#nano-banana
1 messages · Page 7 of 1
So it's a scam website ? Or what
faaakiiing im mister faaakiiing im nobodyyy turururuuuuu oooooh"
Bruh I don't understand what's going on
So why bring it up to us?
Why claim you ‘have access’ and not share?
Why claim ‘only 34 people’ have access?
Let's move on from this conversation please @crisp rain @chilly dragon
thinking of posting it on YouTube
I'd recommend muting the other person.
I’m thinking of posting it on YouTube 😉
Anywho
That was fun for all of 10 seconds
So what's the general situation with nano banana now ? Still only LMArena and Yupp have access to it ?
Yes
Have any Google's officials teased or announced release date for it ?
Bro, in the form of a bot in an application, and the number of its users is only 34,
That one guy said ‘3’
pineapple lets be real this kind of persons are more into ( suspicios behavior ) in the future who tell me if i ask to him something that he say got access , and then say thinking of posting it on YouTube and he send me a link and is a malware or something will be correct? just asking
The Logan Patrick something guy ?
Yeah
A few days after the launch of the Google Pixel
I hope it comes in Google AI Studio Lul
It will but let’s be honest, it’ll be censored, guardrailed and have corporate handholding and oversight
They’ll tell you what you can and can’t edit
Yeah it'll be a while until open source community can catch up
I like how Yupp is giving me more credits per scratch off for feedback than I'm using 😄
It’ll be a very long time I think
It will be released on the new Google Pixel phone first in Google Photos and then later on to the rest of the phones, and there is a possibility of it appearing on Flow. I do not have information about Google Studio, but I hope there will be.
Not for long I think, did you just open a Yupp account?
They do that for few first tries
I wasted time writing long feedbacks and they keep giving me nothing scratches
Exactly
The same happens to me
tell me the promt? i sub
Yeah, just trying it now for the first time. It gave me 5000 the first one and now they are all 200-300 but the generations are only 150
I feel betrayed so for the time being I'm just sticking to LMArena
Me too
LMArena is truly unlimited
You can do Battle mode with no end
love to LMArena
When I request a mod that requires a "fight" scene on the Yupp site, it cancels the process and replaces it with another model.
i made this for a friend
Because a lot of corporate models do that
You want fully uncensored you use open source
Plain and simple
That’s why I’m able to gen locally
I won’t be told what I can/can’t make
You don’t tell people what they can draw or paint or whatever
Or what kinds of fanfics they can write
hlo everyone
I'm from third world country and I don't have strong PC, only a weak laptop that can't run any model locally
I’m very sorry 😞
so my only choice is those online service AI models
I have a computer, but I have no idea how to do this.
Have you used Civitai?
You can gen stuff on there
Yeah but the credit (buzz) is not generous
25 daily credits, but one generate takes 15 Lol
I wouldn’t know since I just download models and LoRA from there
you want uncensored i use Awesome-LLM , need at least 500gb for ia models for work i recommend an a 2070 or 3080 gtx , 32 gb min , rec 64, and a i7 or i9
its open source
*i was rhinking to put the tutorial on youtube *
Come on! I won’t know about it if I don’t sub to your channel and smash that like button
I said that, and if you have eyes, I said "think" but I didn't, it's for my own good and my ability to use it, and anyway, didn't we get over this a few moments ago?
So why are you bragging about having something that you know others want to use?
I don’t see a point in that other than to be a troll and make people mad
Did you think you’d accomplish something here?
Dah, human nature, it's fun to show off something others don't have.
becouse in the eyes of people sound entirely different , you have access but you think like hey i got this and you not blah.. you know the point? is like giving a light to the dommed and turn it off and laugh
Alright, clearly a troll.
Time to block.
This guy will fade into obscurity when it launches for the rest of the world shortly
Not worth it
Ok so please move on
Why don’t you just remove them?
They’re clearly starting trouble
so that is the point about it, enjoy your human nature i stop talking and answer you.
Bro, your presence is like nothing, and it's not like I can't live without you
Is that imagen
/image nanobanana
here is the info you're looking for: #nano-banana message
How do I transfer style on nano
Lmao that's crazy
i need to generate an image using nano banana
Yeah if you read the pinned message it explains how.
its not sufficient to understand. and that link doest direct to nano banana
check out matt wolfe's video about it. that's how I found out about NanoB and LMArena. And was able to figure out how to gen images and videos. Very easy! 😄
who's that
search the name on YT 🙂 search bars were cool before AI 🙂
That link directs to where you access nano-banana
that link directs to lm arena but i can't specifically access nano banana.(very rarely it apperars)
Yes, because it only appears randomly
Yeah it's only in battle mode
yea
ok
can i DM you to ask some questions about Civit?
Nano banana is very good for image to image
Yeah so far for me I've noticed that FLUX Kontext and Nano Banana are both REALLY good!
Can someone help me replicate this art style to another image
what is the 2nd image?
This
Ok
Read #1397655624103493813 to learn how to use the bot
@wispy holly can you submit bug issues in our #1343291835845578853 channel? There are likely similar issues that you can add to instead of creating a new post.
How and where you got all that credit??
What's most impressive is that the model can do professional image editing in literally 10 seconds. Hopefully, when it's released to the general public, it won't be nerfed with things like: it's not possible to process the image because it contains people's faces.
1000? I got 5000 yesterday for signing up
@agile oxide is this an error , cause why it comes up everytime i generate an image ?
Just a security thing
happens to everybody
THATS WHY I GOT 50€ BILL
It wasnt the video generation
Holy cow
what did u do?
Nothing waiting
it just charged you randomly?
Nono I see the credit go up everyday
but today it got up by 48€ after some veo 3 generation and some images
Output from Nano Banana
Nano Banana (left) x Imagen 4 (right)
Hello! Can anyone help me take a measurement using the nano banana? I'm trying to make a simple clothing change and I'm struggling to get the ideal result.
I'm trying to modify the wedding dress in the first image with the dress in the second image.
I've tried several prompts, but I can't reproduce the reference dress in the second image.
Let me try something
I'll wait.
@steady flint
@topaz helm the neckline never looks the same. Note that in the reference dress it has a very pronounced curve
alright. thanks for trying to help friend
np
Veo3 is on ai studio?
Ah
i find it funny how picky nano is with prompts, sometimes you get the greatest image edits you've ever seen and other times you just get
could gemini 2.5 flash image generatioin preview be nano banana?
It is
and is coming very soon
Get him to drive a black Mercedes, the latest model, and a nice dress code.
NANO BANANA IS NOT ACCESSIBLE IN HERE!
You CANNOT generate images in here
It’s hopeless 😔 You guys will never understand because you’re not understanding the rules
sigh
hehe the cycle continues
Maybe that was the intention 🤷🏼♂️
Also
Anyone else see that?
I love when the Ai messes up. Sometimes the mistakes are hilarious.
For instance:
Original image
The mess up
But here’s the nano banana result I was hoping for
Check the pin for more information on where to use nano-banana: #nano-banana message
I’m finding more and more that a lot of models make really cool things that nano banana can’t
First image is Qwen Image Prompt Extend, second is Seedream 3
The paint styles are beautiful and accurate
what did nb fail at here?
I haven’t gotten the result yet but it usually doesn’t do as good a job
Especially with traditional Asian styles like that
dataset I suppose?
You’d think since it’s Google the dataset would be massive
but not exhaustive
Well..this is nano banana
It did amazing actually
But the other two also are really good
I like using all kinds of models, honestly, because they all have strengths and weaknesses
GPT1 is amazing at characters and pop culture and stuff
But it has that yellow look to it
Unless prompted otherwise
Nano banana is hands down the best image editing I’ve ever used
But it lacks in pop culture and changes faces a lot
Do you prefer this model over GPT-image? For image editing, I believe nano-banana is unbeatable at the moment. But when it comes to generating images, it falls a bit short. What do you think, since you've tested it a lot?
I really like GPT1 for the sheer variety of characters, shows, movies, games, etc. it knows
You can make all kinds of things with it
You can change the style but you have to be very specific
Nano banana excels at image editing
It’s also really good at just prompting
However
Yeah, it's definitely way better
For the eagles?
yeah
It’s very simple actually: A Joseon Dynasty painting of a Golden Eagle
Joseon Dynasty is Korean
I use my knowledge of history and culture to make things 😊
I’ve always been influenced by East Asian cultures since I was a kid
I did Shaolin Kung Fu and Ninjutsu for 16 years when I was young
And Bruce Lee has always been my idol
And I used to live in Tokyo in Kitashinjuku
very cool, I'm a fan
A fan of what 😮
the prompt & generations, just looks rly satisfying
Oh yeah
I love art like that
I’d buy it myself if it existed
I could try to get the design on a shirt
I need to figure out how to do that
I suppose I could do Redbubble
So I’m gonna say I think I prefer Imagen the most
Imagen and GPT1
Nano banana comes in third for me
These are Imagen
Oh wow
Gpt?
Yes?
Actually no!
I'm new here, I'm from Brazil
Pretty good image
Thank you 😊
You can do it by typing /image in one of the creation channels and on the LMarena website
I recommend you doing it on the website
But you can't choose nano-banana. It's random
Can you send me the link because the site I was on was just the battle arena of im
You can generate images on the Battle Mode, and then choose what model is better
Is there any arena for creating videos?
Yeah, here on discord
What is the command?
But how do I do it? I just ask in the chat, like, create a dog in the bathtub?
I'm in a chat, what now? What do I do?
Read this again
#1397655624103493813
Ok but I can't create ☹️
why not?
I write the Prompt and it doesn't generate what do I have to do?
Did you actually read #1397655624103493813 ? Because obviously you haven't as you aren't doing it properly. Read again
I got it, thank you very much to everyone
Can I create images like this?
In battle or direct chat, yes
These are all different models I’m using, though
So I can only use it in battle? But if I can only use it in battle, are there other types of models?
Yes
And you can select models in direct chat to use those models specifically
Can I here in the chat? How?
Bruh, I can't believe nano-banana lost to GPT...
Happens quite a bit actually
At least with the stuff I make
Your mileage may vary
What commands can I use to create images? Here in the chat
Rare nano-banana L
?
The only part I don’t like about GPT-image is its love for the color yellow and its imprecise image editing, aside from that, it performs better than Nano Banana.
I was talking about nano-banana losing to GPT, I wasn't replying to you
Kkkkkkk
OK
Looks like a project I did 3 years ago, with MidJourney.
https://www.reddit.com/r/midjourney/comments/z7fzn1/what_happens_when_you_mix_ancient_egypt_with
Is Nano-Banana a diffusion model like MidJourney?
Those are reall good!
I've always loved Cyberpunk stuff so I make a lot of things with it
Qual era o nome do site mesmo?
Yupp ai
Blz
?
Back then, everything was a bit archaic, and video models weren't available. I'm thinking of recreating this project today with much better images and then play in videos
nano-banana is likely gemini 2.5 flash native image gen, so it is probably autoregressive
i mean, google accidentally charged people for this model so...
What does that have to do with nano-banana?
Manda aí na DM que aqui é só inglês
How? Do you not see all of the stuff I make? Far from it '-_-
GNOMES ON CAPYBARAS
niiiice
Which model was yours
its a tossup for me, sometimes nano banana handles a prompt better than imagen, sometimes worse
its been like that since it first came out
Actually
Every model I've used has been giving me great things
Which is why I love Battle
You never know what you're gonna get
Believe it or not, sometimes even Seededit delivers something better than Nano Banana.
Flux, Qwen and Gemini have given me awesome things, it just depends if the prompt aligns with the strength of the model
Yep, this
I've shared stuff earlier that were Seededit that I loved
@late thorn 😂
nano-banana is a stealth model and all evidence suggests it's by Google, the art style, some responses the official Google AI twitter account including a banana emoji in their tweet recently, everything
We all know this already
and it's likely 2.5 flash because it's an editing model like their last one and the timing makes sense
nano-banana has to be 2.5 flash native image gen
The composition 🤌
and google accidentally leaked the name anyways
Thank you 🙂 That one is nano banana
I only used one image of Zhou Yu from Dynasty Warriors
Where?
The supposed 'Gempix' or whatever?
it can't be an on device model because of how good it is and its world knowledge, but it can't be a pro sized model because of its lack of world knowledge compared to 4o
this
people were actually charged for an unreleased product
and google had to refund them
Do you use prompts in JSON or in plain text? Because your results are great.
Plain text 🙂
I just type my things out
Things that I have in my head
Cool
Years of things I've wanted to make but couldn't before
Now it's all possible 😄
The whole reason why I love ai stuff is because it brings my imagination to life
@late thorn This is Flux failing btw
which one, there are 3 flux kontext models
Dev
I hate this Kontext texture/grain 🤢
Wrong, it's going to be part of Google Pixel's stuff
They want it on devices
well that one's the worst one (open weight, 12B parameters)
They've stated this
are you saying that nano-banana is on-device?
That's what's been indicated
The Google Pixel photo editing features
The new Pixel phones
But what Google will do is have censorship, guardrails, and corporate handholding and oversight
it's not on-device, most on-device models are 4b max and usually quantized to 4bit, image gen in this quality makes this very hard to believe + look at your image above
You see, the Gemini site itself doesn't let you make images of copyright characters
The stuff that we get to test out lets us
Idk why it's so hard to believe, Google wants to make it available to the masses
I doubt a 500m image encoder like what Google uses for their small models are anywhere near capable of this, probably can't even get to SD1.5 level let alone gpt-image-1/Flux kontext max
the only way they can do that is by firing up a lot of TPUs and run the model on there
Google is gonna do it one way or another
If they've managed to add it to their photo editing on the new Pixel phones
September 3rd also seems to be the day we'll find out
According to that Google employee on Twitter/X
That is a few days after the launch of the new Pixels
most devices people run are 6-8gb ram, flagships like pixel 10 are 12-16gb ram, plus device has to load and prep model which takes like 15 secs and doing autoregressive gen like that takes a bit, faster is you use MoE like what Google is likely doing plus TPUs
People want to be able to edit on the fly, hence cell phones
There is no indication of that
None whatsoever
Just because people got mistakenly charged for something doesn't mean squat
No one here is a Google employee either
So it's all hearsay
All speculation until we know for sure
well, that's I think nano-banana is, we'll just have to wait and see
Yep, and that's what we're all doing
Playing the waiting game
On the plus side
My 37th birthday is next month
Nano-banana will be a good present 🙂
I also have far too many games to buy the rest of the year 🙁
<---- Huge gamer
I have a PS5, Xbox Series X, Switch 2, Switch OLED
Left is nano banana, right is lucid origin
Did some tests here.
The only problem is now it's not Zhou Yu from Dynasty Warriors 🙁
Solved haha
Bananza is my GOTY
I own it and haven't played it yet 🙁
I have hundreds of games 🙁
I buy way too many
Plus GamePass
I have an addiction
That and energy drinks
I know your pain, haha. I also have all the consoles here and I don't even play 5% of all of them.
I have a Meta Quest 2 and a Quest 3 and I've only played a few times this year. The problem with having everything is that you end up with too many options and don't take advantage of them.
When I was a kid and was poor, I only had a GameCube with four games. I played those same four games for years and found it really fun.
It's funny how things change.
Yep, that's definitely my issue as well
In all honesty, I play with ai every day
I make stuff all the time
That takes up my gaming time
And time I could spend reading my books on Kindle
Creative minds are thirsty to create, haha. I also can't go two days without creating something.
Do you work in a creative field? Because I always thought that when I worked with that, I'd lose my hobby of creating things. But luckily, that didn't happen, and I can work and still use it as a hobby.
Sadly, no 🙁
I am a laborer
I work for The Man haha
I'm part of the corporate machine
I should've done better in college
I played around way too much
I'm too social
I didn't focus enough on my schooling so...
....My job sucks 🙁
You've never tried? It's a waste of a creative mind.
Like most Americans
Nope 🙁
I also think I have some form of ADHD
I have problems focusing unless I really sit down, but ai helps me focus
Let's keep conversation focussed on nano-banana please.
I think it's fine if we're not hurting anyone? At least we're getting along in here 🙁
There are so many cool areas you could explore:
3D, VFX, Matte painting, Compositing, Storyboard, Retouching, Photography.
And use AI, ofc
I think nano banana is going to change the game dramatically
It's definitely gonna give the middle finger to Adobe LOL
It already has and it's not even out yet
I remember when I had to get Photoshop by less than savory means when I was younger
Adobe is a joke. They're just stuffing their package with APIs, like Kontext, because they can't create anything on their own.
Photoshop is a great software. The problem is Adobe 😩
The problem is also Google, sadly
All that money and greed
They control what we can/can't make
If they let people have creative control over what they're making and let us be responsible for what we make things would be better off
Which is why GPT sucks as well
I meant OpenAi
They're far from 'open'
Btw, this is one of my things I made locally 🙂
This is concerning, because we have the SOTA (state-of-the-art) of AI images and videos in Google's hands, and I think it's difficult for another company to compete at the same level. Let alone an open-source model (my dream).
Imagine a Nano Banana open-source model that you can train LoRA on and create a workflow with it in ComfyUI?
The open source people will work on it, but it'll be light years behind Google sadly
Which model?
Google has the compute and resources
One of the Illustrious or NoobAi ones 🙂
And that's my Madara Uchiha LoRA
kohya?
< (Imagen 4) . (Nano Banana) >
Imagen 4 and Nano Banana are pretty equivalent, both are great. It's funny because for commercial photography, Imagen 3 does a better job.
That's a really cool concepts.
Which models?
I'll try to do the first one as if it were realistic, with Lady Gaga herself.
I'm glad you knew who it was 🙂
That's what's important
Nano Banana keeps messing up on Yupp
But I managed to do this
I'm having a hard time getting the Nano Banana model in LMArena, I must be really unlucky.
Niice
Yeah I'm not really getting nano banana either
LMFAO!!!!
When a model messes up, it messes up badly
That’s why LMArena is so fun, because there are some horrible models that create these hilarious images haha
I agree entirely
I've laughed at some bad images
Ok, maybe this model is way better than I imagined. The edits it makes are insane. Congrats, Google.
In UHD 🔥
So Nano Banano is great to create design? Good to know
Insane
I feel like I'm gambling in Las Vegas every time I roll haha
Except I'm not losing money here
But not in one shot lmfao.
Holly
That's Qwen Image Prompt Extend
90% of the time Qwen sucks.
That's impressive
That's why I was so shocked because I thought for sure it was Imagen
Even Photon did a great job
The consistency in the people who are farther away is something that impresses me too
Imagen won here haha
I'm just prompting now, not image editing
It’s a shame he didn’t know how to make Lady Gaga’s face in play.
Right?
I think if more people embraced ai we'd see some truly wild things
Most people don't care 🙁
I've been working in the arts for many years, so I know a lot of people in the field, and I can say that 90% of them hate and attack AI whenever they can. It seems to be the fear of becoming obsolete because of it. I did things differently; I included it in my workflow and dedicated myself to understanding everything about it. But people in the creative field still criticize it.
Oh I know
I've seen all the anti-ai mess on Twitter/X
I stay away from it
Those people are set in their ways
I still hope that there will be peace and that everyone can embrace the technology.
Bacon time!
when can we use nano banana api?
Whether they like it or not ai is here to stay and is going to become more integrated in our lives. The same thing happened with Photoshop. People hated that too.
No one knows.
Yeah, that's true
Video models are so fun
WHOA!
What model is that?!
That was honestly amazing
Sadly I don't think I have enough VRAM 🙁
I need to get a 3090
I hear those are one of the best and cheaper
I have a 4080 RTX
U can rent some powerful GPUs on Runpod and run Comfy there.
Whoa.
Which model?
Wow
Flux 1 Kontext Pro
text to image?
Kontext Pro did a great job.
Looks almost like a real painting
Yep 🙂
I'm just prompting stuff
I haven't uploaded an image to edit since Lu Bu
All these are just ideas in my head
you can use it for free on google colab?
Left is Flux 1.1 Pro, right is Imagen 4.0 Generate
Neat 👌
Kind of, yes. But you would have to configure all the checkpoints and download them every time you wanted to use it. It would be a complete mess.
: (
Ideogram is very good to painting and design
Nano banana is left, Imagen 3.0 is right
Imagen 3 did a better job, IMO
hard test on banana
Yeah I think so too
damn, That's so accurate
true
not good enough, his body looks weird still
@crisp rain
I know you prefer to generate paintings, but if you ever want to explore realistic styles, I made this LoRA a while back and it still works great.
https://civitai.com/models/672963/phlux-photorealism-with-style-incredible-texture-and-lighting
that's true, i meant the face
yeah, its very good
I'll have to try that sometime when I redownload Flux
I deleted it due to lack of space 🙁
Yes, running locally takes up a lot of space 🙁
I had to get a 5TB external hard drive just to store all the models.
I have...almost 10 external hard drives
All 2 TB a piece
All full
All models and LoRA I downloaded from Civitai
They delete a lot of stuff on there, especially now
Bro, the bacon knife 😂
Nano Banana will let us make that a consistent character!
Do me a favor
Retype that in ChatGPT in your language and share the translation
Because none of that made sense to me 🙁
Photon
I absolutely LOVE that since I'm a huge fighting game fan and afficionado
Been playing them since I was a kid
me too! KOF is amazing
I've owned almost every one of those 🙂
I have KOF XIV and KOF XV and KOF XIII
The others were on my PS2 years ago
ps2 is the goat of consoles
#nano-banana generate image with two boys laughing in the class
i know lol, which prompt
I've written down thousands of prompts over the years
Snoop Dogg in Danganronpa, essentially
so nano banana can do celebrities faces?
didnt know
it knows a lot, or just the famous ones?
That's not nano banana lol
That's Imagen
Nano banana isn't good at celebs
You'd have to upload an image to edit for that
Nano banana also fails at a lot of anime styles and characters
GPT1 excels at those
So does Imagen
Qwen Image Prompt Extend
really?
nano banana’s character preserve is awesome
Yes unless you upload an image of them
If you just prompt normally, no
But for image editing it’s great
And keeps characters consistent
hmm I always drop images and do iterations lol
That’s why
If you just prompt yourself, the characters come out wrong
may be it’s a model designed for image editing
Most likely
It’s good at straight prompting as well, but lacks in a lot of ways
It’s strengths definitely lie in editing
might be a distilled model
it’s incredibly fast
too small to hold so much pretrained data
Not really
That's the only model that gets a ton of characters right away
Like Peter Griffin above in the Family Guy style
i2i i means
as for me I usually do i2i
Outpaint/extend the current illustration downward so the canvas shows the character from the top of her head to her knees. Keep the same anime/semi-realistic style and color palette. Preserve all existing details: short silver-white hair with a thin blue streak, cat ears, blue ribbon and white plumeria hair clip on the right side, blue eyes, small bandage on the nose, white short-sleeve blouse with scalloped collar, navy bow at the chest, and both hands forming a heart in front of the chest.
For the new lower half, add a navy-and-black uniform pleated skirt and white semi-sheer stockings. Use a top-down (overhead) camera angle; she slightly tilts her head up and looks up at the camera. Her figure is tall with slender legs. Keep the softly blurred warm bokeh background and soft natural lighting.
gpt-image-1
just look at the hair, it obviously became lighter
Akaarii?
Akarii?
That's a Macintosh Plus
Ah
You're reading the katakana?
The text.
I used nano banana to change my OC outfit, and the consistency turned out amazing!
Here’s the prompt I used:
A high-quality 2D anime-style character illustration. Keep the character’s original face, hairstyle, proportions, color palette, and overall art style exactly the same. Only change the outfit and accessories according to the following description: <outfit description>. The character should still look consistent with the original artwork, as if drawn by the same illustrator. Preserve line quality, shading, coloring, and rendering style. Background: pure white.
I wonder if it’ll be open-sourced someday?
Romantic moments between the couple and kisses in the bedroom
Romantic moments between the couple and kisses in the bedroom
does any one prompt for product
where i can use nanobanana
in battle mode its totally random
is there a reason you use AI to generate yourself pictures of underage anime girls?
I think imagen 4 ultra handled this prompt better but I kinda like how the nano banana one looks
"this prompt"
meanwhile doesnt give any prompt 🤨
@sage locust Don’t take the bait, they’re just trying to start something
Ma Chao the Splendid has arrived!
What about showing us the 3rd, he was cool. Or oruchi
I have Orochimaru stuff because he’s also one of my most favorite characters of all time 😊
This was also my first anime when I was like 10. Big fan.
Hb kakashi
you may want to check #1397655624103493813 to see detailed instructions on how to create content in the server
Very highly unlikely ):
how do i land on Nano-Banana ? is there a way of getting image generated only by nano ?
no, you just gotta reroll and be patient
More information can be found here - #nano-banana message
Please describe why eggs explode in microwave ovens. Use 4 panels.
Some spelling mistakes but overall very nice
which model is this? seems very good
nano banana?
yes
I also tried imagen, but interestingly, ultra didn't work as well.
Once it does perfect writing it will be insane
how can I try this model "Nano-banana"?
The pinned message should help: #nano-banana message
thanks
guys, what do you think is best to create consistent characters for games ? A model like Nano Banana while keeping the base images of the character or a LoRA ?
side
just tried nano-banana for an illustration and I'm quite impressed!
nano banana tomorrow?
forget nano banana when are we getting macro banana?
Most likely September 3rd
You're looking for Video Arena, read the #1397655624103493813 for information on how to use it.
have you seen this:
https://x.com/patloeber/status/1959914630366494985?s=46
But what do those have to do with tomorrow?
this week
Which part of either of those Tweets indicate that it’s coming tomorrow?
I mean if it comes this week I’ll be pleasantly surprised
The new Pixel phones launch this week
@crisp rain ✋
What’s up? 😊
I can hardly wait, but I'm afraid that EU regulations will mean I can only use the new model with a VPN again... :(
why do they say "this week" tho, sept 3 is next week no?
u think its gemini 3?
And nano banana
ohh
im praying rn for it to actually release
where can i use nana-banana
I’m very curious: did you not read the pinned message?
It clearly says how and where to use it
apologies
Not sure if I should put it here but oh man, nano-banana seems to be the only model (as far as I can tell) that can perfectly handle reference images and does a great job
yEAH, ITS THE BEST
How to use it ?
On LMArena or Yupp.ai
On LMArena you can only get nano banana in Battle
nano banana
Check the pinned message for more info: #nano-banana message
When I upload pics to edit in battle it's getting errors like not made for pic editing
See like this
i have the same error
Looking into ^
Can you share with me what image you're trying to upload? I'm unable to reproduce
I think check whether it’s a Live Photo
I don't think that'll cause problems, I was able to upload a live photo just fine.
Original on the left, nano banana on the right
I do understand why nano-banana doesn't change the style of images now.
It's not because it's a bad model. It's because it was actually trained intentionally to always stick to the style of the reference image. On some occasions, it will change the art style just a little bit, but it won't change it to where it's a completely new image. It'll still maintain the same character, same style, the same art style.
And hence, it's very good at editing the reference image to be exactly the same but have one minor thing changed and to be in the exact same style too.
Yeah, in my image it kept the original background and only changed the character.
Yeah, exactly.
I asked Copilot about it, and it gave me that answer.
Apparently, it's because the image generator actually first goes through a thing called the multi-layering system where it takes apart the image (background, characters, art style, etc.) - even down to the colors. And it's able to understand what it needs to change about the image just from that. I believe.
bruhh
flux is very censored
yes
how do i make images in this ratio ? my output is always in 9:16 , can u help ?
It uses the same ratio as input image
#help
What are you looking for?
How to create images please?
If you're looking for nano-banana you can find more information here: #nano-banana message
Thank you
Nano Banana with a lot of editing
I've gotten better results transforming sketches into caricatures with GPT-image-1. Nano-banana is too sensitive to the prompt and relies heavily on the quality of the sketch, more than the reference style image.
Nano banana is especially good for editing and merging images.
When using multiple images, it is offered at least half the time. If you want to edit a single image, you can send it twice and you will have a good chance of getting it.
Also, its generations are sometimes superb, often completely botched like an old crude cutout, you have to learn to recognize its style and not forget that the retry is unlimited.
If you have any doubts about the model, download the image, it is the only model whose outputs are around 2mb.
So it’s gonna be on the Gemini app or…?
That means it’s gonna be hella censored, guardrailed and corporately handheld 😌
You can’t even make copyright characters on the actual Gemini site 🙃😬😑
So I wonder how it’s gonna work out
Posted by Demis too
I hope it’s just as good
The camera slowly zooms out.
Do not write in the video area
ctrea video
amazing detail nano-banana 🍌
Hey @worldly summit Please read #how-to to learn how to use the bot. Thanks.
So, how soon do you think Elon or China or OpenAI or whoever releases a copycat for nano banana?
What a mess
Qwen distilled nano banana two month after release? It took a while for them to distill seedream for Qwen image but they seem to be doing more eimage stuff
Hi guys ! I’m new here
Just heures about nanobanana it’s impressive
Heared*
How can I use it constantly ?
I don’t see it in the ”direct chat”
Hey, please check the How to channel https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to generate a prompt, and go to the video Arena 1
Hello, since I am new here, I just used LMArena to edit images. I got incredible results with nano banana, but today I don't see any results generated by nano banana, is it gone?
Hey, please check the How to channel https://discord.com/channels/1340554757349179412/1397655624103493813 to learn how to generate a prompt, and go to the video Arena 1
Nope, it's still there.
Nano-banana is only available in Battle mode. More info can be found in the pinned message #nano-banana message
wrong channel, read #1397655624103493813 and use #video-arena-1
use #video-arena-1 or the other two video arenas please
it won't work in this channel
ok
this channel is for discussion about nano-banana, not for generating images with it
read the pinned messages
Got it. Thank you.
It's launching today
Really ?
You sure?
Who said? Where’s the proof? With no proof it’s just speculation.
multiple tweets by google with banana picture.
even by Demis : https://x.com/demishassabis/status/1960105069690655116
That’s interesting but that doesn’t say anything about launching today
google dont hype it too much in advance... launch is inevitable.
I think it also really needs to be said that we should temper our expectations regarding nano banana because: It’s gonna have censorship, guardrails and corporate handholding because it’s from a big company as opposed to being open source and open weights. Don’t be fooled into thinking Google will let you create whatever you want 🥴 They control the narrative.
11
not everybody is waiting for this model to generate nude stuff, tho
Who said anything about that?
I’m talking about arbitrary stuff
Like characters from different fandoms and stuff
I don’t care about generating nude stuff with it, I care about being able to make my Dynasty Warriors and other game characters stuff with it
Way to misinterpret something 🙄
Qwen Image is nowhere near Nano Banana
wouldnt expect google to release this as open weights anyway
No one said they were
We couldn’t handle it locally any way
It’d be one of those ginormous models
You’d need a personal data center to run it
I mean they just released Grok 2 locally
Way too big for consumers
we dont know the size of it and why they call it "nano"
there's some rumors that it could run native on pixel 10
My biggest guess is it’s gonna be on the Pixel phones
It coincides with the launch of them so that makes sense
Meaning I might have to get a Pixel…if it still works how it’s been working for us on LMArena
If not I won’t waste money on it
But if it continues to let me make my fandom and pop culture stuff, then I’ll definitely buy it
I’m just waiting for the bait and switch like OpenAi pulled
Letting us make fun things when GPT1 launched and then yanking it all away
Hi, I have a question about what is the most efficient promp to mix a person with caracteristics of a reference in the same picture like a costume
no
/imagine ddddk
i think its really true, because v2.0 is really same type of manipulating
nano banana x hailuo 2
https://x.com/D_studioproject/status/1960313157953675693
What if one image was all you needed?
I tested Nano Banana to generate a full cinematic sequence from a single still.
No reshoots. No manual matching.
Just pure consistency across shots — tone, light, mood.
It’s the kind of tool that changes how we tell stories.
Workflow? Easily 10x faster.
Powered by
@hailuoai_official x @wavespeedai
@...
Pixel phones edit live and beard thru grape vine September 3rd release
Did they remove nano banana from lmarena ?
nano banana is rolling out to gemini
nah
Yeah I just made this
Google’s most powerful drawing model, Nano Banana, has just been released — you can call it through the Vertex API.
Model ID: gemini-2.5-flash-image-preview.
#nano-banana Place this almond pakage on a supermarket shelf with the rest of the products
It's available on the Gemini app already!
No more need to fish for the banana model on lmarena
I'm using it on the 2.5 flash model
wait really
Check our pinned message for more info: #nano-banana message
yeah
true
I think so, doesn't seem to get imagen anymore
but thats boring trying many times to just get nano banana people are excited to try it but.....
Lemme see
It doesnt seem to work for me
is it out on LMarena
well it works for me
no need to use lmarena
oh
Simply go on gemini.app
Generate a 360 photo shoot of a model wearing this sweat in a white box
wtf
thats cool
wow that's unreal!
Easy marketing
It's on the app too on 2.5 pro and 2.5 flash
And free
Replace the text with “Satik”. Keep the exact same style, typography, colors, and design. Only change the letters.
Lmarena also has it now
yeah we can probably even add celebrities
"Generate a 360 photo shoot of a model wearing this sweat in a white box, pinterest type, streetwear, no face, blurred face"
ayo
lemme try now
Lmao what a difference
does it work on browser doe
what the sigma
Replace the text with “Satik”. Keep the exact same style, typography, colors, and design. Only change the letters.
Cooked the competition in editing, very deserving
so good bro????
yours more better
fr
2 and a half million votes 😭
uh so its probably problem with my name
cuz yk how ai works
it would have more images of satik in its database
nah its okay
Replace the text with “Satik”. Keep the exact same style, typography, colors, and design. Only change the letters.
bro thats so cool
lemme try...
the new name is so annoying bro
Holy jesus!
I find it funny that it actually exists.
So fire
Also good to knows!
The model in Lmarena is slightly less censored than in the Gemini app
fr
Comp.
Don't ask
lmarena got its own word filters
They're better ngl
Better as in, you have more freedom in lmarena
yeah
🍌
i cant seem to get it to work in gemini

Literally 1984 on the Gemini app

bro wtf was that
Whoopsie
Want the prompt or just confused?
Freaky ahh 💔 🥀
i only saw a bit
i just opened discord tab and saw that then u delted it
Barack Obama and Donald kissing.
Was to show the kind of censoring
Can't generate real people my ahh
so my banana is gemini 2.5 flash preview