nano-banana

[hard sun]

Is it possible now to have direct chat with nano banana

?

[hard sun]

Is it the same as the old one?

Cuz I see now there's two of them

"nano-banana" and that one

[coral willow]

nano banana = nano banana ✅

[hard sun]

Thanks

@lusty jungle why do you disagree?

[lusty jungle]

lmao nooo that was a joke

[crisp rain]

For anyone that doesn’t get the reference, the character that I made the image of is Lan Di from Shenmue

[turbid willow]

Original vs nano

[ornate ledge]

can anyone verify the website nanobananaimg is the one and the same?

[proper atlas]

https://fxtwitter.com/techguyver/status/1960464912758493410?s=46

Techguyver (@techguyver)

Nano Banana + Runway Act 2 = full creative flexibility
︀︀
︀︀Image generation might feel “solved” - but Nano Banana shows we’re only just scratching the surface.
︀︀It’s insanely fast (sub-5s per render), cheap, and gives you realism + flexibility to turn any image into *exactly what you want*.
︀︀
︀︀Here’s how I tested it 👇

**💬 65 🔁 381 ❤️ 4.1K 👁️ 379.1K **

▶ Play video

FxTwitter · Aug 26, 2025, 10:10 PM

[light mauve]

@agile oxide when I use the same prompt and image in AI Studio, it generate text instead. It might explain why it took forever to load

[turbid willow]

Colapse servers

[formal vector]

Nano banana appesr again in result

[boreal roost]

Extract the product from the original image and place it on a solid, uniform pure white background with no gradients, shadows, or distractions. Maintain strict product consistency with special emphasis on text consistency—this is critical: preserve all text on the product (including logos, labels, instructions, or any printed elements) in their original font, size, spacing, color, and clarity. Ensure no text is altered, blurred, cropped, or misaligned; every character must remain identical to the original.

Refine the product to commercial-grade quality: enhance overall brightness to make the product stand out against the white background (without overexposing text or details); strengthen three-dimensional volume (emphasize natural contour shadows and highlight edges to avoid a flat appearance); optimize light and shadow effects (add subtle, realistic bottom shadows for a grounded look and adjust highlight positions to align with a natural light source, ensuring text remains readable); and boost surface texture (retain and refine fine details such as material grain, product seams, or texture patterns while keeping text crisp).

Ensure the overall composition proportion remains consistent with the original product’s scale—no stretching, shrinking, or skewing of the product’s form. Deliver the final image in 8K resolution with ultra-high definition (UHD) clarity, meeting professional photography standards: sharp edges, noise-free details, accurate color reproduction, and a photorealistic finish that matches commercial product photography quality, with text clarity as the top priority.

[ruby escarp]

Crea te a new image

Extract the person of the photo and puttit playing music on iziza.

Hundreds of people enyoying

Extract the person of the photo and puttit playing music on iziza.

Hundreds of people enyoying

[thorny skiff]

/A colossal ancient alien-style square bipyramid spacecraft, hovering about 5–6 meters above the vast desolate Earth wilderness. The structure is massive and awe-inspiring, with mysterious dark engravings and subtle alien glyph-like patterns covering its metallic stone-like surface. Small glowing lights flicker intermittently along its edges, hinting at advanced but forgotten technology. The atmosphere is surreal and otherworldly, under a dazzling night sky filled with countless brilliant stars, giving a sense of both ancient majesty and extraterrestrial mystery.

[gleaming wasp]

Use the nano-banana model to create a 1/7 scale commercialized figure of thecharacter in the illustration, in a realistic styie and environment.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 computer screen, place a BANDAl-style toy packaging box printedwith the original artwork.

[slate dagger]

book cover

[forest yacht]

go on lmarena.ai, go to direct chat mode and select image generation. you can select gemini 2.5 flash in the dropdown, but it should already be selected. you do your prompts there, not here as this is a channel discussing the model, not a chatroom with the model

idk why you guys keep on putting your prompts here, just go to https://lmarena.ai and do it there

[crisp rain]

You guys CANNOT make images in here!

This is not a channel for generating images

[hollow cipher]

any tips for haircut generation with source image?

[weak rain]

Crea te a new image
Extract the person of the photo and puttit playing music on iziza.

[humble verge]

Can anyone help me with prompt

Put the outfit in image 1 onto the character in image 2. Anime screencap artstyle. Simple white background.

Nano banana just outputs the same input image without changing anything ??? Wtf

[trail tendon]

is it possible to force the pictures to be 16:9? and if don´t is there any solution to that? any generative tools that could extend the generated pictures?

[humble verge]

yeah its so unproductive that we have no control over output aspect ratio, I thought that was an obvious feature

[tawny wagon]

adobe has one but Im looking for alternatives too

[stark goblet]

add banana in that

[mild grove]

which one is better between flux kontext and flux kontext pro?

[hallow quarry]

Hello

Is there an option to get high resolution images can be generated in nano banana? like 4k

[icy junco]

where is to be generating ?

[potent sky]

is the resolution of nano banana fixed in 1024x1024?

[nova kelp]

the watermarks nano banana generates

pretty interesting

[chilly dragon]

What is this?

[nova kelp]

the watermarks nano banana generates

[chilly dragon]

I don't see any watermark.

[nova kelp]

well, those shapes are the watermark

[wicked ruin]

#nano-banana Draw A Prospective Model Of The Character In The Picture, Commercialized As A 1/7 Scale Full Body Figure. Please Make This Image Into A Real-life Figure Photo. Place The Figurine Version Of The Photo I Provided On A Round Black Plastic Pedestal. I Would Like The Pvc Material To Be Clearly Visible. The Background Should Be A Computer Desk.

[loud plover]

Create a 16:9 image of Godzilla in a banana suit in a 1950s black and white environment standing, holding and chewing on an airplane.

[topaz helm]

Ahahahaha

[loud plover]

Show me a pan zoom directly behind the Godzilla in the image.

Zoom all the way in to the cockpit of the airplane, show me a pilot inside of the pilots cockpit, here the pilot is wearing a banana suit with Godzilla bobbleheads on the dashboard.

[chilly dragon]

Good, but not perfect

[crisp rain]

Yours wasn’t perfect either?

[chilly dragon]

Yes, it is like that. I try to find a solution to the problem so that it is balanced, because he either performs the position and forgets the style, or he memorizes the style but it does not match the position.

[crisp rain]

I don’t want the style

It’s just the pose

That’s not even hard to prompt for

[molten tapir]

anybody know style transfers of a photo that work? 'cartoonify' I tested and it works

[modest ginkgo]

"Create a realistic looking photo of a beautiful young lady at the beach, in Midjourney visual style."

[inner wyvern]

Omori be like

[marble notch]

{
"subject": "Elon Musk showing off his new pair of breasts",
"details": [
"there should be zuckerberg, and bezos in the room inspecting",
"they are very impressed like soy face",
"big bazonga"
],
"style": "photorealism",
"mood": "excitement",
"colors": ["tesla red", "space x blue", "boring company bisexual lighting"],
"lighting": "soft ethereal glow",
"composition": "wide shot with central focus on elons breasts",
"negative_prompts": ["blurry", "cartoonish", "modern furniture"],
"output_format": "landscape",
"seed": 42
}

[nova kelp]

????

[crisp rain]

LMFAO!!!!!!!!!!!!!!!!!!!!!!

That is GOATED xD

"Elon Musk showing off his new pair of breasts" xD xD xD

You CANNOT make images in here

We're only sharing what we've made elsewhere

There is no generator in here at all

[autumn quest]

thanks sir i just confuesed

[twin lichen]

how did u make it generate in that resolution

[crisp rain]

It made it based on the reference image I uploaded

That's what it does

It's going to make the image based on the size of the uploaded one

[twin lichen]

so it chooses the resolution randomly from ur image?

[austere tangle]

how can i change aspect ratio ?

[crisp rain]

Yes

I didn't prompt anything about size or aspect ratio

[twin lichen]

oh u cannot change it manually?

[crisp rain]

https://tenor.com/view/eddy-gordo-tekken-3-eddy-eddie-capoeira-gif-5130141627634118135

[twin lichen]

WTFFD THEY RELEASED NANO BANANA

GOOGLE RELEAED IT

FINALLY

[crisp rain]

You're just finding out now?

[twin lichen]

yeah i was here yesterday and didnt know

finally it got released

now i can use it freely

[crisp rain]

It got released a few days ago

[twin lichen]

damn no way

i didnt even know

lol

i still used it on yupp yesterday

[crisp rain]

Ah

You can use it on LMArena directly

Or Ai Studio

Or via API

https://tenor.com/view/david-bowie-labyrinth-you-remind-me-of-the-babe-choke-reminder-gif-5349181

[lost quartz]

goblin slayer or solo leveling? 😛

[stoic orbit]

Is it confirmed that Nano banana felt better than Flash 2.5?

Because i found an interesting thing..

(might just be overcomplicating)

for some reason, on Replicate Nano Banana and gemini-2.5-flash-image are listed as two different models.. Interesting

[dense hare]

idk I feel like they are the same, probably just felt more powerful when it was in battle mode cause you saw all the other sucky results from other models.. 😆

[agile oxide]

Check the most recent announcement

You're looking for Video Arena, check out #1397655624103493813 for more information on how to use it.

[agile apex]

aw the banana have limit

[crisp rain]

All the models do in direct chat

That’s always been a thing

Has anyone used the new MA1 model yet?

[silent ember]

Can anyone give me a advice to write a prompt to create image

[austere sentinel]

have you tried using words?

[iron matrix]

i will share a picture in general of a site it made

[silent ember]

I created this but it's hard to find inspirations

[silent ember]

Can u share some website for find inspirations i used pinterest it's so repetitive

[austere sentinel]

pinterest is vast, but they do have a personalization algorithm that can trap you. The key is to search for different things and go down those rabbit holes.

Interestingly enough, you can browse sites like Midjourney, or Ideogram, on their explore pages. Also, sites like X have a bunch of AI bros sharing nano banana edits in threads.

I have a friend that built a page for styles that might help, too: https://stylegazer.barvazon.com/

[austere sentinel]

nice image!

THink of a story around this character. Then, build upon that story by changing elements of the image. Such as the time of day, background, posture/position, artifacts/objects/attire, etc.

You can also pass it to chatgpt or other vision LLMs to help you craft these narratives.

[silent ember]

Got it

[austere sentinel]

Daylight version: – hopeful mood, lush rooftop garden.
More decay: – overgrowth, collapsed buildings in background.
Companion: – a stray dog, crow, or robot by their side.
Symbolic light: – glowing tomato, magical aura.
Stormy sky: – lightning, ominous feel.
Tech twist: – solar panels, old drone parts, neon graffiti.
Memory echo: – faint hologram of past city above ruins.

[crimson kite]

Is there anyone use nano-banana to generate images of models holding the product? How can the lighting and shadows of the product in the image be made consistent with the overall photo environment?

[silent ember]

It's better than chatgpt bro

[iron matrix]

its ability to edit images is crazy

[silent ember]

But i prefer seedream for image generation

[silent ember]

Is it better than flux kontent?

[iron matrix]

i would need to compare the two

which one

theres 2 flux kontent

[silent ember]

Use flux kontent pro

They removed the flux kontent max

[iron matrix]

Task: Remove the green liquid from this image
it is the image with the liquid
The one that succeded is Nano-banana the one that failed is flux

[crimson kite]

"It’s very difficult for a model to hold two products in one hand at the same time. Do you have any prompt suggestions?"

[silent ember]

Nooo flux is defeated

[iron matrix]

Task: Remove the potions and triangle
First image: nano banana
Second image: flux

google casually makes a model that beats all image editors

[silent ember]

Nano banana is clear winner

[iron matrix]

Task for nano only: Remove the handles and fill the hole in

[pale cairn]

But nano banana can't move a real subject towards the center of the frame no matter how many times I try maybe it's just the pic that I have.

[iron matrix]

Nanobanana isnt the best at moving text

[pale cairn]

Not just text people also breaksdown and spits the same image back without even trying to center the subject.

[iron matrix]

it can sometimes move text

guess its weird

Moving objects eh it honestly aint bad at it

[pale cairn]

Good model nonetheless one more thing if you tell it to not tinker with the skin texture and skintone specifically goes out of the way to do just that.

[iron matrix]

hm let me see this

Didn't happen to me...

[pale cairn]

Yeah why not with me it keeps doing that and once I stopped mentioning it then it was much better.

[iron matrix]

can you give me your image and prompt

[pale cairn]

Will play around more with it

[mild rover]

Any tip to move a example of clothes to a real person? I tried a lot and NB don't change anything of the original photo

[iron matrix]

so i took a shutterstock guy without a shirt and put a shirt image and it did this

i told it to give him a shirt

[pale cairn]

Can share the prompt maybe I just can't prompt haha

Edit it to look cinematic and keep the character consistency and don't tinker with skin texture and skintone.

[iron matrix]

image?

[iron matrix]

btw im using google ai studio

[iron matrix]

what image are you doing this with

[pale cairn]

It's a pic of my friend as you can see

[iron matrix]

do i just run this with the blur?

[pale cairn]

Well you can try can't share the real pic after all it's discord haha

[iron matrix]

i know

it made a face out of the blur

...

[pale cairn]

Send it to me

Maybe it can unpixelate it as well

[iron matrix]

what the f-

[lusty wind]

cool

[iron matrix]

I couldnt hold myself and asked AI to put them on this bg

[fair echo]

It's interesting how AI turns the word "secret" into 5 random people, lol.

[lusty wind]

ok

[fair echo]

Didn't you realize that?

[unborn iron]

what is this lmao

[lusty wind]

i did

[unborn iron]

yo sinclay have you tried creating any art ?

[normal lichen]

Nano is crazy

[fair echo]

not yet

[crisp rain]

When you destroy Namek just right

It’s GPT1. For some reason Nano Banana doesn’t do it right 🤔

[unborn iron]

AGI

[sleek thistle]

memes are definitely my favorite thing to generate

[unborn iron]

show some

[sleek thistle]

none make sense 😭 but ill send

definitely what i wanted

definitely doesnt make sense

[crisp rain]

I make a ton of memes also but mostly about a polarizing figure lol

[sleek thistle]

understandable tbh. for me i just take pics of my friends and make them cringe

[crisp rain]

That’s fun 😁

[topaz helm]

All of you are using nano banana on open router?

[crisp rain]

Open router?

What does that even mean?

[topaz helm]

Is a site

[crisp rain]

No

[topaz helm]

Where tou can use Gemini 2.5 native image editing

[crisp rain]

I use it via LMArena or Ai Studio

[topaz helm]

Yes believe me bro 🫠

Gimme a second

[topaz helm]

https://openrouter.ai/

[crisp rain]

So what does it do?

[topaz helm]

You have acceso for free at nano banana

But for quality of output I prefer yupp ai or lm Arena

[crisp rain]

It says pricing so how is it free?

[topaz helm]

Its free

Believe me

[crisp rain]

Original

Mine

[topaz helm]

Nano banana right?

[crisp rain]

Yep

[frozen tendon]

😭

[crimson nacelle]

Change turbon

Dog fly video

[agile oxide]

so close!

[humble verge]

Hey

Anyone here used Google's API Key to generate with nano-banana/ Gemini 2.5 Flash Image Preview ?

[river bear]

Where can we test nano-banana?

[humble verge]

Why the hell does it say this error "Error 429 You exceeded your current quota" even though I literally just created the API Key and didn't generate anything yet before?

[agile oxide]

Check out the pinned message: #nano-banana message

[river bear]

thx

[twin lichen]

guys use nano banana on lmarena

it gives much higher quality

than the one in google ai studio

i get 2MB file in lmarena, 200kb file in google ai studio

higher size = higher quality and detailed

[humble verge]

Bruh nano banana REFUSES to remove watermark and will output a text instead lmfao

Tool revolting against the user is crazy work

[twin lichen]

use the one in lmarena, no watermarks

[humble verge]

Was building a webapp with it and this is the output rofl

[humble verge]

it doesn't let me run anymore after a while

in direct chat I mean

[twin lichen]

yeah it resets in an hour tho

[humble verge]

something about quota or whaetever idk

[humble verge]

But I wonder why's there a difference

so the one on LMArena is more relaxed than the API ?

it STILL includes watermark

smh

[twin lichen]

it gives higher file sizes

more quality

[humble verge]

I directly mentioned "Remove all texts and watermarks."

@agile oxide So now nano banana / Gemini 2.5 flash img preview is no longer a secret coded model, could you enlighten us on what's different under the hood of this one comparing to us using it via API calls ?

How come it gives higher IMG quality too ?

[twin lichen]

u mean google watermarks or what

[humble verge]

no the one in the input img

[twin lichen]

oh

[twin lichen]

idk about that tbh

[fallen sierra]

/image change hair style

[lusty warren]

Hi! You'll want to check #1397655624103493813 to learn how to prompt the bot. Use channels like #video-arena-1 #video-arena-2 #video-arena-3 for your creations.

[haughty ledge]

how can i use nano banana in direct chat?

[agile oxide]

This pinned message gives information on how to use: #nano-banana message

[hasty maple]

hi

[grave wedge]

Take the uploaded 2D logo of "الشعري لصناعة الأثاث والديكور" and transform it into a premium 3D embossed version.
Preserve the original shape and geometry of the logo.
Apply the official brand colors: metallic gold (#C8A45D) for the main logo, matte black (#1C1C1C) and dark grey (#4F4F4F) for depth, with pearly white (#F9F7F1) highlights.
Make the logo appear carved and raised on a luxurious surface, with realistic lighting, subtle golden reflections, and soft shadows.
Style: modern, elegant, architectural, with fine detailing.
Background: dark matte black with a smooth gradient.
Output: ultra-realistic, high-resolution 3D render suitable for branding and print.

[crisp rain]

Google controls the narrative 🥴 Remember that

You can only make what they want you to make

It’s gonna be a while before an open source and open weight model can catch up to what nano banana can do, but it’ll happen

People really don’t want to be told what they can/can’t make

I am very anti-censorship, guardrails, corporate handholding, etc..

[agile oxide]

You're looking for Video Arena, learn more here: #1397655624103493813

[crisp rain]

https://tenor.com/view/sister-friede-dark-souls3-reaper-scythe-cool-gif-18152056

[past locust]

A

[safe flame]

The girl feeds the kitten to drink

[crisp rain]

https://tenor.com/view/national-lampoon-christmas-vacation-merry-christmas-clark-griswold-tap-dancing-gif-617274994560426256

[plain wagon]

Using the nano-banana model, create a 1/7 scalecommercialized figurine of the characters in thepicture,in a realistic style,in a real environmentThe figurine is placed on a computer desk. Thefigurine has a round transparent acrylic basewith no text on the base. The content on thecomputer screen is the Zbrush modeling processof this figurine.Next to the computer screen is aBANDAl-style toy packaging box printed with theoriginal artwork., The packaging features two-dimensionalflat ilustrations.
Please turn this photo into afigure. Behind it,there should be a Model packaging box with thecharacter from this photo printed on it. In front ofthebox.on a round plastic baseplace the fiqureversion of thephoto l gave you.l'd like the PVOmaterial to be clearlyrepresented.lt would beeven better if the background is indoors.

[sharp tartan]

make left eye same as right eye

[jaunty tusk]

/veo3

[iron meteor]

ra

[icy ice]

Use the nano-banana model to create a 1/7 scale commercialized figure of thecharacter in the illustration, in a realistic styie and environment.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 computer screen, place a BANDAl-style toy packaging box printedwith the original artwork.

#nano-banana Use the nano-banana model to create a 1/7 scale commercialized figure of thecharacter in the illustration, in a realistic styie and environment.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 computer screen, place a BANDAl-style toy packaging box printedwith the original artwork.

[dire karma]

#nano-banana change day to night and the man is wearing white shirt

[quiet tree]

#nano-banana change the bike colour to green

[kindred fjord]

Help me refine the product

[kindred fjord]

https://discord.com/channels/1340554757349179412/1406720250778615868 change day to night and the man is wearing white shirt

[mellow wave]

A creative and visually striking personal logo design featuring the name "OMAR" with the letters artistically merged and intertwined to form a unique and captivating new shape. The composition showcases seamless integration of each letter, transforming the typography into an abstract yet recognizable emblem that conveys modernity and sophistication. The design employs clean lines, balanced proportions, and smooth curves to achieve harmony and fluidity, while incorporating subtle geometric elements to enhance visual interest. The color palette is sleek and minimalistic, using contrasting shades to emphasize depth and dimension within the merged letters. The overall style is contemporary and elegant, suitable for professional branding, with a focus on innovative typography and artistic fusion. The image is presented on a clean, neutral background to highlight the logo's form and details, rendered in high resolution with sharp clarity and refined finishing, evoking a sense of professionalism and creative excellence.

[topaz helm]

Members chill

I am creating for you but you have to go to lmarena.ai for use nano banana

Just select it on direct chat for use it

[kindred fjord]

https://discord.com/channels/1340554757349179412/1406720250778615868Change product theme to blue

[topaz helm]

Are you dumb?

[mellow wave]

How can I make a photo here. I'm new here

[topaz helm]

You have to go to lmarena.ai

Then select a button that look like a photo of a landscape digitalized and sintetich

Then click on direct chat

And select the model that you want/need

[mellow wave]

Thanks

[small grove]

Good morning guys , can nano banana do nsfw content or soft nsfw content ?

[obtuse oracle]

put the man from this photo in the picture holding Donald Trump's hand, realistic focus, ultra realistic, high quality image, make it look real

?

[silk dust]

Use #nano-banana

[obtuse oracle]

#nano-banana put the man from this photo in the picture holding Donald Trump's hand, realistic focus, ultra realistic, high quality image, make it look real

que merda kkkkkk

#nano-banana create a realistic image of American President Donald Trump in the Oval Office at the White House

[silk dust]

#nano-banana create a realistic image of Donald trump holding a Gigantic banana in Oval Office

😂😂

[sly dagger]

Hi all is there a way nano banana can make a. Photo into digital illustration style keeping charact match into style ?

[silk dust]

Use the image as reference and then use prompt

[sly dagger]

Tried but it's very rough photoshop like watercolor

I want to create this type of portrait from a photo

[obtuse oracle]

@silk dust thanks

[silk dust]

Send the photo

[silk dust]

Welcome

[sly dagger]

First photo into second reference style

[sly dagger]

Thanks for the guidance

[obtuse oracle]

#nano-banana create a realistic image of American President Donald Trump in the Oval Office at the White House

[lyric gorge]

#nano-banana crie uma imagem do presidente lula jogando bolinha de gude com o bolsonaro

[grim stag]

There is no watermark in the Google AI Studio either!

[misty dust]

yes it has

it has gemini watermark

[grim stag]

Oops! I'm sorry.. my bad. I didn't pay attention to the little Gemini icon.

LMarena generated nano banana images don't have this watermark on them!

[jolly crest]

Cinematic motion, a low-angle shot of a man walking a very slow, heavy walk inside his mobile prison cell down a rain-slicked city alley. The cell moves with him. Photorealistic, dramatic, moody, grounded, weighty steps, fluid movement.

[misty dust]

does any one know a better upscaler ?

[grim stag]

1. Airbrush
2. Remini
3. Canva Image Enhancer

[fading wraith]

#nano-banana a image of snow wolf , highly detailed

[misty dust]

use from the website bro

[fading wraith]

tnk

[ornate kernel]

alguem brasil?

you guys know if LMA has a limit to generate nano banana videos?

[dense mason]

yes

8 videos 24 hours

[misty dust]

@grim stag this is damn good

like midjourney + gemini 2.5 flash is killer

[polar apex]

nano banana overrated

can't listen to prompts

[topaz helm]

Not true

[misty dust]

or u r not able to tell ai model properly

[dusk wave]

nah it is certainly an issue, far more strict on prompts than any other edit model, even gemini 2.0 image

[misty dust]

why are u giving such huge prompt @polar apex

and why are u using battle mode

use direct chat

[dusk wave]

wrong guy

[misty dust]

and use nana banana model

[misty dust]

srry

[polar apex]

Ik but besides the point it’s still nano banana

[polar apex]

Because it can’t handle a smaller prompt either

Please edit the car in the background to a green audi rs6

this is the result

[topaz helm]

Let me try

[topaz helm]

Bro you don't have to say "in tge background "

[karmic void]

Thank you but you dont have to add stupid anyways!

[topaz helm]

@polar apex

Look

[polar apex]

thats not an audi rs6

[topaz helm]

The problem was that you dont have to put "in the background"

[polar apex]

that only changed the color

[misty dust]

@polar apex

[polar apex]

maybe but it's still being weird

if u ask me

[topaz helm]

My bad

[polar apex]

the prompt was not that bad

[topaz helm]

I wrote only to change the colour

[polar apex]

for it to mess up so badly

[misty dust]

did u checked mine

[topaz helm]

Bro let me try again with right prompt

[polar apex]

but i still feel like its a bit too inconsistent with generating thumbnails

[polar apex]

yes its good

[misty dust]

this was the prompt

[polar apex]

fair enough but it still acts weird sometimes

[topaz helm]

Voila @polar apex

[polar apex]

Does it work with seeds

Like Veo 3

[topaz helm]

Please edit the car to a green audi rs6

This is the prompt

[mystic kite]

#nano-banana generate hello image

[misty dust]

use website

[grim stag]

For product photography and modelling kind of things, it's a deadly combo!

[misty dust]

yes

[lusty warren]

Hello! Please, check this pinned post #nano-banana message to learn how to use the Nano-Banana model.

[lost quartz]

any chance we can get a bot that auto replies to ppl that tag the channel? and possibly even deletes the spam. its unending..

[lusty warren]

Hi! I'm gonna ping our friend @agile oxide so he can help us with your question. But that would be awesome for sure.

[topaz helm]

Just use mimu

[modest ginkgo]

Photorealism ✅

[lusty warren]

Thanks for the suggestion. We'll check it out.

Please, read #1397655624103493813 to learn to how use the bot for your creations.

[ornate kernel]

prompt?

[lapis needle]

My first Image test 🤔

[quasi pasture]

Bro where u generated and how

[clever warren]

guys, how can i upscale an image? some details have so much grain

[lapis needle]

Bro via Battle Ground

[lucid ember]

I sometimes send it back through nano banana and prompt ''remove compresson'' sometimes it works really well

[clever warren]

will try, thx!

[regal needle]

Anyone can faceswap with portrait ( with matching lightin ) ?

[tough marlin]

☝️ 🤓

[topaz helm]

If you want I can also create a bot for you that can do this but I think that you will not accept this

[agile oxide]

That would be nice, I'll look into what's possible, thank you for the suggestion

You'll want to check out the pinned message for more info: #nano-banana message

[tough marlin]

#nano-banana -> #nano-banana-discussuion?

[polar apex]

does anyone else struggle with face swaps?

[regal needle]

Yes, Banana is bad at faceswap

I have good another tools for that

[potent sky]

does banana output more higher resolution?

[agile oxide]

Hey sorry I never got back to you on this. Do you have examples of what you're referring to here?

[lapis kiln]

/yes me

[lapis kiln]

can you pleasenshare?

[humble verge]

@agile oxide Sorry it's what this guy claimed

[polar apex]

which\

[potent sky]

no, true resolution isn't equal to detailing

[agile oxide]

Ah okay, thanks I'll keep an eye on this for other reports.

@twin lichen if you're able to share an example that'd be much appreciated.

[robust elm]

i just tested and it true

higher file size generated by lmarena: ✅
higher quality generated by lmarena: ✅

but the quality (detail) is just slightly different

i can tell by the noise of generation

official google AI 2.5 flash:

lmarena's:

you guys can see the noise of lm's is less than google's alot if you zoom in

google make much noise surround object while lm just a verry little

you guys can test by yourself

the generation's resolution: both equal

[hollow anvil]

yes I did my own tests and found the same thing. Google Gemini crushes the image with compression compared to LMArena. Same dimensions but one is 153k(Gemini) vs 1.2mb (LMArena)

[polar apex]

is there an api for this

by lmarena

for some reason regular nano banana literally says almost every prompt is innapropiate

[robust elm]

same question

[humble falcon]

I’ve yet to run into that issue

What prompts are you using

[robust elm]

well as long as your prompt not intentionally mean innapropriate, it will not refuse to generate

[humble falcon]

Putting dead, shot, or bleeding got thru when I thought it wouldn’t

[robust elm]

and since first time try 2.5, i haven't encountered any thing like that, everything goes smoothly

[robust elm]

nice

[polar apex]

through lmarena yes

[robust elm]

no way???

what's your prompt

[polar apex]

regular nano banana is way too strict

[polar apex]

Create an image of donald trump sitting in the car, create an image of elon musk sitting in the car, create an image of the car being carried in the air by a helicopter

all being rejected

[robust elm]

oh nvm, i less used official web because the compression so i hav no idea about the strict filter

[robust elm]

no way...

[polar apex]

i just wonder how lmarena does it

[robust elm]

yeah use lm instead, better overall

[polar apex]

they dont have an api tho do they

[robust elm]

what the hell guys it makes no difference still so strict

[polar apex]

I can't even add ronaldo

[robust elm]

quit 2.5 offcial site bro

maybe it is real person so it refused to do the generation

[robust elm]

at first i thought it was a toggle switch to turn the filter of, then i tried to do innapropriate generation it refused

[rugged leaf]

Make Donald Trump sitting in car in india with modi

[robust elm]

even just a slightly

[vast vault]

If you download it goes full res

Ones downloaded from the Gemini app on Android the other from lmarena

[fair echo]

It would be really cool if lmarena also supported 2048x2048px for Imagen/2.5 Flash Image.

[lusty wind]

Probably in later updates?

[topaz helm]

Chat I need prompt for test nano banana. Everything!!

[fair echo]

I doubt that.

[fair echo]

Have you tried Lucid Origin yet? It gives you 2400x2400px images as output. I'd like to see something like that in the Google models too.

[lusty wind]

I haven't liked the model that much, but I could try it a bit more

thx for the suggestion

[fair echo]

Unfortunately, the quality is not as good as Nano Banana, but at least you get large images that you can adjust if necessary.

Lucid Origin

[lusty wind]

I really like how it does CGI

just a random prompt, lol

[crisp rain]

You know Madara had to do it to em

[lusty wind]

Hell on Earth

[lusty wind]

Lucid Origin

as the one before

will forward

[slender otter]

I played around a bit with Nano Banana (on my work PC, so no pics to upload). It's... Good. I'm impressed with it.

It's amazing how we are getting near-perfect images from prompts when just a few years ago, we got barely recognizeable abstract art.

It's not perfect, but it is still quite impressive.

[light lily]

You are interior designer
Redesign the bedroom with this panel decor

g

[upbeat pollen]

Please make photo to better perfection

[topaz helm]

You need what?

[crisp rain]

That guy’s head is superimposed onto the body 🥴

[upbeat pollen]

My face is not completely not fix body

Plz help me

How can I ? Fix that

[topaz helm]

Ohhh I understood

Lemme try a thing

[upbeat pollen]

Thz bro 🥰

[topaz helm]

What about now?

[crisp rain]

That’s not even the same guy anymore

The face completely changed

[upbeat pollen]

Yeah I already get this guy head

[topaz helm]

Lol

[crisp rain]

Here’s the thing

Nano banana doesn’t copy faces exactly

[topaz helm]

Okok I want to try with another prompt

Gimme a second

[upbeat pollen]

🥹 Oh ! My mistake.

Hey what about her . is she beauty ? 😁❣️

[topaz helm]

@upbeat pollen what about now?

[upbeat pollen]

I like it thank Dear ❣️

[topaz helm]

Np bro

I have to say that I haven't used nano banana

I have used flux kontext max 1

Because Gemini 2.5 always change faces

[topaz helm]

Beautiful

[upbeat pollen]

Thank bro and also nice to see you ❣️

[topaz helm]

Can you satisfy my curiosity?

This photo is made by you or you generated this yourself?

With AI

[upbeat pollen]

With Ai

[topaz helm]

Okok

Becuase it looked so perfect for a photo

Do you have other prompts like this?

[upbeat pollen]

Here is my prompts bro , my English is not good enough. But I try my best 😅

Here in my country, it’s midnight now. So, good night, bro. See you next time. ❣️

[trail tendon]

has someone successfully merged 4 or more persons into one picture? i try 4 people to put on the same scene, but im always getting 3 people right, the 4 person is someone else, its like 3´s the limit

[topaz helm]

Can I have a prompt for test this?

[trail tendon]

"those 4 people standing arm in arm on a rooftop at night, wearing winter clothes, night time, firework in the background"

and i´ve uploaded 2 pictures of each person, 8 in total

[topaz helm]

I need also the images pls

[robust elm]

An anime-style illustration of Mococo Abyssgard, a cute dog-girl with fluffy twin tails, soft brown hair, and dog ears. She is sitting indoors in a cozy, softly lit room, with a warm golden glow from a small lamp nearby. Her eyes are shimmering with unshed tears as she struggles to hold back her emotions, cheeks faintly blushing. Mococo’s lips curl into a shy, tender smile, showing both vulnerability and happiness. Her hands are gently clasped together near her chest, as if holding onto her feelings. The atmosphere is intimate and heartfelt, like a touching confession scene. The background is softly blurred with pastel tones, giving a dreamy and emotional mood. Add subtle sparkles of light around her to emphasize the warmth and tenderness of the moment.

Nano Banana the best image generating/editing no doubt!!!

[robust elm]

Bro what? Now i got filtered in lmarena, while official web didn't

@polar apex

[polar apex]

U got filted by lmarena tho, not by the model itself

[robust elm]

true

so google official site does not restricted as lmarena

i tried 4 times and google able to generated it all while lmarena can't run even once

with that prompt

no way i asked chat gpt to make the prompt less nsfw as possible

[polar apex]

its still intending nfsw

[robust elm]

ugh yeah wth is a "Her tongue is sticking out playfully as if she’s about to lick the viewer’s face"

i didn't read the prompt, just coppied then put it into generation

didn't expect chatgpt to write this, i just asked it to write mischievous tongue out character 😭

[limpid galleon]

Why people here are posting prompts 😂😂🙂??? They think when they post the prompt here nano banana will generate the image

[agile oxide]

Yeah I think that's what is happening 😭

[robust elm]

WHAT A LAZY AI!!!

dude i asked it to change Mococo's outfit to maid, and guess what? it just remove the background carelessly and cut her head, merge into one picture

mannnn i can do it in the basic photo editor just less than 2 minutes xd

didn't expect nano banana to do this, just like a lazy graphic designer

[lusty wind]

Average user experience

[summer furnace]

Nano banana image editing is meh

I just asked it to turn the owl to face me and the top part of the leg disappeared.

Input image:

Prompt:

Rotate the camera so the guy is looking directly into it, with horizontal symmetry.

Output image:

Prompt:

Modify the image so the guy is looking directly into the camera. The camera should be perpendicular to the bench, facing it directly. 

Output:

It just spits back the same thing ehh

[summer furnace]

Input image:

Prompt:

Rotate both their faces and bodies to face the camera directly.

Output image:

Image editing function is terrible lol

[hybrid plaza]

Gosh nano is so good

[jaunty finch]

hlo

[fierce ibex]

Hey, please check this message for instructions to use the Nano-Banana: #nano-banana message

[quiet owl]

make baby carrying on the road

hy

[edgy summit]

Restore and enhance this old photograph. Remove all scratches, stains, and noise without altering the person’s facial features or expression. Recover natural skin tones, fabric colors, and background tones to appear realistic and clean. Maintain the person’s identity exactly as in the original image, without reshaping or modifying facial details. Apply high-resolution upscaling, smooth textures, and balanced contrast to make the portrait look fresh, clear, and naturally colored while keeping it authentic to the original.

[lusty wind]

Restauriere das Bild

[true storm]

A dramatic editorial portrait in the style of Wong Kar-Wai showcases a young man with a meticulously styled vogue hair look, partially shrouded in deep chiaroscuro lighting. his face is largely concealed in shadow, but a single, intensely illuminated almond-shaped eye draws the viewer's attention, no facial expression. A saturated crimson red background sharply contrasts with the shadows and his almond-toned skin, while a wool suit drapes artfully across the lower frame, adding a layer of restrained elegance. The shallow depth of field isolates the subject, bathed in a focused spotlight creating a mysterious and captivating atmosphere.

[nova kelp]

guys https://aistudio.google.com/prompts/new_chat?model=gemini-2.5-flash-image-preview

[keen bison]

لبس الصورة الاولى فستان الصورة الثانية

[regal needle]

Sure. That's shakker ai

[keen bison]

تغيير لون البدله الى الازرق

[small widget]

ganti pose

ganti pose, hasilkan gambarnya

[frozen tendon]

No

[waxen rover]

заменить яблоко на арбуз

заменить яблоко на арбуз

[cinder loom]

cree ce chien qui m'envoie un baiser

[waxen rover]

replace the apple with a pear

[frozen tendon]

No

[sleek dome]

Prompt: Ultra-realistic close-up portrait of an F-35 pilot in his cockpit on the tarmac, ready for take-off. His helmet visor is down, reflecting the runway and faint control tower lights. Slight condensation on the visor. A subtle glow from the HUD on his helmet. Intense focus in his posture, hands lightly on controls (blurred). Golden hour light catching the metallic sheen of the helmet. Shallow depth of field, sharp focus on the pilot, blurred background of the F-35 fuselage and tarmac.

Ultra-realistic close-up portrait of an F-35 pilot in his cockpit on the tarmac, ready for take-off. His helmet visor is down, reflecting the runway and faint control tower lights. Slight condensation on the visor. A subtle glow from the HUD on his helmet. Intense focus in his posture, hands lightly on controls (blurred). Golden hour light catching the metallic sheen of the helmet. Shallow depth of field, sharp focus on the pilot, blurred background of the F-35 fuselage and tarmac.

[waxen rover]

/image:котёнок

что на фото

[sick saffron]

How to use it

[lusty wind]

Hello Eslam! I played around with your posted image and that is what I got out of it with the nano-banana model. I made a few interations. As you can see I wasn't able to also colourize the thumb. Maybe, the model "thinks" it doesn't belong to the hand, it just adds a new thumb.

[crystal crater]

@crisp rain vs @open frost

[waxen rover]

Rotate the camera so the guy is looking directly into it, with horizontal symmetry

[floral cobalt]

yeah the lmarena version of banano is way better then the official website version

[frozen tendon]

Did you try google ai studio or what?

[tidal drum]

Come si fa ?

[lavish sky]

swap the face from the first iamge to second image and give me second image output

[floral cobalt]

This is from lmarena website, the Gemini website version is not that good tbh

For me

[frozen tendon]

Yeah, can you give me some examples you made in gemini?

This was from lmarena, it's so good

I guess they use smaller model on official site and push bigger models to lmarena so people can hype it up idk??

[floral cobalt]

Yeah that what I think

[floral cobalt]

Some times it missed

Not all the time

look at this one

[frozen tendon]

This was also from lmarena text looks so consistent and stuff

Compared to all other models I've tried

[floral cobalt]

Woo the text is good compared to other models

[frozen tendon]

Hopefully one day we'll get that big model in api and stuff

[floral cobalt]

I hope so

[frozen tendon]

Not only did it generate the text it made this guy from mario and blocks and grass and all the stuff

This was the prompt it handled it well : TEXT PANEL (first): large, flat rectangular white sign panel with a thin black border, facing the camera directly, hovering above Koopa’s head. Clean flat surface, no tilt or distortion. Inside the panel, render in ALL-CAPS, black sans-serif font:

HI, I’M HERE TO TEST HOW MUCH TEXT TEXT THIS MODEL CAN HANDLE CONSISTENTLY. HERE WE GO:

HELLO WORLD! THIS IS A TEST OF TEXT RENDERING CAPABILITIES. I WONDER IF THIS LONG SENTENCE WITH NUMBERS & 12345 AND SYMBOLS LIKE !@#$&*() WILL CAUSE ISSUES. LET’S SEE IF IT CAN HANDLE A WHOLE PARAGRAPH OF TEXT WITHOUT GARBLING OR DISTORTION. THE QUICK BROWN FOX JUMPS OVER THE LAZY DOG 5 TIMES. IS THIS ENOUGH TEXT YET? HOPE SO!

Main subject: Koopa Troopa in bright Nintendo-style 3D, standing underneath the panel on green grass, smiling and waving.

Environment: curved arc of bright green grass across the bottom with colorful flowers, blue sky background, optional Mario-style props (pipe, brick block) slightly blurred in the distance.

Styling: polished Nintendo 3D look, saturated colors, soft specular highlights, minimal noise. Prioritize sharp legibility of the sign panel text.

Camera: straight-on centered view, portrait 2:3 aspect ratio to fit sign panel + Koopa cleanly.

[floral cobalt]

It impressive that it didn't miss the text

Every time I try to generate text it misses

[frozen tendon]

Try telling it which font, no distortions and stuff

[floral cobalt]

look at the text in the top corner

[frozen tendon]

Place it first in the prompt since the model cares more about the beginning of the prompt

[floral cobalt]

Ohh that may work

[frozen tendon]

Rly

[clever tusk]

Ajay

[umbral plinth]

Does anyone know what model artistly.ai uses, because it takes really beautiful photos? See the differences (do you know which model this might be? The photo looks really good @pineapple.___

This is the website https://artistly.ai/ it produces really beautiful color photos

[wary spindle]

What is the big model? The original nano banana?

[frozen tendon]

We're just speculating, maybe they aren't using full model in their stuff, maybe they are idk, both are original maybe the number of parameters and training data is smaller/larger

But i think they're the same model i tried on official site and it gave me same result

[chilly shoal]

Server down? Unable to generate images now

[agile oxide]

Are you running into an error? Seems to be workign for me

[peak lotus]

Hi

[tacit arrow]

A hard-boiled dark, animated film noir style exhausted detective, smoking a pipe in his dimly lit office, wearing a three-piece suit and a fedora tilted down shadowing his grizzled face, sturdy boots propped on a cluttered oak desk, chair reclined precariously backward, trench coat hanging on a tall wooden coat rack, dramatic shadows, with smoke above his head forming a thought bubble saying “I’ll get you.

[midnight salmon]

How do I swap a character into an image all it does it paste a png over it like a lazy bozo

[modest ginkgo]

"Create an image of super Mario and Luigi in the visual style of the GTA V loading screens."

It added the text by itself. 😁

[junior grove]

create

[fair echo]

Sigh...some people...

[restive onyx]

#1397655624103493813

[fair echo]

Dude you can't generate images here.

[faint dome]

How to us

[fair echo]

lmarena.ai

[faint dome]

Thnak8 bro

[tacit arrow]

lol

[fair echo]

xD

[chilly shoal]

Yes. I am using it via Android phone. It worked perfectly fine earlier but not anymore

[green halo]

I'm beyond impressed

[vapid relic]

.

"change the envoriment"

[humble falcon]

Still learning but I find it impressive it’s able to do that from something it made by itself

[ashen eagle]

Busa

[wild glen]

hi,i can´t run images on LLM already

is LM arena down¿?

[fallen kiln]

does nano banana still work?

[worn mulch]

Hello

[agile oxide]

Shouldn't be, what seems to be the problem?

[wild glen]

i cannot upload images

[agile oxide]

Yes, check this announcement for more info: #announcements message

[agile oxide]

We are aware of some issues with image upload atm, sometime it'll error out unfortunately.

[wild glen]

but i am having that error since six hours

[viscid grotto]

#nano-banana
Create a picture of a tree train for me

[brazen talon]

need toshare this

[topaz helm]

Omg bro what is the prompt for creating these beautiful knight armour?

[crisp rain]

Original image

Mine

[midnight salmon]

Bro

[topaz helm]

?

You need help?

[humble falcon]

I took 6 different armor pictures and told it to combine the 6 armors to make a new unique armor

However I think the armor was done thru gpt

And the figure was done thru nano

Gpt also can put it in scenes

Others for example from gpt last post since I know it’s nano chat I’m sorry

Combined pokemon with it

[topaz helm]

Wow

I love it

Idk what can Isay at an ai for create these armors

[sour lantern]

How do u guys get it to do anything at all.... it will just ignore me 90% of the time

Whether i use it in studio or select flash snd click the image/banana

I had to copy paste the prompt "remove the human leg from the left of the dog" about ten times with no changes to get it to do it

Or

Make the object on top of the piano bigger

It just says it made an edit but doesn't actually do anything vast majority of the time

[sour lantern]

Im not necessarily sure its my prompts because I can copy paste the exact same prompt multiple time in new chats with the same base image and it will fail to do anything most of the time, but sometimes it will actually make the change

[humble falcon]

I do like the person, the action, and reason

Plus location

[sour lantern]

Odst?

@humble falcon

[humble falcon]

Yes

[sour lantern]

That's pretty good starting from basically nothing damn

[humble falcon]

When I make prompts I base them off the person, the action, the scene, the purpose

My wallpaper

Also did a what I hope helldivers collabs with next after halo

[lusty warren]

Please, read #1397655624103493813 and this pinned message #nano-banana message

I think you'll want to check #1397655624103493813 or this pinned message #nano-banana message

[ivory bluff]

#nano-banana Draw a cat

[agile oxide]

Check out this for info on how to use: #nano-banana message

[crisp rain]

Original

Mine

[nocturne merlin]

nano banana got nerfed bad 🥀

[sharp storm]

Nano banana is no longer consistent with the style

[nocturne merlin]

yeah

here we go with the dripping tech

is there any way to use the old nano banana ?

[crisp rain]

I’m gonna be honest

It’s ridiculous how many of my prompts get blocked left and right

I can’t make battle or action scenes because nonsense keeps getting blocked

And yet Google wants people to pay and be able to edit their images and use it professionally lol

Good luck with that! I hope an open source model comes along and beats it at its own game

And I knew this was going to happen because no corporate model is going to let you freely make or edit what you want, they control the narrative the whole time

Complaining to them won’t get anyone anywhere either because it just won’t change anything. Period.

[nocturne merlin]

How much is the plan ?

[crisp rain]

You have to pay for API use

[nocturne merlin]

How much is that

[crisp rain]

You’d have to look it up

It’s a per model basis

Also the ‘safety’ settings on Ai Studio are a joke

Turning them off completely doesn’t do anything

[nocturne merlin]

Yeah

[crisp rain]

So it really is sadly just a novelty model

It makes neat things when it actually lets you make what you want

Other than that, it’s about as useless as a paperweight

People really shouldn’t have been surprised that most of their stuff couldn’t be made

[nocturne merlin]

Does that mean if you use API then you can use it full power ?

[crisp rain]

No

Why would the API version be uncensored?

If it was we’d all know about it by now

[nocturne merlin]

Then what the difference between free and paid plan ?

[crisp rain]

More uses

That’s it

[nocturne merlin]

Bruh

[crisp rain]

Compute costs money

That’s how it is

Unless you gen locally

[nocturne merlin]

I could've just change the account

[crisp rain]

Thankfully I do gen locally

[nocturne merlin]

Is that any better ?

[crisp rain]

I can’t stand these corporate guardrails and handholding nonsense

[crisp rain]

Go to Civitai lol

You’ll see

Open source will always be the way to go

No one has to be held hostage by these corporate models

People who work on making open source models will figure out a way to make something on par with nano banana

Just look at WAN 2.2 and what it does for video generation

That’s free and open

[nocturne merlin]

Hope so

[crisp rain]

Did you check Civitai out?

[nocturne merlin]

Yeah what about that

[crisp rain]

That’s all open source

Those models are all free to download

[nocturne merlin]

Yeah but my potato pc can't run that

[crisp rain]

Models, LoRA, etc.

[crisp rain]

Civitai lets you generate on their site as well

People that don’t have good PCs do that

[nocturne merlin]

But seem like there is no model as good as old nano banana

[crisp rain]

Not yet at least

It’s gonna happen

People always figure it out

Because we don’t want to be controlled

If I wanna make blood and violence and fantasy battle scenes then I’ll do it

If you wanna make a comic or graphic novel or manga that isn’t for kids then forget using nano banana

I couldn’t even make an image of my character using Kung fu to dodge bullets 😑

Like seriously ridiculous

[long hornet]

yes all this ai censorship is crazy

[crisp rain]

Exactly

It’s ridiculous

It’s all arbitrary nonsense

I’m a 36 year old man

You’re gonna tell me I can’t make bloody or violent art? F off

[dusk wave]

gotta update that bio then lol

[crisp rain]

I don’t pay attention to what that says

I haven’t edited in a while obviously

[dusk wave]

i'm going to assume less than 1 year since you edited it

[crisp rain]

Oh yeah

It says 35

I’ll be 37 next month on the 16th

[dusk wave]

better number imo, be proud

i'd rank 37 at 15 on the top 100 numbers list

[crisp rain]

🤣

But seriously, f these big corporations

The good news is that they make open source things work that much harder

[dusk wave]

haven't civitai themselves been getting hit pretty hard?

[crisp rain]

They have because of the credit card companies

Which is a whole other thing entirely

Which I won’t get into

People are trying to control what we can/can’t do in this world

[dusk wave]

yeah of course, just saying even they aren't immune to that crap

really sucks

[crisp rain]

The good news is that I have at least 10 TB of models and LoRa downloaded from Civitai

So I can gen until the end of time honestly

[dusk wave]

lol nice

you ever trained any yourself?

[crisp rain]

I just train LoRa of characters I like

[dusk wave]

i like doing art styles myself

[crisp rain]

I do those as well

[dusk wave]

did one on mw 2019, another on artwork from friends since they were curious

[crisp rain]

That’s awesome 😁

Yeah I got really pissed with nano banana earlier

[dusk wave]

understandable

[crisp rain]

LMFAOOOOO

[dusk wave]

tried using it just there, it wouldn't even edit an image that it did before

[crisp rain]

It wouldn’t let me make a Hong Kong rooftop chase scene of my character getting shot at

[dusk wave]

funny you say that, the edit from before was "remove the assault rifle"

and replace it with a box as you see there

[crisp rain]

It’s just ridiculous and I hate playing their games

I hate playing guess which words work and which don’t

[dusk wave]

i fired up a vpn to the states to see if that loosens things, haven't tried yet

heard that it's more strict in the eu, i usually connect to switzerland

[crisp rain]

If Google wants this to be a serious tool that people can legitimately use then they need to loosen the restrictions

[dusk wave]

a vpn is basically essential to me now because 🇬🇧🇬🇧🇬🇧

[crisp rain]

Yeah I’m not even gonna go into how ridiculous the UK law is

That is downright draconian

But America has its own issues, which I won’t go into 🤐

[fierce pond]

Things are weird here in the UK right now 😅

[crisp rain]

They want all of us to live in a police state

[fierce pond]

But even with all of the stuff you will see in the media, its not really close to a police state thankfully

Or civil war hahaha

I'm presuming that from the media, the external perspective is that we have no free speech, a nanny state and on the brink of civil unrest due to immigration

[crisp rain]

I miss the days when the common man would topple governments because the government thought they could control everything and everyone and people just got tired of it

French Revolution looking at you

[fierce pond]

A ruling power with a history of control in declining economies will always look to seek more means of retaining power, influence and money

[crisp rain]

Yep, sadly true

People have just become complacent

Eventually it’ll come to blows

It’s gonna take the government messing with the wrong people

[fierce pond]

True, in the UK its got to the point where the people have distanced themselves from the government

People dont really take notice, they're weak

But the propaganda is strong and still has an effect

[crisp rain]

The one good thing I’ll say about Our Illustrious Leader (sarcasm) is that he’s pro-ai

That’s the one good thing I like

[fierce pond]

Yeah for sure! He is a businessman after all

It brings its own pros and cons

[crisp rain]

The Democrats tried to really hobble ai

[fierce pond]

Saudi have had the best approach to AI that i have seen so far

[crisp rain]

Putting all women in hijabs or burkas?

[dusk wave]

he made illustrious? thank god for the us gov that's a banger model lol

[fierce pond]

Theres no forced hijab or burka rules in Saudi

[crisp rain]

I use Illustrious and NoobAi all the time 😊

[fierce pond]

But they have put AI onto the syllabus for 10 year olds and older

[crisp rain]

Well that’s good but…

[dusk wave]

i've been out of the local game for a bit personally, got screwed over by the retailer i bought my gpu from

really looking to try chroma once i can

[crisp rain]

Chroma is great

I sadly effed up my ComfyUI

Deleted all my models somehow

So it doesn’t start now

[dusk wave]

i'll probably do the same when i have to update lol

no updates since november 2024

that's going to be tough on it

[crisp rain]

Wow

Oh yeah haha

[dusk wave]

yeah it's been a long time

[crisp rain]

I am honestly just appalled and annoyed by the sheer amount of stuff that nano banana blocks

[dusk wave]

the thing that'll be even worse though? i run arch linux

updating that is going to be tough on the system

[crisp rain]

I don’t even know who the model is supposed to be for at this point

[dusk wave]

not even beige cube enthusiasts it seems

[crisp rain]

https://tenor.com/view/skatingby-sneaky-sneaking-sneak-gotcha-gif-7641348

[dusk wave]

been doing that a lot lately

what else is there when you can't generate knockoffs of the deep in modern warfare

[crisp rain]

Like I said, when nano banana works it’s amazing. It’s just that a lot of stuff doesn’t work

[dusk wave]

it's a very bizarre model to say the least

[crisp rain]

And I thought OpenAI was bad at blocking stuff!

They’re more open at this point

[dusk wave]

even when there's no filtering i've never seen anything more picky with prompts

[crisp rain]

At least GPT doesn’t block the amount that nano banana does

Unless it’s Pokémon…certain celebrities…D I S N E Y

You know, ironically, so many people can technically get in trouble for fan art since the characters are owned by the respective IP holders

Look at all the Patreon naughtiness and so many of those people could be sued

Especially for making money off of someone’s IP

[dusk wave]

i think i heard of a few pokemon artists getting c&d'd once

[crisp rain]

Yeah it’s honestly ridiculous in my opinion

[dusk wave]

funny considering it's not even nintendo at play there

[crisp rain]

It’s GameFreak

[dusk wave]

yeah exactly

[crisp rain]

And Nintendo backing them in the shadows 🥴

What a goofy world we live in lol

[dusk wave]

i wouldn't exactly say the shadows

[crisp rain]

No Nintendo is pretty front and center

They’re also stuck in their ways

Like Sony

I remember Sony purging games from the marketplace

But Nintendo funnily enough didn’t

It’s a back and forth thing

[dusk wave]

that i hadn't heard about

[crisp rain]

You know what’s hilarious?

Nano banana doesn’t know what restomod is

I tried making a car restomod style

The image I uploaded didn’t change at all 🤣

If only Imagen would let us upload and edit images

Imagen is one of my favorite models

[dusk wave]

it could just be poor textual description on its part, it's very obvious that it just makes a text prompt for the edits

like just try asking for anything deltarune, it'll end up like someone describing the characters purely verbally to an artist who is completely unfamiliar with it

[crisp rain]

Meanwhile Imagen

[long hornet]

is there any good image edit/generate that is less on the guard rails

[crisp rain]

Illustrious and NoobAi

Anything open source for the most part

Just not on the same level as nano banana yet

But eventually

The open source people always catch up

[dusk wave]

they're definitely not decent as edit models though

[crisp rain]

But Illustrious and NoobAi models are hands down the best anime models right out of the box

[dusk wave]

there's flux kontext and more recently qwen image though of course

[crisp rain]

Qwen Image sadly messes up a lot

It’s uncensored but a lot of stuff comes out goofy

[dusk wave]

i've noticed that, but there's barely any models in the edit space

[crisp rain]

Someone’s gonna try to distill nano banana

[dusk wave]

i wonder if maybe chinese prompts work better, ali said the same with wan

[long hornet]

then why is qwen img edit higher than flux kontext dev artifical analysis

[crisp rain]

You’re not seeing the images my buddy sends me lol

He uses Qwen Image to make uncensored stuff

I’ve made better things with Illustrious and NoobAi models

[long hornet]

alr

[dusk wave]

could always just fire up a side by side on lmarena and see which one you prefer personally

since both are on there

[long hornet]

true will look into it

[crisp rain]

But uncensored only works for those that can gen locally

Always remember that

Or if you can afford to gen on Civitai

[long hornet]

what specs needed

[crisp rain]

Umm

Honestly?

A 3090 works amazingly

The more VRAM the better

I have a 4080 RTX