Nano banana (on the left) completely ignored the request to keep it in the artstyle of Madoka Magica.
GPT Image respected it, for comparison
Prompt engineering is not without its limits ):
What was the original image?
The Girl with Her Mother
Impressive
I would have never guessed it was edited.
Thanks 🙂
You're right, it really looks natural, looks like its his little sister.
Yeah, Nano is very good to fake celebs selfie
The prompt 
Where I can test nanobanana?
where can i try nano banana ? cant find any link
Should be it
I mean I got like 2 credits
Wym? Is it a scam?
i check lmarena and didnt find it
yes, it's probably kontext or qwen-image, you can test by giving the same prompt to it and lmarena, they won't give the same result
Damnn that’s why i can’t find
Some mod needs to create a specific section on how to generate images in Nano Banana.
The HowTo guide doesn’t explain anything about it.
Make a step-by-step tutorial with images, explaining that it only works in battle mode.
Every 30 minutes someone shows up here confused.
I will test it now
No, yupp has nano too.
@agile oxide
weird how nano can be unable to do the same thing it did to a similar image
flux kontext max is the best ai editing image i used but its out of list for few days idk why
Reddit said that nano should release around 3rd of September right?
no matter how many times I prompt it, it can't do the pose on the left on the girl on the right
Yess like a list of models and where we can find them (only battle or even in direct chat)
meanwhile the left always works, even though sometimes with an extra limb
it's a stealth model, not really meant to be publicized like that
Damnn
it's just that nano especifically is so far above the rest it blew up in popularity
:)
Nano b
The sea right now in barcelona
/image - make this shot look more realistic.
Nano banana so dumb rn
Too many people trying it so they are probably toning down the compute level
make a full button jersey
Send me the prompt let me try something
And the images
Place this bag in the scene with proper lighting and shadows so it blends naturally into the background.
Position the bag in this exact spot and angle without altering its orientation or design.
Adjust the colors, lighting, and shadows to ensure seamless integration.
Add reflections and refracted highlights that catch the light, with edges that glimmer with a crystalline glow.
alien and man riding horse
The color is not right, needs to be more warm.
Like this output from this morning
Probaly nerfed
I changed the prompt
With and without the images as reference. Both nano
My eyes like this more idk I guess preferences
But it's important to keep the same scene and composition
It's the briefing
Did some photomanipulation in Photoshop
Oh gosh the spacing perfect
nano isn't very consistent
might be why Google is taking so long to properly release it
?
Same lmaoo
When it happens reload the page, don’t refresh it
I can’t find nano anymoree
Nevermind
Nano
oops
Hahajaha she about to fall
veo 3 fast 3 free videos for all users for this weekend
wrong channel, use #video-arena-1 and read #1397655624103493813
Best game of the year
It's always the damn GPT that is slow like a 🐢
where do you guys use the nana banana?
Same thing happen to me
Maybe it’s the prompt fault
Nikola Tesla riding a Tesla ? Lol
yo thats nice
in lmarena battle mode
got it
I created this using a technique I developed and am still testing.
Soon I’ll reveal my workflow here.
The reflection on the car screen is insane.
nano banana is so good
Yes brooo
Omggg
Would be more good if there was text-to video or image to video part
But it will be not free as we want
Cool
My new workflow
So cool
Looks like Steganography works well with nano banana too
like a small additional prompt
Hi! Do you guys know a way to test nano-banana outside the Battle mode in LM arena?
theres no other way
Oh, okay. Thank you!!!
theres news that it might come next week
ye
looks like it works well with steganography too
it gives better guide
Woah! I didn't test it that way
try it out
The second image changed your artstyle entirely
The GPT one has a much better style
That's the objective
To make it worse?
Art taste is subjective
the lighting looks better
This ultra yellow GPT and texture is awful for me
I find the GPT one better still
The sky changed completely as well in the second one, while the GPT one sky is more moody and atmospheric
Well, we have different taste to art.
I work with art +15 years, so my references are different from yours
You act like I’m young lol
I have many years in art, too, buddy
Don’t be a snob now
yeah, it looks soulless.
idw, to enjoy the GPT aesthetic, dont look you have good taste, sorry
Personal attacks, huh?
Well, sorry the overly bright second image doesn’t impress me
This image is trying to look photorealism, and not artistic bro
Art is subjective
Photography is also art
Artistic, done
Hell, most people would consider all of the Ai stuff we do ‘not art’
I only consider art if you generate every part with control and manipulate after.
Only type few words and get the output is not art
Yeah I’m not touching that one
The bad thing about GPT is that you can recognize it almost every time
Of course
Just like nano banana also has a default artstyle if you don’t prompt differently
Most models do
But some creations with gpt image are good and artistic
Yes, that's the reason i'm using other AI to remove the GPT look and only use the composition.
GPT texture and colors are awful and boring
But GPT is great to composition
That would require subscriptions to all of those things
Not everyone has the money for that
I know. I have all subscriptions because i work with it.
And i need to test everyday to adapt my workflow to next projects
i have a collection and i easily get them for free lol
i also use yandex
I'm from LATAM ☠
I use Yandex to get images for making LoRA
I make stuff locally because that’s specifically what I paid $3,000 for
Me too haha
Because I don’t trust corporations
I won’t be told what I can/can’t make
Also, and this needs to be stressed
99% of high end models are trained on native English and natural English
Nowaday i use AI to work with advertising.
I need to upload more projects i've done
If you speak English oddly your results won’t be good
I’ve seen a lot of people calling the Ai ‘him’ or ‘her’ and it’s neither of those; it’s an ‘it’ because it’s not even a human, it’s a machine and machines don’t have pronouns lol
with nano banana you can easily create one
Well yes
the database i meant
All my nano-banana images are all made using only 1 image
Or two if I want two different characters
Or none if I’m just prompting
i did found out steganography works well with nano banana too
its like place a small prompt onto the image
just inside the image's code
like writing guides
or description
i tried it out on gemini flash image
and it indeed increases quality
GPT to Midjourney + Nano Banana
last time i tried out gemini flash image, it indeed made this amazing album cover (no nano banana)
So, you first do the image in gpt because of the good composition? Did you try generating the source image with Nano-banana? What was the result?
I did, but i found the GPT composition far superior
Like this example. Base generated in GPT
Hahaha what a good image
fortnite literally
The composition is great, but the colors and texture from GPT is awful
both characters added in fortnite
Unfortunately same.
We play the game of life in hard mode
Get me out 😭
It would be amazing if Nano Banana could generate a wider range of art styles—like cinematic visuals, high-fashion photo shoots, or even distinct video game aesthetics
….It can
You just have to have a knowledge of that stuff in order to prompt it
I’ve made plenty of nano banana images of various video game styles
Okay, so generate the most cinematic image from Banana, please.
Xbox 360, Playstation 1 graphics?
You need to specify specific games
It’s not just ‘Xbox 360 style’ because there isn’t one default style
I'm no talking about a game, but about the console aesthetic
Again
You need to specify the graphics engine
If you specify these things that’s when you’ll get what you’re looking for
If you just slap on a generic theme then it won’t
Okay, so send a cinematic image and bully graphic style image (PS2 era)
Nano can't do a really cinematic image
I’m not going to prompt a ton of images just for one thing
You get nano banana more if you’re editing images
Straight prompting makes you less likely to get it
Well, you told the banana generate cinematic images, but it's a lie.
I just wanna you to show me
maybe this version is just the beta version
What kind of English is this? 😑
not the release
Not native, that’s for sure
Of course, how many languages do you speak?
Did you not listen when I said that all these Ai models are trained on ‘native and natural English language’?
English and Japanese buddy
whats native
I’m American and I’ve lived in Kitashinjuku in Tokyo
I also speak some Mandarin and Cantonese
And Spanish because of my job
But English is my first language
Bro, generate a f* cinematic image and PROVE you are not wrong. Because you are talking bs
Like speaking English like a native speaker
You CANT generate cinematic images with nano banana
Not whatever gibberish that is
example?
You can it’s just a skill issue buddy 🤷🏼♂️
What example do you want? Like how I’m typing now
Prove me.
You can't
That's simple
This isn’t native English
So the Ai has no clue what you’re trying to say
If you type in native English your prompts may work better
I'm wainting the image
I’m just telling you how the Ai are trained
Stop generate these slop anime stuff and prove me
See? You say bs and can't prove nothing 🙂
i mean, nano banana doesn't get descriptions decent; just with simpler or lack works better, like 'use character in blahblah pose'
Because you’re using natural language and not a word salad prompt 😊
like stable diffusion
Does that make sense?
A word salad prompt is one of those huge prompts with a ton of nonsense words
Natural language is speaking simply and straightforward
Also
Also lot models have token counts
Bro, keep uploading 2 images from characters and prompting to generate the most generic japanese style
If you go past the token limit it simply won’t listen to your prompts
Ukiyo-e is a classic and historic style
Been around for thousands of years
or maybe the top P is too low
It just happens to be one of my favorites
i feel like this version has it low as test period
and maybe once it releases, it will increase
We found this out a long time ago with other diffusion models
Some only have a 75 token limit
So it ignores the rest of the prompt
That’s just a hard limit baked into the model
I have no idea what nano banana’s token limit is or what any of the other settings are
since nano banana is possibly by google, i dont think it will be lackier than the gemini flash's one
it shall be way better
As we’ve noticed, the simpler the prompt the better it works out
It’s been a long time since the models have worked that way, because of the T5 XXL encoder.
Like Flux
If you type a giant prompt with a lot of nonsense words it won’t work out so well
i tried pull a dithering filter, but it doesn't know what dithering is
If you have like 20 different art styles in one image it’s gonna end up messy
I’ve seen prompts like that
Where people used a giant prompt and the result looked nothing like what they wanted
i did use steganography on gemini flash's edit, and it does increase the quality better but never messed up
I haven’t tried that before
What you’re saying only works for older models. Nowadays, with the T5 XXL encoder, models can understand very long descriptions.
Reve, Ideogram, Imagen, Recraft, Flux.
it'd be strange if this nano banana doesn't work well with larger descriptions
Just because it understands them doesn’t mean that’s what the output will be
All google imagen models (including nano), lack creative angles, which makes every image look the same as the last... his not wrong.
You can specify poses and the like
tho nano can do literally any style which no other model that i know of can do like it
You can use depth maps if you have to
i did generate some album covers with the flash's one and they went too great
I’ve been saying that not one model is a one size fits all
Nano is a different case, since you can give him a reference of the pose/angle you want and he will absolutely nail it
They all do different things
but imagen sucks
Oh?
Here are some Imagen things I’ve done
Nano banana struggles at this
like this one
Nice one
very good style emitating, still...the angles honestly lack depth and feel ai
its always front angle
or this
change bird to human keeping same style
yep that is indeed nano
Easy task, simple images. Even MidJourney from 2023 would generate these images
yeaa its so good
You’re wrong.
Midjourney is horrible at it
its crazy we can detect images and check them they're AI
These images have no depth or composition; it's just a character facing forward
nah midjourny has its own unique style and is not good at mimicking
Now you’re getting blocked
I asked you to stop
You keep going like a whiny little baby 🙄
Are you offended by something you didn’t even create? The author of these images is the AI, not you
oh... you guys are upto something?
Do I have to say it in Brazilian Portuguese for you to understand?
I doubt you can do it.
Great use of AI
Eu lhe pedi para parar de ser um idiota e você continua fazendo isso.
thnks
And now we’re done here
this is not worth a fight, both of you are right.
That guy is being snobby over nothing
i did like this one pose that gemini flash's edit generated (not nano banana)
And acting like he knows more than the rest of us
I used to go to art school myself 18 years ago when I was younger
I couldn’t stand snobs like him
Pretentious jerks
I said Nano couldn’t generate cinematic images, and you claimed I was wrong. Yet you still haven’t proven it. You’re the typical guy who talks a lot without knowing and shows nothing.
Because I don’t want to waste my prompts on your stuff? lol
Go make it yourself
Go figure it out
See? LOL
Use natural language and see if that helps
or use gpt 5
Amazing, dude
Or that
That doesn’t prove anything
I literally said
Now listen closely
I don’t want to waste my prompts and proving a stupid point to you
no one needs to prove anything, its just opinions you can agree or not, thats it. be fun guys.
You can make the image yourself
wont be fighting be a rule breaking here
I’ll give $50 to anyone who can generate a truly cinematic image with Nano Banana. It simply has this limitation—accept it. Don’t speak without knowing.
what type of cinematic though
Go look on X
I prove my point. I don’t speak without knowing.
See what people are making
you know that you can give nano a reference for an angle or pose and he can easily do it...right?
Nano isn’t a ‘he’ it’s an ‘it’ because it’s a machine and machines don’t have pronouns
Cinematic does not have a rigorous enough definition in this context, thus it's unprovable whether something is truly cinematic.
Like these.
But from prompt only, and not using as base to change something.
Cinematic light
Cinematic Color grading
Cinematic camera angle
i feel like that would be easy
But it’s not according to the Great Wise One
add soldier in full gear facing castle
the problem is depth but still easy
From text to image?
Nah
Skill issue
inner depth i meant
According to the rich ai expert, any image can be rejected just because he wants.
i was talking informally, and this is just discord why would i need to prove that im a master degree at language.
Here’s an example of an AI that can generate cinematic text-to-image results: reve.art
i wonder how images are going to be in 2 years
Nano Banana can only create cinematic images through image-to-image, by providing a base image. It can’t generate them from scratch
I just get confused why a lot of people give pronouns to a machine lol
maybe the accent
It’s because non-native English speakers speak that way
The thing is
it stretched resolution a bit
Since nano banana is random you’d have to prompt a lot to maybe get the image
If you upload an image to edit your chances of getting nano are increased
im getting it on first or second try on webiste
Because you’re editing images 🤦
what about steganography?
yeah probably
I haven’t tried it
Do you think it’s possible to generate cinematic images like the example with Nano Banana using text-to-image?
idk that works
you can even bypass filter through that method
Use GPT
It is extremely simple to get a prompt that’ll do it
like upload restricted images to gemini using jailbreak
placing the jailbreak within steganography
i found out steganography by a jailbreak discord server
If that guy wants cinematography so badly he should just use a GPT prompt
The problem isn’t the text, but the Nano Banana model wasn’t trained to create images like that. That’s why it’s an impossible task, regardless of the prompt
it totally work
If it’s a Google model that doesn’t make sense
Veo 3 can do that
Nano banana is presumably Google as well
Cinematography is an easy thing to make
why so?
Hell I did it with Stable Diffusion when it first came out
What I mean is the style
Nano Banana has a different structure from the Imagen 3 and 4 models. I repeat, it’s impossible to generate those cinematic images from text-to-image with Nano Banana.
illustrious makes it easily
I mean you can also do a ton of different photography styles
u can ask gpt 5
He doesn’t want to listen 🤷🏼♂️
yea.... but i wanna know why? like because of more data available for thoose or what?
i always use ai myself when questioning something
He wants to make us all sound like idiots
Bingo!
yea it make really good prompts
there you go
The more data available the more easily a style is
That's no cinematic
Yes it is
Okay, I have two people saying it’s possible, but I haven’t seen anything yet. Please generate the image and prove it to me, and I’ll agree. But just words won’t prove anything.
There are a TON of different cinematography styles man
wtf do you mean by CINEMATIC
i generated... the same pose using the prompt by Chatgpt
just use gpt 5 for once
you are being a pain in the ass
Because all cinematography styles are the same apparently
and cant accept to lose an argument
THIS is cinematic.
The color grading, the fog, the depth, the composition, the color palette
the banana is imminent
it looks like the usual youtube ad you'll see 24/7
You want an HDR or DSLR image?
give me one image and i bet you my life if i cant imitate it.
its the easiest thing like ever
You can’t because he says we’re all wrong 🤷🏼♂️
im even spanish myself and still can write english well
That looks like The Matrix
so... the challenge begins?
the bad thing is when i speak it vocally, not writing
He just wants to make us feel bad
That’s what he’s trying to do
And claiming what ‘art’ is
Or ‘cinematography’ in this case
I doubt you can create an image with that cinematic quality using text-to-image
That quality isn’t the best, honestly
I’ve seen better
Go watch Jurassic Park
That is cinematic genius
Madara Uchiha the official brand ambassador of nano banan
Great, then it’s an even easier task for you. Go ahead, create it.
gpt 5 waiting patiently
I think @simple remnant is doing it
I, on the other hand, have better things to do than to inflate your already convoluted ego
Better things to do? You’ve been here for hours just talking BS
No, that’s you mate
Trying to act superior over everyone
I gladly suffer fools 😊
anything else??
i just found out that the noun 'simulacrums' works better for style switch in nano banana
Yupp ai?
This image is screaming AI from miles away. Well, if you don’t know what qualities a high-quality image needs to have, obviously you won’t have the vision for it.
im gonna do the joke...yupp
yeh bro i kneel to you, you are better than me and every human kind that ever lived
cant you accept defeat?
prob im having ideas now
or ask gpt 5 better nerdy nouns
since is the most advanced model, it could help more experiment
"Look at them side by side to understand better. Your image is just a standard AI image with no quality at all, while the real image has depth of colors, proper placement of colors and lighting, composition, and professional color grading
He just wants to be right. That’s it.
hey i got this from .... leonardo 😂
I can generate images with that level of cinematic quality, but in other models. My point is: Nano Banana does not produce truly high-quality cinematic images.
@crisp rain btw
Someone should just pat his head and call him a good boy
Leonardo is a better model to cinematic
LMFAOOO
@crisp rain you have watched the entire show right
Yessir
that model is outdated dude
yea that's what i think
I made this on my own on alight motion as my pfp, please rate
hell illustrious is even more a trend than leonardo
Yes, and it still delivers a more cinematic result than Nano Banana.
I love that 😊 Great job!
Thanks
still it looks like a YouTube ad
You did an excellent job at it
bro switched side
No I mean the eyeball LOL
I just tried it here.
Much more cinematic than Nano Banana.
That hideous yellow in the middle
ik but its funny u switched from conflicting to happy
A man has many sides 😉
i guess you have problem with light.
you like it when its barely visible
i guess because it hides the ai imperfections?
He’s trying to sell stuff
Now everything is clear.
You don’t know anything about cinema or filmmaking.
Makes sense.
So his clients don’t know it’s Ai
You’re really being a rude jerk
now the ego is too high man
This is why I hated art school
Jerks like him
Thinking they’re better than everyone else
I can’t stand those people
You’re also part of the machine, buddy! You’re no different
If you don’t know about a certain field, don’t try to act like an expert. Be a listener and learn.
Take your own advice and don’t be a jerk
almost everyone here in Latam is starting to sell out stuff with AI, which is becoming humiliating, honestly...
I know very well the field I’m talking about.
Now, if you start talking about otaku stuff, I’ll stay quiet because I don’t know anything about that world of anime 🙂
That guy is an expert, remember that
all you did is you focused all the light at the center point and blurred the background and make it dark...
anime's unrelated to this?
You act like I only like anime, which you are very and sadly mistaken
You don’t know what ukiyo-e is
Which you thought was anime
sounds like a stereotype
i always liked pvz's one
Man ask for job in marvel/Disney they are looking for jerks like you to keep doing flops and telling that it's viewers fault
Said MadaraUchiha with an anime avatar.
well
Ok? And your point is…?
and we barely meet
Come on Mr. Nikola Tesla sitting in a Tesla 😉
we're chatting right through the internet
I have a recent project posted that I did for Disney.
Thanks for mentioning it
Ew
That explains a lot your behavior is a perfect match for them xD
Obviously, I’m going to defend something I understand, of course.
Disney isn't even a trend anymore
I like the new Disney. They do hideous movies and it's the viewer fault 😂
LMFAOOO
He’s using CHATGPT TO TALK!
HA!
Fraud!
Charlatan!
Yes that's what they say when they lose millions
You’re only allowed to speak English here 😉
what
But I bet you understood.
Said the anime character.
maybe you shall relax a bit
release ur shoulders
damn this escilated quickly huh?
He’s using personal attacks to make himself seem like a bigger man than he really is 😉
from a banana to literal insults
Mental illness
Now everything makes sense.
I understand why you identify so much with the model.
Nano Banana LOL
joe rogan jokes
No pelien
did you just call us monkeys?
Now you’re gonna get reported 😉
He is japanese
Racist
now its turning to racism?
I’m not Japanese at all, I just happen to be an educated American with an education 😊
Reading works wonders
mentioned in ur bio
Open a book sometime
NAH... I meant if we read your bio he could've understand that u're from california
Hey can we be more respectful to each other and refocus the conversation back to AI/nano-banana?
Oh yeah haha
We can’t because we’re all stupid
That guy says we know nothing
And is belittling all of us
I believe in us
Yes, let’s talk about Nano Banana.
That anime guy over there knows a lot about it
lol
I'm talking about nano banana
No, that’s your interpretation.
Hey how can u spend all day doing nothing?
Simple: I have my own agency, I pay people, and I have free time.
no one still tries out steganography in nano banana
So.. have no talent... to do anything?
no purpose in life
Why is everton catching so much hate?
@agile oxide Make something cinematographic so he can insult you, too!
just talk to him 2 min... u'll get it
I previously did seems nice
so maybe today he is high...
So like I said let’s keep this focussed on AI. If you’ve got an issue with someone I’d recommend the mute button. If not we’ll start to step in
Funnily enough this has been about Ai
Because I work in the film industry, and they want to know more than me, and they didn’t like that I debated about the subject I work with and actually know what I’m talking about.
You’re no Stephen Spielberg that’s for sure
i think mods about to do job
Looked like it was getting pretty personal to me
And you’re not Madara Uchiha.
At least I actually exist.
can you send links to some of your works im really interested. genuinely
just post them here
u have to become his client.... then u can recieve some of his " creations"
I wouldn’t pay him a penny
I googled and he seems to be on LinkedIn, legit to be honest.
okay.... well am leaving now .... it's seems it will take my all day
Anyone can make a LinkedIn page
he is now upset and dosnt want to send links...
im sorry if i hurt your feeling, i really have nothing against you.
At the end of the day we’re going to continue to do what makes us happy 😊
I have a job so I just use Ai for fun
That’s it
here comes
Hello chat
hi
Tons of Veo 3 I see
Text to video generator
what am I missing?
You can’t generate stuff in here
I believe that’s in the video channel, not this one
What is it?
its like encoding a secret prompt inside the image
Can I have an example?
for example a pose reference pic, once you download it, use steganography and encode such prompt or guide, like 'example for height and pose reference. Face hidden by hand. Cinematic mood. Hand held waist.'
its like injecting
it also works as jailbreak too
Tbh this guy @late thorn is ass but he's not wrong
We’ve all moved on from that 😊
I’m finding that Seededit can do some pretty neat stuff, also
A lot of it is bad but sometimes it makes something great
Nano banana actually did a bad job and failed
I love the Seededit version more (the first one) because that’s what I prompted
he's a prick who wants to hate on nano banana and has an arbitrary definition of "truly cinematic"
but yeah the nano banana text-to-image doesn't get the lighting just right, the color grading is a little flat
it doesn't mean it's a bad model, it's probably better for other things
He is not wrong and knows what he is talking about.
Even if he is a prick.
I suppose so
but it doesn't mean you have to be rude to other people
shouldn't true professionals be nice to others? but anyway, ai stuff
inconsistent sadly...
and also i asked nano banana to make a sunny scene rainy, it just spit out the same image
How can I use Nano banana
Now using Ideogram in workflow
create a 1/7 scale commercialized figure of the character in the illustration,
realistic resin statue style with highly detailed sculpted textures,
place the figure on a computer desk using a circular transparent acrylic base without any text,
on the computer screen display the ZBrush modeling process of the figure,
next to the screen place a BANDAI-style toy packaging box printed with the original artwork,
studio lighting, ultra detailed, photorealistic product photography
I’m not getting resin style from that
It looks far too plasticy
It looks like digital artwork, but not a statue
I would’ve gone with PVC style
The one with the Anti-ragging theme was a real handmade poster, second & third images I created with Nano Banana
It works very nice with reference images
The only problem is nano banana suffers from spelling errors
Yesss true but for a general broad reference it nails it, the different testing which I was doing on LM arena, everytime Nano banana was winning
For me nano banana wins at some things but fails in others
I’m surprised when I vote against it and find out the model was something I didn’t expect
Like Seededit
I’ve been dismissing that one a lot
And DALLE 3
But they manage to surprise me sometimes
Chatgpt & flux kontext was also doing great job but it depending on tasks, chatgpt was very finished with images with texts , but it literally changes styles, over-all reference etc,
I guess if flux kontext upgrade version will come it'll be something next level
That poster example I sent, it's a good starting point especially for students who wanna make a poster for competition etc, they can get a visual representation idea with that
I think every model has a different use case
There’s no one size fits all model
You have to use different ones for different things
One more
In terms of image editing from what i've been trying it's so far ahead of all the other ones. It does fail sometimes but i've been voting and I might have chosen it 90% of the time
I expect an elo of 1350 maybe on image edit
i'm sorry if it's a stupid question, but is Nano-Banana open source? Like, can I use it locally? Because I can only use it via lmaerana.ai.
It is not even released yet, and since it is Google's model, it is unlikely to be open weight ):
There's a few things I don't understand. Did Google actually say this is their model? Because I never saw that. And if they did, I’ve never seen Nano-Banana on Gemini or AI Studio. Also, when people post about 'Nano-Banana with workflow,' is that coming from a ComfyUI workflow, or are they talking about something completely different?
Some models are created by asian companies. It makes sense that they do manage it better since their native language is not really English but they were trained with asian language
Adidas sport suit ✅ 😂
If you ask Nano-banana to tell who created it, it'd tell that it is Google. A Google engineer also recently twitted "🍌" just in case
It's been said all but explicitly that it is Google's model
It's an unreleased Google model so no. You can use Qwen edit as is the closest
i understand. thank you for clarifying.
I thought he wrote by mistake on "x" instead of Tinder 😂
it's also crazy for colorization even though it's not a model specialized in that, beats all the other ones every time
#1397655624103493813 use lmarena to generate image from image in lmarena disord
shame the resolution is lowered a bit
meanwhile the human colored versions were like this which hm lacks a lot of color
meanwhile with qwen and flux it doesnt look real. Qwen way overshot saturation with the soldier one, and flux is decent but it lacks saturation cause trucks really were this blue back then
vẽ ảnh 1 con gà
IMO Flux did much better job here vs Nano. Speaking as Eastern European, it's rare to find any bright colors outside.
that's in usa, the truck should be more blue and uniforms had these colors
i feel like in human colorizations too it's like they are always undersaturated, like they think the past looked more dull
add a baseball hat
a girl japan
cat
gere a imagem deste homem segurando um energetico rosa, com nome "Ve se vira homem"
you can use it on the website (lmarena.ai) in battle mode, not here
thank you
elaborate why it didnt add veganism
or else ban (joke)
MAKE PHOTO
You can make one via the "/image" command.
/image make a futuristic of people dead by tomatoes
ok
Please read #1397655624103493813 , or use the website.
convolutional neural network
Make a girl
This is not we use the bot
You have to go on https://lmarena.ai/?chat-modality=image
Send your prompt and vote for the best result
Then you will see the name of the two models
If you're lucky, you'll have Nano
They were trying to use the bot, but that has been disabled for whatever reason. It seems to only be available in #video-arena-1 .
True.
uncanny valley thing
The AI that does this is Flux Kontext
Hey @gentle jasper
can you tell me how to generate character
in the same art style? or animation style?
Sure.
yup it would be a looooooooooottt help!
Yeah, I got you.
trying this for hours..... hope you can do 💚
I know that nano-banana is very good at preserving a specific art style even while manipulating an image.
Not really
Well, sometimes it doesn't, but those are rare occasions. Most of the time, it will keep the same art style for the associated character.
Minecraft style
It literally made my fav HomuMado look like a-1 slop >:c
Sure, it may not be a perfect image editing model, but it still tries its best.
Friends should add a video editing option to the page.
Well, to edit videos like changing clothes, improving quality and changing style
Yeah.
They should also put all the video models on lmarena
Yes, bro, we need the Google Veo 3 video generator in lmarena.
True.