images-canvas

[severe crater]

Those were fun

[wraith arrow]

those were the days

[severe crater]

the studying 2 cracked me up

The cat is astonished at the lack of intelligence

[wraith arrow]

obviously

[severe crater]

i love comedy

Mont Saint-Michel

Giza

Parthanon

[cerulean apex]

this one is truly epic! ❤️

[severe crater]

pyramid of El Castillo

Saint Basils

Ziggaraut of Ur

charactures

The Taj came out lovely

Game Cover art 1

2

3

which one are you playing, art alone

front and back on the box

[wraith arrow]

for mont-saint-michel sure but there's some flooding

The Picture of Dorian Gray

[woeful bough]

happy holi

[severe crater]

@wraith arrow

[severe crater]

a good day

[mint owl]

Happy Birthday holi

[severe crater]

The TwinOracles got updated.

[wraith arrow]

New theme sounds fascinating

[placid oar]

enjoy it!

[wraith arrow]

Ty you too!

[placid oar]

I am, already got cool images in queue

[wraith arrow]

Nice.

[placid oar]

not sure if I should have posted this one instead

[severe crater]

What's up B?

[mystic mango]

damn, missed the summit theme

[severe crater]

awww

[wraith arrow]

wasabi!

[severe crater]

Wassaaaaaap

hahaha

[wraith arrow]

he's making wasabi

Mr Coffee

[marble stump]

Really like this one. Those pipes are laid out so seemingly purposefully 😄

[placid oar]

getting really cool stuff

[marble stump]

You have discovered such good visual styles 😮

[placid oar]

yeah lol

[marble stump]

To me this looks like part disney classic and part modern digital concept art.

[placid oar]

it's somthing along those lines tbh

still exploring it because I still need some adjustments before I can add it to my documentation

hand looks weird

[gentle wagon]

An enchanted forest where neon fireflies illuminate the night sky, casting a surreal glow on the ancient trees that whisper secrets of bygone eras to those who dare to listen

[wintry delta]

Adobe Photoshop and Illustrator icons relaxing after a long day and talking each other, white background, highly detailed images.

[severe crater]

thats what im talkin about. simple puns

[wintry delta]

Adobe Photoshop and Illustrator icons relaxing after a long day and talking each other, white background, highly detailed images. | dall's-canvas

[severe crater]

give the in dalle thumbs down and regen on those for proper text if you're contributing to database

[placid oar]

do something with a PET bottle

[severe crater]

polyethylene bottle?

[placid oar]

with lots of sugar inside

you know, keeping the world diabetes free and the ecosystem plastics free

[severe crater]

ah, you want a name brand test?

[placid oar]

doesn't have to be dr pepper

unless you do an actual Doctor being a Pepper

[severe crater]

prompt: A polyethylene(PET) soda bottle(clear) labeled "Fizz" sitting on a picnic table

[placid oar]

lol

[severe crater]

prompt: A polyethylene(PET) soda bottle(clear) labeled "Sugar Water 'lite'" sitting on a picnic table

prompt: A polyethylene(PET) soda bottle(green) labeled "Sit-Up'" sitting on a picnic table

Prompt: A 2-liter polyethylene(PET) soda bottle(green) labeled "Bubbly" sitting on a picnic table

prompt: NOW A cup of yogurt labeled "So-Hurt"

[placid oar]

tried to do monty python like alchemy, it didn't turn out good...

[placid oar]

please keep in mind, images considered scary for a G setting should be marked with spoiler. other than than that, I'm scared

you can repost the image, then it's all ok

[tawdry sonnet]

Alright ill fix that quickly with spoiler on

[placid oar]

can't mark images as spoiler once they are uploaded

[tawdry sonnet]

What have i created lol

[tawdry sonnet]

I am also scared

I also have this one, which i kind of think is actually pretty symbolic for my struggles 😂

[placid oar]

that looks like something I would not want near or under my bed

[tawdry sonnet]

Hmm, yeah, i agree with you on that one lol

[placid oar]

that embodies more than anything my "monster under the bed"

Monster under the bed is where Copilot Pro draws the line

[tawdry sonnet]

I actually quite like the "old oil painting" style

[tawdry sonnet]

Lol, i see copilot draws a hard line

[wraith arrow]

Dupe

[severe crater]

@wraith arrow

i forget what i share

[wraith arrow]

Even with an IQ of 168?

[wraith arrow]

tron-ish

[placid oar]

yeah

even tho I wasn't going for that

and just took your motorcycle as reference as I'm out of ideas currently

there we go

[wraith arrow]

ooh

purty

[placid oar]

and this is what you get when you add "glowing sprinkles" ...

[wraith arrow]

noted for future reference

[placid oar]

lol, I'm not really creative today. I'm just fooling around

[placid oar]

this is def a sign, penalty box on OAI and this on Copilot

i'm gonna stop doing images for today

[wraith arrow]

a sign of something new?

[placid oar]

yeah

or a sign I should take a breather and give room for resting

[wraith arrow]

lower levels start at just $1m/month, call for penthouse specials

[placid oar]

$4/month gets a window?

[wraith arrow]

$4 won't get you a storage closet in that building

[placid oar]

lol

there goes my hopes and dreams for an apartment with a windows in a filled with smog area

[wraith arrow]

in this building you look down at it

dall-e struggles with spiral staircases

[steep oyster]

Showcase your DALL·E creations to the world!

Collaborate with our OpenAI Instagram page! Just invite @openai as a collaborator before you post. If selected by our team, your work will be featured on our handle, giving you more visibility.

Quick tips for selection:

• No copyrighted or profane content.
• Your profile needs to be public.
• Stay mindful of sensitive topics.
• Outputs that closely follow the prompt are preferred.

Feel free to share reels, carousels, or just a single image!

[severe crater]

The constellations of meow.

Theres Dontchu Doit The mischievous sign. Then Wherjugo The Coward. Ah, Twopets Nomore the picky sign.

[brisk root]

DALL-E gave me air head

[brisk root]

wow

[placid oar]

wow good or wow bad? depends on your prompt and expectations

[brisk root]

Wow good. A style I've never seen in DALL-E.

[placid oar]

it does look cool

I'm a bit tired so my textThroughBrain model is kinda slow today

[velvet solar]

#art-and-illustrations "In an old, abandoned mansion, strange noises have been heard echoing through the halls. As a group of friends decides to spend the night there, they uncover the chilling truth behind the mansion's haunting past."

[cerulean apex]

If you are trying to generate image head out to #image-bot

[velvet solar]

How to generate free images without limitations? Please help me plz plz plz🙏 🙏

[placid oar]

you can do 5 images per day for free, limitless on #image-bot

[velvet solar]

@placid oar Thank you so much!

[placid oar]

thank OAI, I'm just pointing you there

[paper lake]

Can you create a 11th grade bridge prototype image?

[placid oar]

dall-e doesn't work like that

you have to pass the prompt to the model, you can try using /draw on #image-bot

[undone pier]

very beautiful

[marble stump]

Pretty strange generation 😄

[brisk root]

Uhm. What is gasoline pump doing inside the supermarket lmao

[placid oar]

I'm off for the big weekend. Have a great time everyone!!! I'll be back next week.

[rapid carbon]

An I fashionably late?

[blazing frigate]

An expressive oil painting of a basketball player dunking, depicted as an explosion of a nebula.

[olive plover]

Hey! You can try DALL·E over in #image-bot with the /draw command

[blazing frigate]

Thanks mate

[thorny walrus]

Happy Easter

[wraith arrow]

this theme tests the limits of our imaginations

[outer hinge]

Well, you see how fast that person's moving, they're a blur!

In the distant future, everything's electric powered, and gas pump-like displays have come back into style, it's very retro and a reminder of how much better things are now.

The pump selectors dispense different flavored safe-to-consume beverages, there's cups of various sizes on the other side of the machine, off what we can see. Notice that there's a hose for each dispenser, to help keep the flavors separate.

Story about it here: #1070954956963721216 message

[hybrid spire]

its for people who drive into stores

[mystic mango]

Stardew pillow

[burnt stone]

Hii

[dark locust]

#how to convert video to animation

[prisma panther]

Ayo this is cool, how did you make this?

[mystic mango]

haha, nothing to special, just a simple prompt and some luck, I asked ChatGPT Please illustrate a girl with short brown hair and tired eyes waving in hello. "Yo!" in a speech bubble by her head. A webcomic style with slight anime aesthetics. A reaction image.

[prisma panther]

Wait, so if I just go on ChatGPT.com, it can do it?

[mystic mango]

well, you have to pay a subscription to use GPT4/Dall-E 3 with ChatGPT, but there's other ways to make it for free, like Bing Image Creator...

[prisma panther]

I see, is that option not that great or somethin?

[mystic mango]

I think it's fine, You might be more limited in how many generations you can make in a day/week/month. I'm not too familiar, but the images should be the same quality

[prisma panther]

Oh I see, if the fact that websites or just image generators place a limit on users, in a way remembering them if I'm understanding how that works right

[mystic mango]

if I understand what you are saying, then you're correct! Most AI image generation services do limit and track usage per user

[prisma panther]

I really, really hope some AI can remember everything...

[mystic mango]

@junior spruce

[mystic mango]

when you run out of generations...

[bright arrow]

cute

[trim nest]

[bright arrow]

[fallow glen]

Sup everyone. What has been your experience with the new in-painting feature?

mine so far has been... funny

[placid oar]

something particular on how to use it? or is it being rolled out slowly?

[outer hinge]

I don't seem to have it yet either, but details here: https://help.openai.com/en/articles/9055440-editing-your-images-with-dall-e

[fallow glen]

dalle GPT

[outer hinge]

Ooh, I didn't think to check the Dall-E GPT itself. However, even that, I don't have the in-painting features. I still have the thumbs up/down instead.

[fallow glen]

does it show when you click on the picture?

[outer hinge]

This is what I get:

[fallow glen]

hmm, well. seems like they are rolling this out. You can try loggin out and in again, sometime it triggers new features on-boarding

[outer hinge]

Thanks! Yeah, I was auto-logged out earlier this evening, and logging back in, it's still the same. When it is my turn, I'll explore!

[cerulean apex]

Avelible in EU that dalle thingi ?

[cerulean apex]

oh f i think i don't got it sadly

[brave marsh]

Dogs of war...

[analog pier]

Cats in bed

[brave marsh]

I got automuted for cats in bed, but dogs are ok I guess 🙄

No automute this time. Weird.

[placid oar]

what matters is tha the kittens are safe!

[brave marsh]

Tiny kittens engaged in purrracy.

[placid oar]

dall-e really overdid it with the zoom in

[jovial aurora]

hullo all

[placid oar]

heyo

[lucid lotus]

May or may not have used the prompt from a Sora example 🤫

[mystic mango]

Sadly pointillism feels so inconsistent

[placid oar]

Think it's now getting consistent with gens

[junior spruce]

@mystic mango

[mystic mango]

InPainting coming in clutch I see 😉

[junior spruce]

Well, you can certainly see how it helped there. 😅 But I'm not sure I'm feeling it. I may have to learn to work with it a little.

[mystic mango]

Nah, I don't think it's you, it definitely needs some fine tuning in its end.

[junior spruce]

For sure. I think it's using the whole prompt in some way, instead of some smaller, more focused prompt for the edits. Adding in or changing stuff hasn't worked to well so far for me, but I only got it today, so we'll see how things proceed.

GPT also talked about the most recent generation as if it was the one you were working on, that was kinda odd.

I tried to give it the prompt to what I was working on and it decided to fuse the prompts...

[mystic mango]

Daily image cap 😭

[placid oar]

Image enhanced with the new inpaint

[frigid meteor]

The editor in the inpaint function works for enhancing a whole image, kinda, it's trying, might tune itself after awhile, a week maybe.

[placid oar]

Love that sword, so shiny and colorful.

[frigid meteor]

Thanks master of dall-e @placid oar , well thanks to the new edit function and dall-e lol.

[trim nest]

I agree, this is an awesome pic!

[placid oar]

I'm no master, just learning

[frigid meteor]

Of course, there's you just learning in your workshop lol.

[placid oar]

Example, I have a project that has about 25-ish files. Because I didn't study enough, when I push to repository, the project generated around 1500-ish delta files and screwed all references.

[placid oar]

oh no, that's too much to do, I will move in there tho. Looks interesting to play with

[frigid meteor]

LOL

[placid oar]

This is interesting, looking into my credentials activity from VSCode, people are tring to get into my account from Saudi Arabia, Jordan and Philippines apperantly

This is me atm, checking security properly right now.

[mystic mango]

I don't think I'll have the chance to post this before the theme changes.

[sour geyser]

It will change in 44 minutes - so you will?

[mystic mango]

lol well that's fair I never knew the exact time

[sour geyser]

(For my time zone) exactly 7 p.m.

[mystic mango]

so 10am for me. I'm always in a meeting so that's why I miss it 😭

[sour geyser]

If you want I can send it for you and add to my message that it's by you.. but yea.. 🙈

[mystic mango]

I appreciate the thought 🙂 but i'll focus on preparing for the next theme.

[sour geyser]

Have fun! 😇

[placid oar]

wow, that was some monumental work there, but finally was able to reset my visual studio code profile for dall-e

[marble stump]

Figured this was a pretty neat concept and composition for vertical format.

[trim nest]

Awesome 😎

[wet token]

awwwwww

[wet token]

lol

[prime bloom]

Can you send me the prompt that u used for this sticker pls?

[mystic mango]

Absolutely, I had to use bing image creator since I was out of chatgpt generations. And the prompt was please make an image of a girl with short brown hair and tired brown eyes in despair. exaggerated facial expression. cartoon style with slight anime aesthetics. reaction image

Nothing too fancy

[prime bloom]

Ah, thanks! I expected something more complex tho 😭🙌

[prime bloom]

"Reaction image", well seen! I would never have thought of that!

[mystic mango]

Haha I can't guarantee it's effectiveness but I feel like it has worked for me, maybe as a placebo. 😂 it can be tough to say sometimes

[proven plume]

The artist who made the origenal for me that I fed in DALL-E is the reason I am now using AI art her work was not even finnished look what we got now

colored it in ❤️

[wet token]

damn

thats not the same picture

[proven plume]

its not the AI got creative

[mystic mango]

I like how this one turned out except...

...for this part

[placid oar]

DALL-E does good enhancements. I'm not that good looking for an anime like this

[placid oar]

If I'm gonna throw myself into the mix, might aswel start meeting my creations no? #daily-theme message

Too bad hit my daily cap for now

I still got copilot but not reliable concepts with that

[placid oar]

This is... weird....

but now I know the model is capable of mixing stuff

[mystic mango]

i do like when it manages to mix styles like that

[placid oar]

saved the prompt to analyze it later

I've been wondering how to do stuff like roger rabbit

[mystic mango]

I've had some luck with MIdjourney, but I haven't tried with Dall-e 3 much

it is definitely trying to merge the two styles

[placid oar]

Yeah it's trying, copilot is trying to adapt between the two styles

[mystic mango]

ooh that ones pretty good

[placid oar]

hmmm, seems I got something going on, but that's copilot lazy prompting

I have to do it tomorrow in a proper way with dall-e

In a sun-drenched skatepark, a real-life teenage boy is caught in the midst of a skateboarding trick, his concentration tangible. Just off to the side, an anime girl with a distinct, colorful style is cheering him on. She's standing firmly on the ground, her presence adding a touch of whimsy to the scene. The two are depicted in starkly different styles, yet they share the same space, creating a unique and engaging visual narrative.

[placid oar]

This one turned better. but seems the size is gonna be a great issue

size awareness seems to be off when blending both art styles

also spacial placement is gonna be a thing

[wraith arrow]

Placing animation in photorealism is a neat concept, reminds me of Roger Rabbit, the most difficult movie ever made for its time.

[placid oar]

that's exactly the motivation

And also

[plucky jasper]

The great Qiao of the Romance of the Three Kingdoms, with a beautiful and unparalleled appearance, is unparalleled in terms of beauty, beauty, and beauty, making him an unparalleled beauty

[wraith arrow]

Fascinating

[boreal zodiac]

画兰花

[outer gazelle]

holla

[fallow glen]

hi

[wraith arrow]

Yo

[trim nest]

Hi Shon!

[trim nest]

Caelumsonus, the God of Sound

Prompt used: ```Envision a deity, Caelumsonus, the God of Sound, harmoniously blending the grandeur of Roman and Greek mythology into an ethereal figure of awe. Caelumsonus resides in an otherworldly realm where the ether vibrates with the essence of sound, each note shaping the fabric of reality.

Description:
Caelumsonus stands tall, embodying both the stoic power of Roman deities and the ethereal beauty of Greek gods. His physique, a perfect amalgamation of strength and grace, is draped in a flowing toga that shimmers with the hues of twilight, interwoven with threads of gold that catch the light, symbolizing the reach of sound from the whispers of dawn to the echoes of dusk. Around his shoulders, a cloak of midnight blue billows, dotted with silver threads that mimic the stars, a nod to the belief that the gods communicate through the music of the cosmos. His hair, dark as the void between stars, cascades down his shoulders, with a laurel wreath of golden lyre and flute keys crowning his brow, signifying his mastery over all forms of sound.

In one hand, Caelumsonus holds a lyre, its strings made from the very essence of light, capable of producing sounds that can soothe the fiercest storms or awaken the slumbering mountains. His other hand gestures subtly towards the horizon, where the line between the celestial and the terrestrial blurs, inviting the resonance of his divine melodies to permeate all creation.

Scenario:
Caelumsonus's sanctuary, the Aetheric Concert Hall, is nestled atop Mount Harmonia, where the earthly meets the divine. This magnificent structure, inspired by the Pantheon with its oculus open to the heavens, allows the light of the gods to illuminate its interior, creating a natural spotlight that shifts with the time of day. The hall is surrounded by a forest of cypress and olive trees, their leaves whispering hymns of praise in the gentle breeze, while a stream, infused with his power, murmurs melodies as it winds through the landscape.

Background:
The realm of Caelumsonus is a dimension where sound influences creation and emotion. It's a place where the chorus of the universe converges, from the softest rustle of grass to the majestic symphonies of the spheres. Here, the God of Sound teaches that harmony is not only a principle of music but of life itself, encouraging understanding and unity among those who seek his wisdom. Through his melodies, Caelumsonus weaves the tapestry of existence, embedding the essence of sound into the core of all beings, inviting them to listen to the music that exists within themselves and in the world around them, thus maintaining the balance of the cosmos.```

[trim nest]

Silver Butterfly-Dragon

[placid oar]

Love how the style for these images turned out.

but.. new theme

[placid oar]

This was also an option, opted for the other one in daily theme

[wise raft]

"Make an image for a Japanese kenjutsu dojo logo"

[prime bloom]

Looks like the lonely tree music

[placid oar]

You will have to elaborate, don't know what you mean by that reference.

[prime bloom]

You've probably heard it on tiktok

[placid oar]

What's them tiktaking thing?

I don't have a tik tok account lol, I barely use social media. I had myspace

[prime bloom]

Wdym my space?

[placid oar]

pre-facebook era

[prime bloom]

Oh ok

[mystic mango]

I know I'm old and it's overdone to say. but it never gets less painful to see something like Myspace not get recognized.

[placid oar]

Also, love her!

[mystic mango]

I'm glad you do!

Lol wish the app I used to remove backgrounds worked better

[placid oar]

But I want to play Uno

more fun with friends

Nothing like a peaceful game of UNO

[mystic mango]

Be careful Uno can end friendships 😮

[fallow glen]

New challenge: removing lights from a scene. It's weird, but dalle3 seem to always create a source of light to illuminate the focal point in the picture. For example, we can try to describe a completely dark room, dalle will always find a way to sneak a source of light in the room.

even the "blackhole room" is illuminated by the blackhole 🤣

[placid oar]

DALL-E OAI or DALL-E Copilot?

[fallow glen]

I tried with the API (openAI)

[placid oar]

yeah, still azure services implenet dall-e different, MS DALL-E has a strong fear against dark places

[fallow glen]

[placid oar]

and I mean really scared

[wraith arrow]

Everyone is afraid of the dark.

[placid oar]

not me, I use Dark Theme with VCode all the time

[fallow glen]

but it's not even a question of content policy nonsense or the revised_prompt including a light. really, even when the revised_prompt describe a scene with no light, dalle3 will add a light in the background no matter what

[placid oar]

no, I was testing it because I was having light incoherence with prompts

made a custom copilot for work testing with copilot studio and added GPT with a slight temp.

images were always well lit even tho I never said anything about light

but then again, the way MS wants those things to work togethere is messed up, they are decoupled completly

their vision is to reduce use of GPT tokens by implementing their Power Virtual Agents first as a first contact

when I get back to the office tomorrow I'll grab some examples

[placid oar]

I don't like something about this image, but I don't know what it is... something is not right and I can't figure it out

[hidden goblet]

A vibrant, colourful grove in an otherwise barren forest. A large, bent, crudely-constructed shack is nestled within, the smoke from its stone chimney wafting into the cool air. Trees, their branches contorted and embellished with purple foliage, envelop the area and obstruct the faint luminescence of the moon. The image is a piece of digital concept art depicting an abstract scene, rendered with broad brushstrokes.

[wraith arrow]

Let me know what you guys think?

[placid oar]

It's a futuristic (Bio-)mechanical Armor of what I could say it's a Ninja/Samurai! Neat!

[wraith arrow]

Thanks

[placid oar]

Here's my take, just with the words I gave you and added dynamic scene to the mix.

Hope you like it

[wraith arrow]

Pretty cool

[wraith arrow]

How many Keywords did u use?

[placid oar]

This was my prompt: A dynamic image of a futuristic (Bio-)mechanical Armor of a Ninja. With high tech features and nano-materials that adapt quicky to any action.

[wraith arrow]

Whoah

Thats cool

[placid oar]

It's just what I saw, nothing else

It's your idea

[wraith arrow]

Have u heard of anyone selling AI generated Images as NFTS?

[placid oar]

Not with AI images from OAI at least. Would be hard to do.

And to be fair, I never got into NTFS, so I haven't looked willingly

[wraith arrow]

Alright alright

[placid oar]

But in the meantime I did this:

[placid oar]

Really cool

[wraith arrow]

Thank you so much

[placid oar]

That's more of the bulcky warrior path. I went more for the sleek ninja variant

[wraith arrow]

Haha yeah

[placid oar]

A good change, considering I've been doing other type of images today

[wraith arrow]

Do u have specific focus

The images i generated above i manage to locked it as base generator

[placid oar]

Here in the community, mostly anime inspired scenes

[wraith arrow]

Which means i able to generate the same variant but customize them into holding a rifle and various others

[placid oar]

for private use, other concepts

[wraith arrow]

I see

Do the community sells their AI generate to make money?

[placid oar]

You can do with your images what you see fit. It's the the terms of use

but activly here, no, people don't promote that

[wraith arrow]

Ohh

[placid oar]

self-promotion is against the rules of the server

This kind of images has been for me today: #daily-theme message

[wraith arrow]

Very cute AI generate

Where did u generate your images?

Directly on OPEN AI?

[placid oar]

all with DALL-E

yes

[wraith arrow]

Can u send me a linnk to the site:

Im using microsoft co pilot as of now

[placid oar]

chat.openai.com

[wraith arrow]

Is it free?

[placid oar]

you have to be plus member to use it

you can use 5 gens a day on #image-bot with the command /draw

[wraith arrow]

Then no need then i will stick with co pilot

[placid oar]

In the dall-e channels we talk mostly about the things we can do or find or could be with the DALL-E model, Copilot is also a DALL-E model, so yeah, feel free to share your creations

[wraith arrow]

Will do and thanks

[frigid meteor]

@placid oar workshop lol.

[placid oar]

Why do you want me there?

Here's another version

@frigid meteorI only need one keyboard and a large screen.

and food!

Here are some other variations

Btw, all this images are story driven

The story began with this image, and then I started wondering, why is she living there, what drove her to live at the top of the mountain.

[frigid meteor]

That's one from your daily theme it's really good

[placid oar]

yeah, glad you noticed ❤️

The prompt I can share, the art style is a secret I will keep forever!!

[frigid meteor]

Keep both your prompt your prompt will just end up in one of the prompt stores

[placid oar]

lol

[placid oar]

@frigid meteorquick think of one kind of person, one verb, one adjective and a number from one to 100!

I wanna try something

[placid oar]

Now we can post 6 images at once? I thought the limit was 5

[frigid meteor]

Marvin Martian, running, scared, 38, 🛸

[frigid meteor]

LOL

[placid oar]

I passed your words through my NLP and then gave the prompt to DALL-E

Here's a better version.

[frigid meteor]

So @placid oar you're making use of Visual Studio creating custom language processing apps

[placid oar]

Well, kinda yeah

I use vcode mostly to work with code and then use it where I need it implemented.

[frigid meteor]

Cool Master Dev @placid oar lol, that is cool though 🧠

[placid oar]

Nah, hobbyist lol

I do have to say, to write cool stuff with code, there are truly great engineers and math ppl.

[frigid meteor]

Same here, slobbyest I am lol.

[placid oar]

I don't think I could do it for a living tho

[frigid meteor]

Not sure to many do now, the ones with insight get preyed upon by the ones that take credit for everyone else's actual smarts.

[placid oar]

hehe, that can be true

unless you are the coder and the smart one

[frigid meteor]

lol, and stay under a rock.

[placid oar]

hehe, dunno, probably a good profile on LInkedIn or similar

@shut sparrow When are you selling me this #daily-theme message ? Love the house.

Here's my take on your prompt. Some people (Hint: maybe me!) are there to buy the property.

I'm loving how endearing this concept is working.

[shut sparrow]

Haha, I'd 100% live here as well!

[placid oar]

I was thinking that when using DALL-E the frontend on the Chat should have a responsive design for wide/ultrawide images. Would be really cool to use the space to see the images and also prompt accordingly.

[placid oar]

I made a small suggestion, I hope this has some use for some of us https://discord.com/channels/974519864045756446/1226506750673555486

Shouldn't be hard to implement and less clicking on images to navigate features

[frigid meteor]

@placid oar created new rendition of the Antikythera mechanism.

[placid oar]

lol, I would love to, that artifact is so intriguing

There's also the water clock from the swiss army, that thing still puzzles me til this day.

[frigid meteor]

LOL

[placid oar]

I'm gonna take a break from image generation for a bit, until the next daily theme, think after doing so many images today I'm about to him my daily limit

[placid oar]

No wonder it looks so low def, if you are running DALL-E on that

[brisk root]

I have time travelled to 95 and created DALL-E

[placid oar]

some of the anime masters are on the loose

Strange inpainting, but I kind of like it how it went

I wanted to change only a clothing accesory, replaced the whole image

[fallow glen]

I'm unsure if inpainting use natural style or dalle2

but we can see it's not the normal dalle3. when we select the entire picture with the inpainting tool, and ask it to generate something, the result is low quality. Like dalle2 and natural style

[placid oar]

It replicates the API behavior with a snapshot of the original image and then the masked one.

[fallow glen]

You mean that dalle2 API route?

[placid oar]

Yes!

It's probably adapted for the chat, but the principle would be that.

[fallow glen]

right. It's what it looks and feel like. It's simply mapped on the dalle2 endpoint. It's weird though. Does that mean dalle3 is unable to perform those operations? That's so strange.

but dalle2 was unable to make large images afaik

[placid oar]

or it's a DALL-E 3 adapted for inpaiting as evolution of DALL-E 2

[fallow glen]

I mean... this is not dalle3 quality

That's more like dalle3

[placid oar]

The snapshot we see, and I only assume, I haven't fact checked, is not what gets sent to DALL-E. Please correct me if I'm wrong. It's probably a resized enclosing crop of the mask we generated.

[fallow glen]

there is no way to fact check this completely, openAI didn't document what happens under the hood when we use inpainting. They just kept it vague in the announcement. The inpainting is available only for dalle2 with the API also. I think I remember they mentioned something about inpainting and images variation in the technical report thing

[placid oar]

Ya, i know, hence the assumption

[fallow glen]

I'm curious though, I'll go check

https://cdn.openai.com/papers/dall-e-3.pdf
https://cdn.openai.com/papers/DALL_E_3_System_Card.pdf
(hope I won't get muted)

phew

[placid oar]

you are a daredevil

[fallow glen]

for science

I can't find anything about inpainting though... hmm

[placid oar]

same

Don't see anything in particular

[fallow glen]

interesting that the dalle3 paper references sdxl

[placid oar]

ok, that is a lead

[fallow glen]

lol, well... it's just nice that they mention the reference. In science, it's normal to use all the knowledge available, and it's good that they don't hide it. It's just sad that they are not making real papers anymore... Unless really there is really nothing more than better captioning and scaling the training data + parameter counts behind dalle-3. But then, it doesn't explain why dalle-3 can't do in-painting (or why it's so bad at it). Even dalle-1 could do it (even though it wasn't planned, it was an emmergent capability)

[placid oar]

What I'm saying it's just a hunch tbh, I really haven't delved into the real of text-image2image models

[fallow glen]

don't worry, it's the same for me. I would not pretend I known the absolute truth. I'm sharing what I observed

[placid oar]

I'm wondering if it's a similar case when uploading an image if it gets resampled and that's what is provided to the model.

passing it to the vision model

would account for some error on more in-depth details missing sometimes

[fallow glen]

I did a break down of this a moment ago. I'll see if I can find where I posted it. (but long story short, it seem to be the same idea: the image is split in a grid of fixed-size squares, those serve as tokens for the vision model)

[placid oar]

sounds like a good lead

[fallow glen]

[dalle3 & gpt4 interoperability](#off-topic message)
well, according to the dalle3 papers, the dalle3 and GPT-4V model treats image and text the same way, as part of the same stream of tokens. an image sent to GPT-4 is subdivided and each portion of image represent a distinct token. It was trained on text and images mixed without distinction, basically.

Dalle was trained to output image tokens while GPT-4V was trained to output text tokens. But both can technically understand images and text.

(in the context of a discussion about GPT-4 tokenization)
yea, dalle interpretation of concepts is a whole different world... I performed countless experiments and continue to perform more. understanding how it process the information is just as difficult as it is with pure text... I just found that article that explains the maths behind the training, very interesting stuff.

I spent the past 40 minutes reading papers to make sure what I'll say isn't 100% dumb. How I can relate the infinitely complex "reasoning" process of GPT4 with the infinitely complex "representation" process of dalle3?... I don't even know how to describe the metaphor, and it's not useful anyway.

I'll just use a dumb simplification: Dalle3 is the reverse process of what GPT-4V does when it describe a picture..

That's oversimplified but I believe that's a good way to think about it. Now, assuming GPT-4V and Dall-E 3 use maybe the same tokenizer (that claim is based on nothing, just a random guess based on the papers I read, but I don't remember it being mentioned anywhere).

(see the picture)

that's DALL-E 1 paper btw. We know DALL-E 3 use a different encoder. But we also know GPT-4 (and maybe GPT3.5) were used a lot in the process of training DALL-E 3.

I don't think I can send the link to the article here. but it's a very nice breakdown of how dalle2 works. I can send in DM if needed

[placid oar]

I will in due time, but for now I think I will be calling it a night. I'm gettins really sleepy, only slept 3 hours this weekend.

[fallow glen]

sounds good.

[placid oar]

And at a good time going to sleep too

[wraith arrow]

Maximum 6 haha i tried

[fallow glen]

right, I never remember what the max is. I know the max allowed by Discord is 10, but this server implements its own undocumented rules

[amber coral]

These are under the same prompt

[lunar heron]

who is similar?

[wraith arrow]

Funniest AI design by far

[brisk root]

Telescope images of planets 🪐 🌠

Do you remember this 70s movie?

[wraith arrow]

@placid oar hey

Check this out

Cute huh?

[proven plume]

Heaven's gate keeper I know I tweeter her already but she belongs here too!

[amber coral]

I had a dream I woke up on the beach of a beautiful city. The floors and walls paved with limestone with beautiful textures, the water pure and blue. There was a giant amphitheater and many structures in the back, but I always found myself coming back to the waters, sick with its beauty. Below the land in the rocks, I found a chest full of clothes, a book, and treasure. The items were all things that I owned, seemingly lost.

[placid oar]

Was gonna post this in the daily theme: "The Dissonant Maestro" but the time is up tor a new post.

[placid oar]

And here I landed, in FF15

[placid oar]

Was getting more final fantasy 15 imags...

Had to change part of my concept to avoid that

Still nice to know it can be done

[placid oar]

Narrative was going so well

[amber coral]

Neo babylon

[placid oar]

@prime bloom here are some gens, marked as spoiler for obvious reasons

[prime bloom]

Can you make the brushstrokes lsigtly more visible pls?

[prime bloom]

Also u can replace obese by fat (i just used this word to make thicker muscles)

[prime bloom]

can u share the prompt pls? @placid oar

[placid oar]

The prompt: A monstrous, large wolf, its fur a dark, ominous black, muscles bulging with power, strides through a grim, dark bottomland. The wolf's movement is captured in an strong motion blur, suggesting unstoppable force. The scene exudes a grainy, vintage feel addingdepth to the composition but the secret sauce is mine to keep

I'm surprised it wasn't flagged as scary content on both OAI and Copilot's filter

[prime bloom]

OpenAI wimp

[placid oar]

no, I mean, GPT in general will refuse to do scary content and flag is

[prime bloom]

Yes, OpenAI wimp

[placid oar]

I klike this big chunky one,

I guess it's safe to call it quits on the daily theme and wait for the technique for today's theme

[prime bloom]

me too

[placid oar]

Some images that were good but won't go to the theme channel

[steep oyster]

Showcase your DALL·E creations to the world!

Collaborate with our OpenAI Instagram page! Just invite @openai as a collaborator before you post. If selected by our team, your work will be featured on our handle, giving you more visibility.

Quick tips for selection:

• No copyrighted or profane content.
• Your profile needs to be public.
• Stay mindful of sensitive topics.
• Outputs that closely follow the prompt are preferred.

Feel free to share reels, carousels, or just a single image!

[placid oar]

This was a candidate for today's theme

We need to get more o the anime pros

I have found my next image

[mystic mango]

I

[placid oar]

there, my next one landed

[mystic mango]

That's quite creative, good idea

[placid oar]

Mentionable honors

damn, being an anime girl creator sure takes 200% of the time, I can't imagen how much the anime artists really work for their personal projects

[mystic mango]

Ooh that first one may have been my pick but you can't go wrong with any of em

[placid oar]

yeah, she's really in a good mood to provide good images today

probably because I added to the prompt "casual day"

removed casual day and this happened

[placid oar]

Here's my take just from what I see

[placid oar]

This time I only had one good image to provide

[mystic mango]

It is a great image to be sure

[placid oar]

found a neat way to play with bokeh effect

the problem is, if you word it wrong, the bokeh effect is too strong and overpowrs everything

[tawny kraken]

sauce please 🙏

[placid oar]

dang it, foiled again, got a batch of great images

[mystic mango]

I'm interested, are you willing to share your secret? If I'm honest the amount of images in the daily theme that don't actually have any bokeh is disheartening

[mystic mango]

A cartoon style illustration with slight anime aesthetics of a girl with short brown hair and tired brown eyes. She is depicted happily walking on a nice spring day in the Pacific Northwest (PNW), surrounded by lush green trees and flowers typical of the region. The illustration features a dynamic pose to convey movement and joy. The color scheme is flat, with no shading, and the lines are bold and clean. Wide aspect ratio.

Here's what ChatGPT came up with

[placid oar]

here's a preview

[mystic mango]

Reminds me of the video game Journey where the sand shimmered like that

[placid oar]

hehe, I'd have to know whcih video game it is tho

ok, here's a small hint to my today's approach, remember with great bokeh comes too much power

volumetric bokeh effect

[mystic mango]

I appreciate the tip. I will give it a try

[placid oar]

I do want an image as payment, and add lots of creative stuff in it

Here's another preview of what's coming

[placid oar]

@mystic mango for your consideration #daily-theme message

[mystic mango]

Ooh I love them all, I think the 3rd one is my favorite only because it feels like the most accurate depiction of bokeh

[placid oar]

I had to choose 5, but got one quadrillion images out of that prompt

[mystic mango]

I love when Dall-E 3 just gives up

[placid oar]

it happens

[mystic mango]

In a vast desert under a clear blue sky, a figure stands in the distance. The figure is draped in flowing red robes with intricate gold accents, evoking a sense of mystery and nobility. The sand around glimmers and shines, with light reflecting off the surface creating a beautiful volumetric bokeh effect in the distance. The image is drawn in a minimalist cartoon style, focusing on the contrast between the simplicity of the desert and the complexity of the figure's attire. The artwork is in a wide aspect ratio, emphasizing the expansiveness of the desert and the solitary figure within it.
Here you go!

[placid oar]

opened a gallery that's just gonna feature candidates for daily themes I made

Posted a few of the most recent ones, I will upload more there, in no particular order and may or may not have a prompt

https://discord.com/channels/974519864045756446/1227353973963489300

dall-e messing with great katanas again

[mystic mango]

I wanted to make a chefs kiss reaction but that is against content policy for some reason.

[placid oar]

lol

I understand the content policy though, I don't want someone else's lips on my food 😠

as you mentioned earlier, I wonder if people are aware of the bokeh effect

so many images that are great but don't feature it

[mystic mango]

Yeah, it's tough, there's a lot of great images but bokeh feels pretty specific. I could ignore pointillism because I feel that Dall-E 3 has a difficult time with small details like that

[mystic mango]

Also you can't stop me

[placid oar]

Why should I?

I give up on depicting good katanas today

DALL-E is starting to mess with me

[mystic mango]

Allow me to clarify

This looks awesome!

[placid oar]

@mystic mango ok, now I'm ready!

[mystic mango]

I'm not a furry.. but

[placid oar]

lol

[placid oar]

oh no, I know that feeling

[mystic mango]

Building a PC is super easy and somehow also incredibly stressful at the same time. At least to me.

[placid oar]

I did that as a student, don't think I'd do it now

something i didn't show

This one, I could add a bokeh

and done, adapted to today's theme #daily-theme message

I had this one also, but posted the ohter one because of the black dress

[placid oar]

$

[amber coral]

That's fire

What's the prompt?

[mystic mango]

I am so frustrated with Dall-E 3 right now

[runic scarab]

Is that your expression rn 💀

[mystic mango]

lol yes

I just want an image of a person eating a hamburger and Dall-E is refusing

[wraith arrow]

my first test of the image selection editing feature -- looks more like a bad photoshop job hehe

[wraith arrow]

nice!

he should definitely keep the beard. what api is this using? it's not a seed thing?

[rapid carbon]

my first too:

[wraith arrow]

i selected the eyes and the beard, changed and removed -- what did you do?

[rapid carbon]

selected two opposing quadrants of the pokeball and tweaked them. northeast and southwest.

the changes aren't that pronounced, which is what i wanted.

it's hard to see the edit.

[wraith arrow]

oh, i see now, i think

that's cool

[rapid carbon]

the northeast one is brighter and the southwest is bluish and less shaded

[wraith arrow]

looks like the bottom half turned 180 degrees? looks like a thermal detonator from star wars

[rapid carbon]

cyborgpt version of a pokeball

i literally typed:

a = a pokeball

[wraith arrow]

that's what makes custom gpt's so awesome

you typed a=a or you input a? that's your menu?

[rapid carbon]

the menu is w, a, s, d

so a = is image creation

and a = a pokeball creates something like that 😄

[jovial aurora]

Unused Scampers Bokeh Image

[placid oar]

I had this or the one that was posted on daily theme

[jovial aurora]

Must've been a hard decision.

[placid oar]

It always is

btw "a corn field in nebraska" is auto time out for copilot

wasn't able to prompt anything for a few hours

[jovial aurora]

are you kidding?

[placid oar]

no

[jovial aurora]

I don't even begin to understand why that would be the case.

[placid oar]

same

[placid oar]

dang it, hard choices again

[placid oar]

Had my doubts about this one or the other one$

[placid oar]

for some unknown reasons... some cars landed on the water...

[solid anvil]

I do wonder what is happening.

[placid oar]

green hat duck is possessed, def that's the plot

[solid anvil]

It’s funny, I gave it as input a page from Donald Duck cartoon and asked it to create a similar image

This is funnier than typing a prompt

It degrades in a funny way

[placid oar]

there, no anime image #daily-theme message

It was either that one or this one

[solid anvil]

I’m not afraid of your monster

[trim nest]

Hehe

[trim nest]

Lito babies

That Orc in the far right corner looks so proud. He's like, "That's my boy."

[placid oar]

Copilot DALL-E going full abstact on me right now...

[trim nest]

These are cool! 😎

[prime bloom]

That's amazing! Brutalism architecture 🤩
Can you send me the prompt please?

[brisk root]

Thanks! Of course.
"90s footage of few soviet blocks out in field from very far away, clear dark sky, empty, 90s music video aesthetics, minimalism and surrealism, street lights, vhs aesthetics, grainy and blurry"

[trim nest]

90s music video aesthetics...that is an interesting perspective.

[marble stump]

Kind of a continuation to my daily post. Liked the concept and style, so kept going with it 😄

Almost tempted to do a whole series of these.

[brisk root]

That's why I use it. It's an authentic and nostalgic feeling for me.

It doesn't always work by the way. It varies on objects and things.

So you have to put it in the front , or in the back, or in the middle of prompt.

[trim nest]

It's a very nice look 😺

[placid oar]

Poor cat, what did you so to it, he's all corned and helpless...

[prime bloom]

Thank you so much!

[random vine]

This was more of an accidental generation, I asked ChatGPT what a live action monopoly would look like, but I meant in speculation, not an image, but it actually turned out good.

[brisk root]

I'm a bit late but this looks cool 🙂

[placid oar]

No more time to post a new sea themed image on daily, so here it goes

[brisk root]

which is better?

[solid anvil]

this is so.. odd. I'm no pink fan, but somehow I like it

[solid anvil]

Somehow I think it should have Beth Harmon in it.

[brisk root]

I don't know the style but should I call it... abstract? minimalist?

[mystic mango]

definitely a surreal minimalism

[solid anvil]

or bad architecture?

[mystic mango]

the pinks and blues give me a bit of a vaporwave vibe as well. or maybe Wes Anderson with that pink

[placid oar]

I hate when this happens...

[placid oar]

You find a magic prompt, and you don't want to let it go away....

same prompt made these

[placid oar]

same prompt

[amber coral]

I have a challenge for y'all

try to create something that looks like Egyptian Hieroglyphics

[placid oar]

for some reason, all images are coming now like this...

so many of these....

[amber coral]

very nice

[placid oar]

It was an very intense concert

[cloud prairie]

Some pirates imagined

Pirate escaping with gold.

Gold and Rum

Vision of an ocean goddess

Pirate diving to get ocean relics.

[brisk root]

Reference - DALL-E 3

Some blurred silhouettes

[humble jetty]

yo you will get this reference from a well known animated film but i made a donkey eating waffles

[brisk root]

DALL-E 3 September 2023 / DALL-E 3 April 2024

[brisk root]

I liked this image

Which is the best one?

[cloud prairie]

A fight in the street

[placid oar]

This is for @trim nest

[trim nest]

Check out my new GPT: https://chat.openai.com/g/g-jgvLB2tCS-spicy-boss-pepper-portmanteau-art-fusion

[radiant olive]

Pretyy cool

[trim nest]

Show what you're making

[radiant olive]

Newest one has been detailed statues

[trim nest]

That is very cool!

[radiant olive]

Thank you 🙂 I've been fine tuning to get very intricate results out of it. Trying to decide what "style" or artistic theme I will go for next

Like here is a ninetails

[trim nest]

Awesome work! Really good!

[cloud prairie]

I have been trying to get some traditional tropical houses from India.

Here is more attempts to get as realistic as possible. Somehow not very easy to get the person to sit inside. But this could pass off as a photograph

Vegetation was spot on.

This house at night was cool. Not sure if people allow chicken to roam at night. But that can still be excused.

[radiant olive]

Maybe that's what should be next, an architectural one

[cloud prairie]

Have you attempted architecture?

[radiant olive]

Not yet for the bots, but I have an idea on how to approach it.

[radiant olive]

THICK Boy bout to mess some stuff up

[cloud prairie]

This is cool. Very mountain vibe.

[trim nest]

He's the daddy of your nine-tailed one.

[radiant olive]

Just make a daddy bot

Only makes beef cake daddy's of various mediums lol

[trim nest]

That. Is. Just. Wrong!

[radiant olive]

Lol

[trim nest]

@cloud prairie Made it with this.

[cloud prairie]

Nice.

[trim nest]

I just said make a daddy king warrior for a nine-tailed fox

And boom

[cloud prairie]

Home kitten is impressed with superhero cat

[trim nest]

Awwww and I just hit my daily max. Now I gotta wait 13 hours...Oh well...it was worth it tweaking that gpt.

[placid oar]

What have you done? Too many gnome images?

[trim nest]

gno

Gnot this time.

[placid oar]

the problems of gnobalization

[radiant olive]

Ok making the "building" bot now

See what we can churn out

Maybe I'll do a gnome building first - modernized

Yep that's what Im trying first

Actually came out very well done

Even for a magical tower, overall great design

Ooo ok asked for higher detail resolution and not upset at all

[cloud prairie]

Pretty neat. All kinda tower like.

[radiant olive]

DANG ok it impresses me

[cloud prairie]

You say you make bots for buildings? Not using existing gpt?

[radiant olive]

Right I make my own custom ones

[radiant olive]

https://chat.openai.com/g/g-PVwHELe2R-wright-s-architecture-bot You can use it here if you want

[cloud prairie]

How do you make bots? I am asking at the risk of sounding clueless.

[radiant olive]

In premium chatgpt platform you can hit the "explore gpt's and in the upper right corner there is a create bot option

[cloud prairie]

Chatgpt plus?

[radiant olive]

I will say that I have refined my instruction sets for my bots to deliver the exact medium they are made for. So it's hard to reproduce alternative results that the bot wasn't desgined for specifically

[cloud prairie]

I see... and you need to train the model with image set?

[radiant olive]

Doesn't mean I have the best bots though lol, and no I didn't use images to train my bots

I just have super descriptive terminology and instructions I feed it

[cloud prairie]

You inspire me to try making my own bots @radiant olive

[radiant olive]

Happy to hear 🙂

The power of making your own bot is you can give it very specific instructions so that your images have a lot of consistency, but knowning what you want to consistently output is half the battle I suppose

[cloud prairie]

I see. We are living the world of magic. This was how I was when I was 3 years old. Now at this age, I could be 3 again.

I could say a magic word and there I had what I wanted.

Got a nice drink working out. Who wants to join?

Here is Vodka in Russia

Japanese rice wine in Japan

Beer from an English pub

[placid oar]

Got something less common? More exotic?

[cloud prairie]

And wine straight from an Italian cellar

[cloud prairie]

Even I hit my limit. So have to take my curfew

[radiant olive]

You just gave me some more inspiration

A liquid bot

[cloud prairie]

Super neat

[cloud prairie]

Here is Coconut toddy from tropics of India

Here is coconut toddy, and fish fry straight from the backwaters.

Bring some nice creations everyone.

So cool to see all the works. I am sure many have exhausted their quota and are forced to wait till next creation.

[cloud prairie]

Man that is cool.

[placid oar]

Here, some Aguardiente to keep you warm in the high mountains in the Andes

[cloud prairie]

Here is some brandy

For the silent whiskey drinker

[placid oar]

some Rum

[cloud prairie]

Very inviting.

Everyone is getting a lot of drinks.

Looks like a good day.

[placid oar]

I don't dink alcohol, but I can images still

Maybe a tequila

[radiant olive]

Rum and brandy at the same time

[cloud prairie]

Alcohol is a pretty thing even if not used

Sight wise and looks wise, alcohol is made the prettiest. Coffee is never as pretty as alcohol.

[radiant olive]

Wanna bet?

[cloud prairie]

Here is some drink on the beach.

[cloud prairie]

What are you betting on?

[radiant olive]

Lol was joking but Imma show you a pretty coffe ready?

[cloud prairie]

Oh sure. I will be happy to lose to coffee

Give us a coffee in a coffee shop. With a pretty barista

[radiant olive]

The bot is tailored to make the drinks not scenary with it lol

[cloud prairie]

That is cool.

[radiant olive]

That's the only limitation is that it hardline focuses on the instructions

So it can do the liquid but doesn't care about anything else