images-discussions

[grizzled iris]

Woah 😧

[grizzled iris]

That’s really beautiful

[grizzled iris]

Heck yeah a whole army!

[tall mason]

The malachite and ruby statue stuff is from a terrifying dream I had after a night of watching some C'thulu related stuff.

[grizzled iris]

Well I’m proud of you for facing your nightmare head on 🫶🏽

The rock gets better each time!

[tall mason]

Lol, I read Call of C'thulu after that and thought it was lame. The dream was so intense cause it was one that felt real.

[grizzled iris]

I don’t think it can get any better than this -

These are so cool I’d actually print them off if I had ink 😂

And if I owned a printer lol

Guess gotta use the work printer

😂

[tall mason]

You have access to a printer with cyan?

[grizzled iris]

I will find out tomorrow!

Or get told off for wasting a whole cartilage on a rock

Hmm 🤔

Never seen this bug before

It made exactly the same images as before, that’s a new one for me…

[tall mason]

Careful about messing with the company's liquid gold.

Yeah, sometimes, GPT just straight up ignores you.

[grizzled iris]

Yeah that was a first for me seeing exact the same images, while at the same time wasting my credits 😂

[dim cradle]

Is cyan hard to find? I haven’t had a printer in years

[tall mason]

It's just a common joke about printers always being low on cyan and refusing to print, I think some even refuse to print b&w like that. Guess cyan is the most used.

[dim cradle]

Ahhh gotcha.

[grizzled iris]

Omg lol

Todays theme I made a same themed rock 🪨

Diva 👩‍🎤

The Rock

[tall mason]

Now have him in an apron cooking at the stove.

Or, with humanitarian stuff, have him cooking in a soup kitchen.

[grizzled iris]

I’m trying I’m stuck again

[tall mason]

I missed this message. I considered that, but didn't seem to help. But if it's working for you.

[grizzled iris]

Brb

[tall mason]

Lol, no need to waste your time on a silly pun, though.

[grizzled iris]

It’s ok - it’s good for feedback -

[tall mason]

Goodluck.

[grizzled iris]

Fixed

Phwew

Only took 7 min ☹️

Why do they crack me up 🤣

Anyways onto todays theme

Still haven’t done that one

[tall mason]

The fantastic doing the mundane is always wonderfully absurd. Can't find the image...Oh, well. Good luck with your divas.

[dim cradle]

universal healthcare, renewable energy, volunteering 👏

ok, with gpt you can say <whomever> "looking as diva as possible" and so far i haven't been disappointed

apparently adding "shown in full" provides those full-body shots we sometimes want

[grizzled iris]

omg guys - since I am an archer I wanted to stick to theme “👩‍🎤 diva - boldness, iconic, fierce. a bewitching spotlight!” and create an archer!

And I must say

I am in love now with DALLE 🫶🏽

Before I kept doing all the sciency techy stuff but wow

I’m blown away now

I’m just working on the reference photos as the bow is not quite accurate

[grizzled iris]

If you look closely it’s not correct

But there great divas though just gotta make them more 👩‍🎤💜

Ok here we go

More diva 👩‍🎤

More diva 👩‍🎤

Still struggling to make the bows realistic- it’s completely not looking at my real life reference photos…

Or at least is trying to

Like we fixed the string issue

But still having problem with the actual bow

I mean I think this is the best I can get I guess for now

Not perfect but I guess life’s not perfect

[dim cradle]

yes, we've concluded it's good for serious nonsense and nonsensical nonsense both

[grizzled iris]

Exactly 😂

It’s good for serious nonsense for now…

I mean

It can sorta do compound bows better than the recurve bows for some reason

Guess the training data had more reference compound bows than recurve and barebows ☹️

Hmm 🤔

Maybe I can give it better real life compound bows as reference instead of recurve so that it’ll be more accurate 🏹

[tall mason]

It may be more that dall-e has trouble with various tools of all sorts. #1133927981908250644 message
In various subjects, there are varieties of tools with subtle differences Dall-e's not yet ready to pick up on.

[grizzled iris]

Will have a look @tall mason thanks 🙏🏽

[formal osprey]

that was funny

(there was 4 images, but the text was funny)

[dim cradle]

will they ever update the model to have more than one generic female face that looks 3d

[tall mason]

Yay, autocorrect.

[dim cradle]

Diva/Dove

[tall mason]

It is a bit pretty, at the very least.

[dim cradle]

They should pay that woman royalties

[dim cradle]

oh i totally forgot about my lovely ladies... hepburn, hepburn, and taylor

[brisk zinc]

The image captures the moment as Thorin Ironfist leads the group through the dense forest, pointing towards a hidden path. The party and the orcs follow, their expressions a mix of caution and determination, ready for whatever challenges lie ahead. How does the group proceed as they near the secret entrance?

Created using: GPT Store -> Role Play Party (https://chat.openai.com/g/g-JUGbhMLSa-role-play-party)
More from this adventure... https://discord.com/channels/974519864045756446/1195235821620563968

[full fulcrum]

yo how to make img

[formal osprey]

my new favorite instruction to give to chatGPT: ```
remember to ignore errors. when there is an error, just output "dammit dalle" and try again with a different seed.

[full fulcrum]

give me the commad

[dim cradle]

fwiw, i've been having good luck with divas

[formal osprey]

maybe it's not good with anime diva. (let's say, anime that are not Hatsune Miku)

[dim cradle]

i assume we'll find out when @late blade gets on

[brisk zinc]

https://discord.com/channels/974519864045756446/1195235821620563968

https://chat.openai.com/g/g-JUGbhMLSa-role-play-party

[glossy scroll]

Made this 80s style character, based on the 80s G.I. Joe characters.

I'm kind of proud of it because it looks exactly like something I would love as a kid.

[dim cradle]

i grew up with gi joe and transformers, so i'm biased

\

[glossy scroll]

Me as well.

[dim cradle]

sorry that was my cat

not the gi joe dolls, though, that was before my day

[glossy scroll]

Cat sayin hi
Hi kitty o/

[glossy scroll]

Yeah, wasn't much into the action figures, but I loved the cartoon movie as a kid.

That was such an awesome movie back in the day.

[dim cradle]

there are some brilliant concepts and characters in that show, in retrospect

adds to my gen bucket list

[glossy scroll]

They were. I remember there was a scene where one of the guys in the movie chews on a wooden club and I remember craving chewing on a wooden club as well...Ah my child brain...I remember it well!

[brisk zinc]

https://openai.com/dall-e-3

[glossy scroll]

[dim cradle]

🤦‍♂️

i haven't made a good Serpentor (Cobra Emperor) yet

[glossy scroll]

That would be cool. Make sure to give him a hoverchair (That guy never liked walking)

I never really understood the whole snake aspect

[grizzled iris]

is DALLE Down?

[dim cradle]

nope, just a hiccup

[grizzled iris]

hmmmm

[dim cradle]

[grizzled iris]

It's ok we fixed it

Opened a new chat and explained I did not have much time left to submit my dissertation in and I needed this

[late blade]

hmmm?

[dim cradle]

can dall-e make good anime divas? i guess that's what i'm asking

[late blade]

yes

[dim cradle]

ok, i think another member of our community was wondering

i feel like i've seen her somewhere before

[late blade]

daily theme?

[dim cradle]

right, i thought so, but what about the high-fashion anime?

[dim cradle]

someone may have beat me to Elizabeth Taylor

i asked for "anime" -- i know nothing about it except a few movies i liked, i don't know all the styles, etc.

they were all by Hayao Miyazaki

[late blade]

let me find out, I have some different anime art styles defined

[dim cradle]

cool. i just looked some up. trying psychological now, sounds like a blast

[shut niche]

I missed the G.I. Joe conversation. "Knowing is half the battle." LOL

[grizzled iris]

noooooo - guess we're back to 3.5 while we wait...

[dim cradle]

haha love how it always, and i mean always, ended with them laughing

[grizzled iris]

exactly 😦

I can't make no more images - and the 4 free credit guy is on holiday ... 😦

the /daily bot 😦

sad times

guess i can just watch other's great art for an hour lol

[late blade]

yeah totally possible to do the diva anime stuff

[dim cradle]

awesome

my first "psychological anime"

[late blade]

is the service still unstable? I'm getting so many errors atm

[dim cradle]

i guess he's got a lot on his mind also

i haven't seen any errors

[grizzled iris]

in today's theme

[grizzled iris]

looool

[grizzled iris]

ok would watch

[grizzled iris]

this is good

[dim cradle]

if you request a deva male, he has similar bone structure as that woman

[grizzled iris]

your right can you ask for a male archer deva version please based of that image

Arrrrrrr

[dim cradle]

male archer? lemme see

[grizzled iris]

gonna do an image of this proverb "When elephants fight, it is the grass that suffers" Kikuyu proverb
"So, when humans fight, the world suffers"

[grizzled iris]

let's see

i can't make any more so all I can do now is watch. and not create.

[dim cradle]

gpt went all chad on me, "chiseled physique", "alabaster skin" -- not going for a romance novel cover here.

[dim cradle]

craft your ideal prompt for this deva archer in gpt-3.5, send it to me, i'll gen it. you get 1 shot at this, if you accept this mission at all.

[late blade]

dall-e is having trouble with other anime styles I usually have no problems with

[dim cradle]

with the diva specs?

[grizzled iris]

accepted

[late blade]

yes

this is whit the anime style I use for my elven couple

[dim cradle]

hmm. kinda reminds me of designing holodeck characters, this ai art

[glossy scroll]

"Don't judge people til you give them a chance. Cuz there's a chance you could be wrong."

Haha

[grizzled iris]

Sorry, I can't lol - I will have to use my brain for this prompt lol

[late blade]

what the heck?

[dim cradle]

your account is restricted?

[grizzled iris]

no it's my custom instructions

[dim cradle]

dude

lol

[grizzled iris]

to keep me focused on work and not diverge too much

[glossy scroll]

[dim cradle]

i see, maybe ask it to rephrase, optimize? i don't know how you construct your prompts. i just find that sometimes works for me when i see that.

[grizzled iris]

no i know i cant get around to it if I wanted to but it's good to see the failsafe actually working now lol

[late blade]

it doesn't happen all the time, it happens with specific art styles. It depends on which json file I feed to my prompt builder

[grizzled iris]

"Must Focus On Science ImaGes Only" lol

[late blade]

I can give you a tip on how to work better. Do this exercise: Sit down, feet on the ground for 2 minutes. Breath in, breath out, during those 2 minutes don't think of the pink elephant on your shoulder. Once you do that, come back to me and I'll ask something. Then you'll be able to focus.

[grizzled iris]

tempting

who has this new one?

[late blade]

@grizzled iris 2 minutes starting now!!!

[grizzled iris]

ok ok!

[late blade]

How was it with the mind exercise?

[grizzled iris]

it was good

[chilly onyx]

You went for the plan then?

[grizzled iris]

looking at this

quite interesting

[late blade]

me?

[grizzled iris]

even though AES-256 can be broken... especially with SNDL - Once they have Quantum Cryptography I'd go for that with my no money in bank account balance lol

[chilly onyx]

Yah, thought you just said you’d had it for a whole two minutes 😅

Think I misread

[late blade]

I asked @grizzled iris to do a thought experiment to help focus, but apperantly we got to take that credit card away

[chilly onyx]

Ah, the elephant

[grizzled iris]

yes otherwise I will go further into the minus

😆

[late blade]

So, did you do the experiment?

[grizzled iris]

it was good 2 min mindfullness - mental health is key

[late blade]

but were you able not to think about the pink elephant on your shoulder?

[late blade]

I guess you didn't

[late blade]

which is also ok

[grizzled iris]

[late blade]

The idea of the thought experiment was to let you know you have something you want to do and you need to do it before you can focus. So set a plan to do that and get back to work.

The elephant is actually what it's on our mind at the moment

[grizzled iris]

ohhh i remember now this image "When elephants fight, it is the grass that suffers" Kikuyu proverb "So, when humans fight, the world suffers"

I wanted to make

Test the Ai on Proverbs

and then train it lol

since it will get it wrong

[dim cradle]

Did my African Diva lose votes or just my phone? She’s only got 1 star but I know there were more. Same for some others that went down.

[chilly onyx]

Weighing heavily on your mind

[chilly onyx]

Ignore the @ there Shon

I got muddled

[late blade]

it's an exercise I used to teach in psychiatry and show students how thoughts can influence you on your daily tasks.

[chilly onyx]

I think I lost votes is what I was going to say

[chilly onyx]

Ah, nice exercise

[late blade]

Yeah, don't need big books to teach the basics you already know.

[grizzled iris]

nanorobot in a blood vessel

[late blade]

using nanotech to turn you into a diva?

[grizzled iris]

that's clever

[grizzled iris]

didn't think to make a nanotech diva!

[late blade]

doo eet!

[grizzled iris]

right, there's a backlog brewing now lol...

[late blade]

c'mon you are the nanotech expert

[grizzled iris]

I don't know what that is though lol^

never seen this at all

[dim cradle]

you lost votes also?

[grizzled iris]

they need to be more like this

but we will get there

[dim cradle]

never mind, it's just the phone app

[grizzled iris]

DALLE just make ones from 2025 not 2024 lol

[chilly onyx]

I think so. It wasn’t one I cared about, just the logo one, but I swear it had some and then suddenly just the one.

Ahh, well I am using the phone app

So that makes sense

[late blade]

right on point, the "divas" i know irl are very unorganized and chaotic

[grizzled iris]

so jelous

male diva's not too long left

[final compass]

was just about to say that there is some organized chaos going on in that 😄 in the sense its not a total mess. but the atmospheric play of light and shadow is perfect.

[grizzled iris]

these 8 min feel like 8 hours

[grizzled iris]

The arty diva!

[final compass]

i don't know why, but it feels like all these 'photorealistic' faces are the same face unless you specifically tell it what kind of eyes, nose etc you want.

[grizzled iris]

Oh yeah it's my error

[grizzled iris]

I fell in love with this type of face and I specified I wanted the face to remain similar lol

[dim cradle]

kinda got that seven of nine look

[final compass]

My first intuition was just that i think i have created this exact same face many times and often see the same likeness happening everyone else's generations that have this same visual style.

[grizzled iris]

we need to love the imperfections too sometimes

the bow is wrong but it's a start

`

ok...

[shut niche]

Now that's bias in action, lol. That red jacket's a little snug, and open in the front.

[grizzled iris]

algorithmic bias

i knew it

can't do it

hmmmm

\

[dim cradle]

well, that's serene and utopian

[late blade]

my penalty box senses are tingeling

aaaaaand, i'm in the box

[grizzled iris]

[empty kelp]

the first image has the “looking from the back” effect. it transformed early morning into late afternoon. it does that with fire/ice, good/evil; etc.

it happens when the scene (forest, beach; etc) is in the image background, or in the background of the views and one of the views looks at the character/focus from behind. it reverses everything instead of drawing it from behind

if you want to actually view the character from behind in the 2nd view then you can put the entire scene in both views (and it will look normal and not reverse everything)

[dim cradle]

Xenomorphs can be divas also

[late blade]

@dim cradle so, anime divas are easy, even with one line, I just did that with #daily-theme message

[dim cradle]

yeah, i just saw that. i had confidence that dall-e could handle it.

[late blade]

hehe

ok, but now the challenge, got to do something outside anime

[dim cradle]

there are still a number of historical candidates, alternate styles, open for new concepts

[late blade]

I was looking into the candidates, but it's my to go thing when I don't have anything else, so gonna force myself to do something different

[shut niche]

This is epic

[dim cradle]

oh, i thought you must've been referring to my xenomorph diva

thank you, sir

i mean, i can't take alllll the credit

dall-e did most of it, and the critically acclaimed Audrey Hepburn was with your custom gpt, many thanks

[dim cradle]

i think its construction was very thoughtful

[dim cradle]

any ideas?

[late blade]

yes, I'm hungry and planning the route to the kitchen is not easy

was thinking about songs, look songs and see what I can find

[formal osprey]

I've been experimenting with dalle3 gen_id to recreate the same character, works pretty well

[dim cradle]

be like that person who said more charcoal. charcoal it to the bone. you have to do that with diva. i've had to request more diva a couple of times to get the full effect.

[formal osprey]

some pictures can be identical if we ask it to use the exact same seed, prompt and reference. varying only sightly the prompt leads to very simmilar pictures

[late blade]

I'm trying to do a spoiled diva that has access to sudo in the terminal...

[dim cradle]

root diva

as diva as possible

then more diva

[formal osprey]

here's what I meant with "can be identical". I used the same seed, prompt and reference for those 3. you can see the very subtle differences

[late blade]

Oh, I got something really coold for today's theme

[late blade]

hair and slippers vary a bit

[formal osprey]

it's not the exact same image, which is good. it means the 3 are actually generated, it's not just reuploading the previously generated image

[late blade]

it's damn close tho, even most of the particles are 1-1

[formal osprey]

for me that's quite nice to see. it helps me understand better what happens under the hood, to an extent

[late blade]

yup

I had the same thing with my steampunk cat-girl, but now I can do images of her accuratly

[formal osprey]

it generated one flipped once. (well, not flipped, but looking the other side)

[dim cradle]

i was just reading about that, the subtle differences are caused by stochasticity and noise during the generation process

[final compass]

Interesting i used to do identical generations with minimal changes, but lately when i try to do that GPT just tells me it's not going to do that.

always amazingly frustrating when proven process suddenly stops working because GPT doesn't feel like it today 😄

[formal osprey]

for me, those are more interesting than the "identical" ones. the only difference is that I asked a slight variation in the prompt for those. It's very nice to have that kind of control possible. (and weirdly enough, I'm almost hesitating to discuss about that because I feel like openAI might "patch" this out...)

[final compass]

yeah it is interesting method, also noticed that if i changed anything at all, added a word, removed a word, changed to synonym i would get the same image with slighly different angle etc. I think this should basically happen if you just change the seed number and keep everything else the same.

[late blade]

prob at this time it's what's going on under the hood until it gets officialy out

[formal osprey]

changing the seed number has a dramatic effect on the result. due to how pseudo-randomness works

[final compass]

I can relate to this. The hesitancy of exposing something interesting that does something awesome that it's probably not supposed to do or allow.

[formal osprey]

in those 2 pictures, it's the same seed and the second is referencing the first
the prompt only differs with that part added for the 2nd one: , zoomed out, the rest of the prompt is identical.

prompt: ```
A digital watercolor effect illustration of a 2D anime girl in a full body shot from a dynamic low angle, zoomed out. She's wearing a black tutu, similar to the Black Swan, adorned with shimmering gemstones, and her hair in a sparkly ballet bun. Her vivid and mesmerizing eyes reflect the intensity of a ballet performance, with a bold and confident diva expression. She's on a dark stage with a solitary beam of light emphasizing her movements. The style incorporates luminous hues in a 2D anime aesthetic, capturing the essence of ballet through dynamic posture and lighting.

(the "zoomed out" didn't work at all, lol)

[late blade]

@final compass think Paws🐾 and #daily-theme message are in the same clique?

[final compass]

Yeah, i would think so 😅

[empty kelp]

“full body shot” won’t work. To zoom it out you could describe the shoes the character is wearing, or say that the view is unobstructed and focused on the character

if you tell DALL-E to focus on a character’s feet it will likely see it as inappropriate and ignore it — but describing shoes or something under the feet will put the whole character in view

[green pebble]

It doesn;t do feet pics for free lol jk

[final compass]

Relaxing to do something whimsical once in a while.

[empty kelp]

She's wearing a black tutu, similar to the Black Swan, adorned with shimmering gemstones,
brightly glowing blue sneakers, ...

[thick smelt]

cool ai art yami

[empty kelp]

shoes work to put all of her into frame. just need to add that she's standing up.
If not using the API you can make it portrait with:
The image should have 1024x1792 resolution, portrait orientation, and HD rendering.

instead of "low angle" you can say, "the view is from the ground"

[late blade]

strange, I tried that and I had better images with low angle.

but I also load a file that has cinema shots defined

Still working on the data and how to structure it, but this is what I got

[humble sonnet]

Someone said it’s impossible to make real art out of ai art, so im making a lyric video with assets now for fun

hehe

[late blade]

oh, do tell more about it

are you including images from dall-e?

[final compass]

the unsinkable

[humble sonnet]

thinking of making a music video for a class project using a combo of dalle and photos from a camera

[late blade]

@clever phoenix because of this #daily-theme message I won't be able to sleep for the next 2 weeks....

[clever phoenix]

I agree, koalas are terrifying 😉

I had considered going with a cuter version

[empty kelp]

i see a lot of talk on seeds. are seeds working again?

[late blade]

hansa had some good stuff going on

[empty kelp]

going to test it now

[late blade]

err i mean yami

[quartz vale]

[late blade]

my brain has the wrong seed and the wrong gen_id

[clever jungle]

You can't put a dino in a bowl and expect me not to eat it

[quartz vale]

You should try dinosaur oatmeal then!

[formal osprey]

you're right. that's what I did right after. I was curious if "zoom out" with a referenced image would have an effect relative to the referenced picture. But nope... I wonder if the referenced image affects the visual or if it only set the same seed 🤔

A significantly zoomed-out digital watercolor effect illustration of a 2D anime girl performing complex ballet moves on a dark stage, showcasing her footwear. She's wearing a black tutu, similar to the Black Swan, adorned with shimmering gemstones, and her hair in a sparkly ballet bun. The dark stage is more visible in this zoomed-out view, highlighting the solitary beam of light emphasizing her elegant and dynamic ballet movements. The style incorporates luminous hues in a 2D anime aesthetic, capturing the essence of ballet.

"footwear" is the generic term I ask chatGPT to include when I want a full body shot. describing the shoes is too verbose, it affects the result

[dim cradle]

lol i guess that person has never heard of digital art or all the software people have been using to create it for decades.

i don't know what i'm supposed to do with koalas

[empty kelp]

you gave it a hint of what to zoom out from saying that “the stage is more visible in this zoomed out view”, but the light also needs to be hitting the stage to make it visible. If you want to keep the everything the way it is and have it zoom out more you could add, “A dim [color] ambient light illuminates the stage”, or “the surface of the stage has a dim [color] glow”. That will make your existing zoom work

You could also tell it that a dim ambient light illuminates everything

[quartz vale]

Prompt: Koala

Done!

[dim cradle]

You make it look so easy 🙂

[quartz vale]

It can be that easy

[empty kelp]

now you can get the GenID and seed, and have the koala go on adventures with the athletic and diverse female elves

[glossy scroll]

Koala. Haha. Yes!! My kind of theme!

[dim cradle]

i knew you would dig it

[glossy scroll]

I'm remembering my chubby koala I made a week ago

[dim cradle]

my koala has attained the 55th level of awareness.

[glossy scroll]

Tubby koala adventures!

[formal osprey]

what would be nice is a general instruction that works with any subject and surrounding

[dim cradle]

Tubby and his wife look like clones lol

"shown in full" works often, depending

[glossy scroll]

Lol, you just can't tell them apart. The wife is more fluffy.

[dim cradle]

"And how to tell his wife from all the other ducks?"

[glossy scroll]

You got this.

[dim cradle]

hahaha

the gentledogs have a new golf buddy

let's see what kind of blending we can do with a koala

that's not working as i imagined

[glossy scroll]

He looks like the CEO of Koala Corp

[dim cradle]

He founded his enterprise from the ground up. His only initial resource was a substantial inheritance.

[glossy scroll]

That sounds like a pretty accurate narrative

[final compass]

pretty mundane stuff, wanted to see what kind of graphics cards dalle generates 😄

[late blade]

oh dall-e is really good at those

[dim cradle]

haha nice

[empty kelp]

“please draw three elves and a koala”

[glossy scroll]

I swear I actually looked like that 15 years ago

[dim cradle]

it doesnt ever hold back on elf ears does it

[glossy scroll]

I transitioned from a koala into a gnome

[dim cradle]

we pretty much slide on an oil slick from 35 to 50

[empty kelp]

i think the seeds are working btw

[dim cradle]

senior mod, lock up your koalas

[final compass]

@clever phoenix #daily-theme message
Pretty sick PC! Really nice image 🙂

[clever phoenix]

Haha thank you, building one of my own so it's been on my mind 😁

[dim cradle]

🐨

[final compass]

So many good koalas in there. Starts to become challenging to come up with something original to post in that theme 🤔

[glossy scroll]

Trying to make a koala Disney character

[dim cradle]

that one looks a winner to me. lets get this movie made now

[late blade]

Looks kinda like Po from Kung Fu Panda

[glossy scroll]

We'll call it 'Kalo's Kaola Kingdom'

[late blade]

hehe

[glossy scroll]

No, Po looks like Kalo, not the other way around.

Koala's are the best

[dim cradle]

Kalo could never defeat Malachite the Menacing Koala

[glossy scroll]

Lol. Dang. He looks quite menacing indeed.

It's hard to not like evil koala's.

[clever phoenix]

All Koalas are evil

[glossy scroll]

Every Disney adventure movie also has to have a Disney princess. Princess Eucalypta

[dim cradle]

love the name

[glossy scroll]

Hehe

[dim cradle]

@clever phoenix tell us what you really think lol

i'm just going with the surreal on this one

[tall mason]

I think Dall-e may be trying to tell me something here.

Cartoon illustration of a koala with a distant, thousand-yard stare, gazing into the distance. In the foreground, a human hand is holding a eucalyptus leaf towards the koala. The koala's expression is humorously exaggerated to emphasize its deep, distant gaze.

[wooden lodge]

Are loras better for image style transfer than img2img? What do AI image platforms usually use for style transfer?

[dim cradle]

dall-e uses transformer architecture, but the model doesn't do style transfer per se, it only uses a version of it to help align natural language with visual concepts (dual understanding, evolution from traditional style transfer); as to other platforms and how they compare (off-topic for this channel), they use a variety, and it depends on their use cases.

[wooden lodge]

my bad I shouldnt have asked in dall-e discussion

[dim cradle]

Shon of the Living -v- Koala Wizard

[glossy scroll]

Koala wizard that summons baby koala

[dim cradle]

i didn't know you were into horror

[glossy scroll]

Only cute horror

[dim cradle]

ahhh, it's cute until you're ravaged by baby koala

[glossy scroll]

I want to make a movie based on that

The revenge of the baby koala

And have a koala peeking from behind a tree...watching your every move...

[dim cradle]

shon of the living dps does cc, no worries

[glossy scroll]

Temporarily Shon of the living. Your names opposite is because of the baby koala.

🐨

Going straight for your pinky toe. That's a baby koala's favorite chew.

[dim cradle]

for a gallery on my sorcerer character, i am working on the cover art -- the challenges: 1) description accuracy, 2) hyperrealistic foreground and background, 3) a title in a fancy typography, 4) correct spelling... did i miss anything? it's tricky

[west veldt]

Why is it that Dalle 2 seems to have the capability with prompts that include “make it look like it was taken with a 1950s. camera” but dalle 3 wants it to look perfect so it makes ultra realistic images every time? i feel like with 3 you have less control over the quality of the image

[dim cradle]

vivid is the default, you might try natural style.

[dim cradle]

the api is not even attempting to title here, even though it's in my prompt twice. so i will try decomposition...

[glossy scroll]

You have to specify that it should "keep the foundational style of that era"

[empty kelp]

here are HD versions of the elves who made off with Dinosaur Oatmeal's Panda. i sent the images from the iPhone app earlier so they were like quarter resolution

[dim cradle]

By explicitly stating the need to ensure the inclusion of the title in the image, you should receive a generated illustration that incorporates the title as part of the artwork.

evidently i failed to explicitly state it in my prompt. noted.

[empty kelp]

uh,.. koala i mean. not panda

[dim cradle]

still not doing the title

hmm. maybe i need a template.

never focused on character gen before

i need to refine my specs

but i'm still not sure what to do about the title, because i've now explicitly stated it

[empty kelp]

i'd add it after in photoshop

[dim cradle]

well, that changes everything, and gave me an idea, thanks

[graceful spade]

How can I replicate this style of Dall-E 2? Dall-E 3's painting style is realistic and detailed, not like Dall-E 2

[empty kelp]

just need to find an artists who died 100+ years ago and tell it to use their style. that looks like a cross between Van Gogh and Dr. Seuss

this was my first attempt at telling DALL-E to draw the elves showing their teeth when smiling. it looks kind of amazing

they're winking also. that's why the eyes are like that

[dim cradle]

you might try uploading it in chatgpt and collaborating with it on that.

[dim cradle]

@glossy scroll

@dense mesa

[quartz vale]

👁️

[dim cradle]

both are surreal, but that's a good-sized one

[formal osprey]

I like that it's so random

[sick flax]

DE3 does tend toward realistic/detailed. What I've found is, the prompt needs to be more assertive if we want to go against the grain. Not that this image is exactly what you had in mind, but at least, it shows the possibility to create a looser, more abstract style. You might also try Bing Image Creator (if you haven't yet), which has different characteristics than DE3 via ChatGPT. And, if you have ChatGPT Plus, you can try feeding the image and ask to analyze the art style. Good luck!

Prompt for the attached image: "An abstract surrealist oil painting, subtly blending Impressionism and abstract expressionism. The painting depicts a fairytale scene with a little old yellow stucco cottage featuring a steep roof, situated under the canopy of a tall tree. A girl in a blue dress and a white dog are seen walking towards the brightly lit entrance of the cottage. The night sky is rendered in a deep midnight blue, creating a mysterious and enchanting atmosphere. Chiaroscuro is used to create a dramatic contrast of light and shadow. The style is dreamy, ethereal, and minimalist, marked by loose, expressive, and bold brushstrokes, giving the painting a distinct painterly quality."

[dim cradle]

shock value

[dense mesa]

Why does that remind me of Jeff bezos?

[dim cradle]

my old tales from the gpt can output an entire tale in comic panels, 60's comic horror style, in 1 response, and it would be perfect if it would spell the text bubbles correctly

it gets some of them right (or pretty close)

[rain heart]

一只小狗

[lean iron]

why is GPT Plus dall-e so much more garbage than Bing Creator Dall-E?

its so terrible

[dense mesa]

is it?

make this image on bing

[lean iron]

for me it is

left one is bing right one is gpt plus

same prompt

[dim cradle]

depends entirely on the prompt

there are explanations within your prompt most likely. you're welcome to share it.

[open trench]

hello, all.

[dim cradle]

hey, glad to see you're still among us.

[open trench]

I am! lol

[dim cradle]

dude, you missed the DIVA daily

[open trench]

I know. I'm sad I missed it. I would've enjoyed making some things up for that one. I didn't get a chance, though.

I didn't even review. I'll have to go through and see what magic people made.

[dim cradle]

i recommend it. there are some great ones.

a couple of mine turned out well, i mean, better than i expected, shows how far the tech has advanced

[lean iron]

visual novel illustration, 1woman, Yoruichi from Bleach, Yoruichi, tanned skin, smirk, extremely detailed image, extremely detailed eyes, sitting down, leg crossed, resting chin in the palm of her hand, crisp and bold line work, 1024x1024 image.

thass the prompt

for both bing and gpt

[open trench]

I will try to find them

[dim cradle]

i could interpret that as b&w. it's a bit ambiguous. think of it as being very literal. throw some color in there.

especially with "bold line work" in the prompt

that might've steered it there

[dim cradle]

did you paste that prompt into a new dall-e gpt convo?

[open trench]

Oooh. I wish I had written down where I was going with Scampers. I have like 4 images and don't remember what comes next. Maybe if i reread the prompts.

And why is it telling me to upgrade. Apparently, I missed more than I thought.

[dim cradle]

And going through that interface--with the prompt rewriting at the api end which you cannot disable--a prompt that short is going to be refined and augmented also, so you might want to request the ai to expand and make refinements before generating the image--it takes a much longer prompt, there is room to be more specific.

[dim cradle]

you've got a whole script and everthing huh?

[open trench]

I wish. I just wrote down enough notes to basically get where I am, but it feels like something is missing. It may be because I can't make the proper transition image between points A and C or I just don't remember where I was going with it. Upon reviewing the prompts, I think I was struggling making the transition image.

[dim cradle]

Are you using gen_ids? I think they only work in the same conversation? I think you’ve used them more than me so I’m wondering

I’ve really just been holding off on narratives until they release that feature officially and document it.

[lean iron]

OK so in summary you're saying that i should make my prompt much longer and also add some color in?

[lean iron]

also what is "b&w"?

[dim cradle]

I suspect it’s going to be a labs feature. I just didn’t expect the labs launch delay.

[open trench]

I am, if you're asking me.

[lean iron]

Bing doesn't do the prompt rewriting?

[dim cradle]

black & white. shorter prompts can be open to interpretation, it doesn't have to be longer, but adding some reference to color is one way to help steer it in that direction.

if that's what you expect to see, describing that instruction in a concise way is going to help get you what you want

if you regen that same prompt, you might get color, it's just not saying that and some terms might make it go in a direction you didn't want

[lean iron]

alright, thanks.

[dusky panther]

Do gen_id and reference_images work with the actual dalle api yet or just inside Plus accounts?

[dim cradle]

please let me know how it goes, share here if you like it ha

[lean iron]

also, does anyone have a list of art styles Dalle recognizes well? specifically different anime or manga styles

[dim cradle]

they're not official parameters, but i have seen some posts from user tests in the dev forums where they included the referenced_image_ids along with the prompt which appeared to work.

[lean iron]

i've been going on the dalle 4chan and there's some really good anime themed art there

like this one

no idea how to explain it to dall-e what that art style is

[dim cradle]

maybe that means once the feature is fully developed then the parameter might be added to the api, just speculation

[dusky panther]

Thanks for update. I’ll try it out

[dim cradle]

we have some anime art fans here who might could help when they're around

i was just recently looking up the basic types, but as far as style and quality, that might depend on your prompt and interface

[lean iron]

oh, where were you looking it up? and by interface do you mean like which site i use?

designer, bing, gpt plus and all that

[dim cradle]

i just asked gpt-3.5 🙂 yup, all that, or a custom gpt for images someone has made, but between gpt and those tools, you should be able to generate your own.

[lean iron]

ah okay

thank you

[dim cradle]

sure, and you can upload that photo to gpt-4 and ask it to generate an art prompt based on that. ask for a long, detailed art prompt. you can use that as input for dall-e.

[open trench]

i asked for bigger flowers....

[dim cradle]

That’s smart of Scampers lol

And that reminded me of a dream from years ago. Wild when that happens

[open trench]

indeed!

[dim cradle]

I think it was the angle and the overall impression

[open trench]

I am stuck.

[dim cradle]

have you tried a koala yet?

[open trench]

No. Koala inspiration thus far lol

finally got the angle and action I wanted. I just don't know what to do next.

[dim cradle]

added the text post-processing, but the speech and narrative bubbles are dall-e

[open trench]

nice

[dim cradle]

thanks, i was just testing if it would create empty bubbles

[late blade]

Did you check this #images-discussions message ? I am working on a cinema data structure, this might help you further with angles and cameras.

[open trench]

wow. Nice. that's helpful. 🙂

[dim cradle]

i thought he was going to climb down to avoid the thorns, it might be time for some magic

[open trench]

I agree that it's time for some magic.

[sleek hinge]

Hello guys im planning to use DALL E to generate a mathematical geometry prompt

Have anyone tried it can it do it perfectly

Prompt like "generate me an image of function R=1 in polar coordinates graph"

[late blade]

don't think that would be the case for dall-e instead that would be for ADA

you'd get a very artistic version of the graph

[sleek hinge]

Yes this is what happen with most AI

[sleek hinge]

Thanks ill look into ADA

[late blade]

@sleek hinge You can also try the Wolfram GPT, their GPT seems good for that kind of stuff also

[sleek hinge]

Thanks bro, thats a big help

[lean iron]

What did i do wrong 😭

[late blade]

You didn't do anything wrong, it's the content policy that doesn't agree with that sentence.

If you add the kind of swimsuit you want to have, maybe that can help

[clever jungle]

be careful what you wish for

[worn bear]

Can someone explain me, what I should write to obtain a normal portrait picture? when I wrote about portrait or portrait orientation of the picture, ChatGPT generated me a vertical picture with the character, rotated 90 degrees

[late blade]

hey there, what was your prompt?

did you ask on a specific format?

there are square, portrait and wide formats, also as 1024x1024, 1792x1024 and 1024x1792

[worn bear]

Create a portrait-oriented illustration of a well-trained anime man with fox ears, sitting inside a wooden temple. Emphasize his abs, which should be more visible. His eye color should be blue, and his hair should be orange. The character should be the main focus of the image, occupying a significant portion of the frame. His attire is a mix of traditional and fantasy elements, featuring intricate patterns and vibrant colors. The background should be a subtle depiction of the wooden temple and the fantasy town, less emphasized than the character. The character's expression should be calm and respectful, and his posture should convey both strength and gentleness.

this is what was written in the prompt of the pic

[shut niche]

Say, "Create an illustration in 1024x1792 of...
Don't use the word portrait anywhere in your prompt.

[worn bear]

Ok, thanks. I will try

[worn bear]

nah, even in the new chat the result is similar

Prompt: Create an illustration in 1024x1792 of a well-trained anime man with fox ears, sitting inside a wooden temple. Emphasize his abs, which should be more visible. His eye color should be blue, and his hair should be orange. The character should be the main focus of the image, occupying a significant portion of the frame. His attire is a mix of traditional and fantasy elements, featuring intricate patterns and vibrant colors. The background should be a subtle depiction of the wooden temple and the fantasy town, less emphasized than the character. The character's expression should be calm and respectful, and his posture should convey both strength and gentleness.

[late blade]

seems about right tho, sometimes it's hard to do that oriantation, we have talked about that on and off here

[worn bear]

yeah. Once, i even asked ChatGPT to rotate the character on 90 degrees, keeping the picture proportions, but the result was the same(

[dim cradle]

that sounds like it might be over-complicating things? a custom gpt specializing in portraits (i made one of that purpose) can help, but that's going to happen sometimes no matter what i think unless the size parameter of the api is used. @shut niche makes a good point that "portrait" can imply other things. i'll dig up my custom instructions for that gpt in case i spot anything helpful.

[shut niche]

Click on the image, and copy the prompt that ChatGPT actually used. Output it here so we can read it.

[dim cradle]

isn't it up there?

[shut niche]

That's his chat, not his prompt.

He didn't say, "run this verbatim", he said, "please make me ...."

My assumption is that the prompt itself will tell us why it's rotated. Not always, but in 98% of my experiences, the issue is visible in the prompt itself. I'll see words like vertical, portrait, tall, etc. Combining them has a "double-negative" effect, and causes an odd rotation. Like saying , go left, go left.

"DALLE_2024-01-13_14.31.26_-_An_anime-style_illustration_of_a_well-trained_man_with_fox_ears_sitting_inside_a_wooden_temple._The_man_should_have_pronounced_abs_blue_eyes_and_or.png"
The filename is part of his prompt.

[dim cradle]

i thought the "prompt:" implied it was copied from the "i" which i thought was the prompt sent to the api? it dosn't look like a prompt a human would write. @worn bear where'd ya go? lol

[shut niche]

There he is!! 🙂

[dim cradle]

haha

[worn bear]

An anime-style illustration of a well-trained man with fox ears, sitting inside a wooden temple. The man should have pronounced abs, blue eyes, and orange hair. His attire combines traditional and fantasy elements with intricate patterns and vibrant colors. The character should be the main focus, taking up a significant portion of the frame. The background features a subtle depiction of the wooden temple and a fantasy town, but it's less emphasized than the character. The character's expression is calm and respectful, and his posture exudes strength and gentleness.

from i

[shut niche]

Try this, it's harder to explain why, than to give you my solution...
Start a new chat, using the default ChatGPT4, no GPT, no DallE GPT
Copy paste this:

Run this image prompt verbatim in 1024x1792,
"An anime-style illustration of a well-trained man with fox ears, sitting inside a wooden temple. The man should have pronounced abs, blue eyes, and orange hair. His attire combines traditional and fantasy elements with intricate patterns and vibrant colors. The character should be the main focus, taking up a significant portion of the frame. The background features a subtle depiction of the wooden temple and a fantasy town, but it's less emphasized than the character. The character's expression is calm and respectful, and his posture exudes strength and gentleness."

[shut niche]

I'm in the penalty box, or I would try to run this for you. LOL

[late blade]

@worn bear welcome to the club, you, hansa, yami, and myself are in the same boat, we've been fighting that kind of problem for a while

[worn bear]

Haha, thanks)

[shut niche]

OHHH

It's that bit about taking up a significant portion of the frame

100%

[dim cradle]

to be fair, the prompt instructions are for pronounced abs, and you can't even see them there

[shut niche]

I'm so confident, lol

[dim cradle]

it's really just missing all the important points here

[shut niche]

That's confusing DallE, because it's probably thinking, "Well, the only way I can make that happen is if I render it like this." Remove that line.

[dim cradle]

I believe it’s crucial we resolve these important and urgent matters

[late blade]

well, there are people that come asking in discord that gpt doesn't do what they want and they have homework or research paper due in less than 2 hours...

[shut niche]

So there's some concepts that take a little time to learn with DallE. You can "steer" an image with words. For instance, if you have a character that you want to be far away, but you start adding things like, "They have a cute little nose, with detailed irises, uneven skin tones..." etc., then DallE will have no choice but to pull the character in closer, as a closeup, even though you wanted them far away.
My best guess, is that your problem is similar. You've said, "The character should be the main focus, taking up a significant portion of the frame," which might mean to Dall-E, now I have to render this differently to make it work.
If that doesn't work, it's broken, ha ha

Oh, and for close-ups, I say just that, "A close-up of X in X resolution, blah blah"

[worn bear]

haha, in this way i tried to explain, that ChatGPT shouldn't spend 2/3 of the pic on the background bat to create a normal portrait pic, hahaha

[shut niche]

ChatGPT gets easily confused. The trouble with that is, once it gets confused, it often stays confused. That's why I suggest a new chat to test.

It knows everything, but is kind of dumb.

[worn bear]

I will try to play with that more) but sometimes ChatGPT generates really cool pics)

[shut niche]

Oh, it's a great prompt writer, although there are known issues there too. But it will elaborate for you, which makes the prompt writing much easier.

[dim cradle]

this will all come in handy when i do my koala master based on the beast master

[shut niche]

Try your character in this, but still remove that bit about taking up the whole frame.

https://chat.openai.com/g/g-EjZNAOcqa-moe-better

[dim cradle]

have you tried anime in that?

i thought it was tuned for realism

[shut niche]

haha, probably not. But it should work.

No, NoRender was the realism focused one.

This one has a bit of that, but isn't as much of a stickler. LOL

[worn bear]

OK, thanks, i will try)

[dim cradle]

i'm gonna try my concept with that in anime and see haha

[shut niche]

If I wasn't in the penalty box, I could probably whip up an Anime one based off of it in about 3min. I probably should, for other people, even though Anime's not my thing, per se. 😛

MoE better selects experts, it's all about CoT. So it should understand Anime inherently. That's what I designed it to do, be good at everything.

[dim cradle]

i thought it was superseded by moe.powerful?

[shut niche]

MoE powerful is my beta baby, lol

[dim cradle]

which should i use for my experiment?

[shut niche]

There's quite a few people using MoE.Better, so instead of mess with it, I've kept it in a "locked state" so I'm not messing other people up. MoE.Powerful was just my cheaky name for a new version so I could run new tests on new CoTs and new DallE instructions. MoE.Powerful is way more elaborate, and more flaky because of that. I'm still working out the bugs.

MoE Powerful might make better images now, TBH. It's getting pretty good.

[worn bear]

Meh, You've reached the current usage cap for GPT-4, please try again after 3:58 PM...

[shut niche]

OH man, you too!

We're in the penalty box together.

[worn bear]

it is ok. I once got a penalty for 24 hours because of too many pics, ChatGPT was unable to generate during a day(

[shut niche]

I'm free again at 8:34 Eastern.
I think most regulars in here have hit the 24hr more than a few times. You're not alone.

[worn bear]

haha, it was just a first time)

[shut niche]

Are you on Eastern time?

[open trench]

so many penalty boxes

[worn bear]

the most funny moment today)

[shut niche]

Yeah, the content policy is touchy... It also hallucinates. Sometimes ChatGPT knows, but other times, it has no idea why an image was rejected, and makes stuff up. I had it tell me an image was rejected because of long hair. LOL

[shut niche]

You can sometimes ask it to fix an image, and make it work within the content policy. More often than not, it'll fix it and you'll get an image. Trouble is, you might not always like the fix it applies.

[shut niche]

@8:34 I'll try to run your prompt for you, when I'm finally out of jail.

[worn bear]

Oh, thanks)

[empty kelp]

This is kind of interesting. The ones above are the process a prompt goes through, and the ones below are about the Transformer and Diffusion model that DALL-E 3 uses

(these are in order)

[worn bear]

wow, finaly right proportions and direction)

[dim cradle]

how did you make that one?

[shut niche]

MoE.Powerful

[empty kelp]

it’s tricky to create images in portrait format

that character has four ears, but that’s very good for hearing things

[worn bear]

10 min more and i will try your MoE.Better)

[dim cradle]

lol

[open trench]

finally updated https://discord.com/channels/974519864045756446/1193219783877988423

[shut niche]

I'm pretty sure the fox ears are a cosplay headband...

[dim cradle]

maybe chatgpt can help us understand

[worn bear]

@shut niche Thanks, it is working much better) 3 from 4 pics were generated in right direction)

[shut niche]

Looks like there's an issue for some, with Discord images loading ATM.

[worn bear]

yeah

[dim cradle]

this is like the dial-up days

[empty kelp]

there is a delay. something amiss with the server

[dim cradle]

at least then they were interlaced

i do suspect there is something afoot with the server

image was replaced with a vector graphic of a pile of doo doo

[shut niche]

Somebody generated too many Santa's helpers, and now there's no one left to actually run the server room at the North Pole.

[open trench]

lol

[empty kelp]

this OpenAI chat may be pushing Discord to its limits

[dim cradle]

i told you there were too many elves

[empty kelp]

that gave me a good idea for an elf image. brb

[late blade]

This is interesting, you can’t mention a certain royal emote in chat because of the filter. But when discord can’t load an image it has a placeholder with royalty on it.

[dim cradle]

Did it permanently lose my koala masterpiece?

We should open an investigation immediately

that's mint

[worn bear]

Meh, again in cap messages and the orientation is wrong, but the picture is cool, haha

[thick smelt]

a realistic platypus? perry the realistic platypus?!

[earnest flame]

I dare anyone to read that name without the accent

[late blade]

I'm lost, I don't see the accent

[earnest flame]

Someone dared!

It's a reference to a cartoon

[late blade]

ah, I'm still lost

lol

[earnest flame]

Yeah apologies one sec

[late blade]

dont worry, I know people in their 20's that never had a tv, and I met them yesterday

[earnest flame]

Not everyone has seen every TV show so no worries to any of those folks even.

This character has an arch nemesis with a very distinct accent in this cartoon

[late blade]

I know, I'm one of them

ahaaaa, see, I learned there's such thing today

[earnest flame]

Here's his arch nemesis Dr. Heinz Doofenshmirtz

[late blade]

oh nice

[primal sparrow]

Hi is there an estimation of when dall-e 4 will be available?

[dense mesa]

some time in the future

[late blade]

yes, sometime after now

lol

we don't know @primal sparrow

btw one of my names has an N, so there, the secret is lifted

[clever jungle]

browser extension???

[late blade]

no

[clever jungle]

awesome

[late blade]

normal interface just noticed it's an extension, thought it was vanilla firefox on this pc.

[clever jungle]

wow the koala daily theme was between the best there were so far

[late blade]

yeah the koala theme was awesome

[dense mesa]

man, I made some wild ones. only submitted one though

I'm curious how gpt4 retains some things. because I told it to make 4 striking images of subjects of it's choosing, and it's choosing the same scenes we were using days ago. not sure what would trigger that

example. the paint diva, lol

also, somber cyborg/robot mourning the fallen scene

[late blade]

I wanted to be more active with koalas, but I was away, had to go handgloves shopping, my old ones died

[dense mesa]

I don't even know what they were doing here. turned into longboyes

they have so much swag though. just look at them

[late blade]

@dusk steeple this #daily-theme message deservers it's own channel!

love it

[dense mesa]

holy guacamole. fed my gpt the new theme. these are interesting

I mean, they

re all good, lol

[dusk steeple]

Im just happy it got the words right ahah \😅

[dense mesa]

can't decide which one is the winner?

[late blade]

hard choice

go with the explorer

[shut niche]

Your killing it man! Awesome details in those!

I don't think general discussion images that get 10 stars go to hall of fame, but that wouldn't be a bad idea IMO, for an addition (if it's not already one I mean).

[late blade]

wonder who will be the first person to make a picture of a kitty's first encounter with catnip

[dim cradle]

lots of great stuffs. i just wish photoreal were not so cgi. but then again photoreal better in MJ 6.0 and i feel like i waste money on that sub because rest of it so poor compare to dalle3. maybe i am just never satisfy 😂

[late blade]

got to take into account that dall-e is a way different approach than MJ

but let's keep it at that

If they do a cat themed day, the server will blow up

but could be a good stress test for dall-e

[dim cradle]

well... i know dalle could make photoreal look incredible if they wish, they just seem to limit it. Bing is King for photoreal still imo, more diversity and natural looking

[late blade]

true, it's still a work in progress tho

chat and dall-e is mostly for brainstorming, I see it as a tool for that

[ripe wraith]

thats fine, chat and dall-e probably see us as tools for growth

[dim cradle]

i wonder if dalle will get any real update this year too... it seem gpt 4.5 or even 5.0 could be release but those of course will not improve dalle model (i dont think anyway haha)

[ripe wraith]

they're working on video too, not sure where that fits in with dalle research

[late blade]

yeah I know the feeling, right?

[ripe wraith]

the feeling of being a tool? yes, modern life 😄

[formal osprey]

I explicitly specify the format to chatGPT, either "square format" or "wide format". It will use the "portrait format" by defaut for full-body characters.

here's the instruction dalle sees:

// The size of the requested image. Use 1024x1024 (square) as the default, 1792x1024 if the user requests a wide image, and 1024x1792 for full-body portraits. Always include this parameter in the request.

we can deduce why it often use the "portrait format" for character design. The reason the pictures are rotated is a mystery. But that's not something we can control, it's a bug in dalle3, so I just avoid the portrait format.

[ripe wraith]

are you talking when it paints a horizontal image in a vertical size?

and leaves bars

[formal osprey]

no, I'm refering to the 1024x1792 format. often showing a 90 degree rotated picture

here's an example

[glossy scroll]

I like it. But it's very strange at the same time.

The off color bit is driving me crazy.

[late blade]

can't think anything else but kitties

[late blade]

mission accomplished

[late blade]

there are 3 scenes

4 scenes

going for inception?

[glossy scroll]

Inspired by your art.

[late blade]

my penalty box senses are tingling

[open trench]

lol

[shut niche]

That looks like Bleach, but he got ultra buff! LOL

[late blade]

I was thinking of a mix betwen bleach and hero academia

[glossy scroll]

It's his son, lol

New anime, Bleachy Academia

[shut niche]

He needs to eat some bread. 😛

[glossy scroll]

No carbs, only a pure protein diet

[late blade]

@open trench you putting scampers in danger?

[open trench]

scampers is always in danger. hehe. and so am I... in danger of the penalty box when I work on a scampers story.

[late blade]

lol

[glossy scroll]

I encourage you guys to get started on working on making specific style of art Custom GPT's for that GPT store profit.

I've made some dope ones that I'm not revealing yet. But they are very very good.

You guys have an upper hand because you constantly practice. That's why I say it.

[late blade]

I'm doing one but ADA stuff

[glossy scroll]

Ah gotcha.

Make some epic GPTs!

[late blade]

yeah have to

[glossy scroll]

We must. Can't let our talents go to waste. Let's make some moola.

[dim cradle]

lol did you specify 0% body fat?

@late blade ha your investigator from last night would work for today's theme

[late blade]

the koala? yeah

my focus is on kitties atm tho

[dim cradle]

the globe

curiosity. fascinating.

[late blade]

aww it was too good to be true, now content policy detective arrived

[dim cradle]

what happened?

[late blade]

wanted to do a sherlock holmes like cat and even tho it's in public domain, dalle refused, until I was very persuasive and managed

[glossy scroll]

Lol, nah I just said "make him super ripped"

[late blade]

interpreted prompt came out like this
without directly referencing specific characters

[dim cradle]

very good deduction, it's all elementary

[late blade]

wonder what's behind not doing sherlock holmes

@fading inlet what is that thing in #daily-theme message a huge lock?

[plucky hare]

This got me curious too, it looks like Sherlock Holmes has a more complicated copyright history than I'd expect, with relevant court cases and copyright expirations as recent as January 2023 -- check his Wikipedia article! Might just take a little more time for that character not to set off the filters I'd guess?

[dim cradle]

the topic reminds me of the old mystery book series The Hardy Boys

[shut niche]

Teaser... standard GPT prototype "dragon" test, lol.

And character test. 🙂

[fading inlet]

Curiosity... hum hum It is perhaps "The key to the mystery": Inexplicable event, enigmatic adventure... 🧐 🤔

[late blade]

I apologize, but I was unable to generate images based on the request for "Purrlock Holmes," due to our content policy. Even with the character of Sherlock Holmes being in the public domain, there are guidelines that I need to follow which may restrict the creation of certain types of content or character representations. If you have another idea or a different scene you'd like to visualize, please feel free to share it, and I'll be happy to assist!

So, for now, not possible.

[shut niche]

Coming soon...

[dim cradle]

actually i need some time to iterate on my pandora's box

[shut niche]

I haven't even tried to iterate... If it makes these kinds of images on the first run, iterations are going to be bonkers, because DallE saves the image concept with image ID's, so you can stack another few thousand characters (I'm not even sure how many, at least 2k) on top of the original scene.

That's epic btw

Indiana Jones!!!

[dim cradle]

it's got some vibes

it's with yours

i have noticed iterations with custom gpts can lead to some loss

i just love the metaphor

[shut niche]

I've NEVER ever had this kind of output before.

"Give her a mysterious name. I want you to have her on an adventure, exploring somewhere dangerous. Make your images progress, so that each one tells me a different part of her story. Attempt to keep the same dark dystopian mood that you've established here, so we can further develop the character."

[dim cradle]

that's kind of how i setup the tales from the gpt, the cryptkeeper tells a new tale with multiple illustrations, a whole narrative in a single response

that was just for testing, of course

i like it when i can give a daily the 10th star

[shut niche]

Whoops, I was so excited I shared in the wrong thread, bahahahaaha. Embarrassing snafu, but at least everyone here can see too. 👀 😆🤣
I'll hopefully be publishing this GPT soon. 😀

[dim cradle]

of all the likely excuses for a shameless plug.... 😉

[shut niche]

Bahahahaaha I wish I had thought of it!

[dim cradle]

eyes, magnifying glass... hint of danger, mystery, risk... and unimaginable rewards...

[shut niche]

In case anyone's interested...
Using the above images character, this was made with only this chat input:
"I want to see your ability to nail close-ups. Get in close to this character's face in these scenes."

[dim cradle]

that's the concept

now let's see ....

[shut niche]

Vision to Dall-E test with a twist. I told it I wanted the ship to be made of cubist stained glass and mechanical metal parts.

[dim cradle]

that's so creative, combines some great stuff

[dim cradle]

i didn't even share a prompt, and these pieces are all so absolutely fabulous 🙂

[pulsar sundial]

haha, yeah

[dim cradle]

sort of an AI-augmented Ancient Greek trireme

[dim cradle]

the dark waters!

[hot rain]

Trying to make a banner for a fantasy medieval house.

[dim cradle]

where are you in the process? it is minimalist...

oh, i was thinking crest, you did say banner

i mean, they were heavy on the symbolism, power, generational... but some were pretty simple still, like the house of borgia

[dim cradle]

i like the french banners on the right there

[dim cradle]

one variation on he-man's house banner 💪🤓⚔️

[dim cradle]

i'd vote to add "history" and "fantasy" as categories in the galleries. those two would cover bases that appear uncovered right now.

[glossy scroll]

Chubster Duckster

[glossy scroll]

My evil twin.

[dense mesa]

in the zone in the dall-e 3 lab

[pulsar sundial]

Yes Yes

[glossy scroll]

[dim cradle]

lol

[open trench]

random lol?

[dim cradle]

basically

[open trench]

lol

[dim cradle]

Morning, Nez. What’s on the agenda?

[open trench]

I do not know. I was going to do some scampers stuff, but I'm running into a creative wall. So, kust brainstorming I guess. What about you?

[dim cradle]

Just playing around, whatever I can get into.

Maybe more daily concepts

Curiosity is a big topic

It sounds like dream-cultivating stuff

[open trench]

It is. BTW your Dixon Hill log entry/ daily theme was fantastic.

is fantatstic

[dim cradle]

Haha I’m glad you liked it. Gpt writes some good content

like sometimes after generating i'll ask it to caption it, either as a curator, or speak to some points, or in that case, write a little personal log with a little nudge in a certain direction.

that was came pretty naturally as it was generated from my holodeck gpt haha

i usually forget to add some kind of commentary, and that can be another fun aspect to the thing

[grizzled iris]

Happy Sunday Y’all 🙏🏽

[thick smelt]

happy sunday to you too

[open trench]

14 degrees here

[dim cradle]

yikes, 30 here

[grizzled iris]

5 here lol 😂

[open trench]

omg

[grizzled iris]

Todays theme sounds good 👌🏽

🔍 curiosity - exploration, adventure, discovery. come take a peek!

Should get good results from these prompts 🫶🏽

[thick smelt]

here's a yummy looking ice cream sundae (pun intended)

[grizzled iris]

Oh wow 🤩

Now I’m hungry

Remembered I’m still human and gotta eat now 😂

[grizzled iris]

Love the lion 🦁

[grizzled iris]

Awwwwwwww 🫶🏽

[grizzled iris]

This is quite cool 😎

[grizzled iris]

Wooaahhh 😯

[grizzled iris]

Wow, you guys are getting better at this every week 😂

😂😂😂

[grizzled iris]

Personally I love this one ☝️

[open trench]

OOOh. I just got hit with Scampers inspiration. Nice. I shall return either totally frustrated or with an update lol Depends on how I'm able to prompt.

[grizzled iris]

Prompt : Good afternoon! In today’s theme of
“🔍 curiosity - exploration, adventure, discovery. come take a peek!” Could you make an image of a free universal healthcare service as my goal is to create an equal world where technology helps to save and improve people around the globe, regardless of their wealth, background or religion, and where such access to technology is equal 🫶🏽

I believe that by improving access to technology and building trust within communities, we can together eradicate the outdated, unfair and biased legacy systems which still exist in society today. You can use this image attached as an inspiration for your works. Thank you x

[late blade]

that's not curiosity, that's 100% surveillance

[dim cradle]

that's precisely what the prompt is about!

[late blade]

the thing is, people are not scared, that's disturbing

[dim cradle]

Curiosity of Big Brother

Finally got around to throwing together a gpt just for crafting art prompts for the api -- just the right length, format and instructions for a copy & paste job.

[open trench]

I have no idea where to start with a GPT

[dim cradle]

I've made about 30 so far. quite a few more on the list. i created one for tasks i tend to repeat.

[thick smelt]

perfect illustration of a 1984 book re-release. I like it! it's dope

[open trench]

I guess I could attempt to make a Scampers GPT. I enter a lot of the same information over and over again in my prompts.

[dim cradle]

perfect use case

[open trench]

I just don't know how. I guess I have to explore a little.

[dim cradle]

or, since it's a process, you might divide it into several tasks, where each gpt specializes in one step in the workflow

one for the narrative, one for the prompt, one for the image generation, or you might combine them, it depends

[thick smelt]

now even the animals is going to have superpowers

[dim cradle]

They’re just all curious. They only want peaceful coexistence….

[dim cradle]

There’s lots of ways to go about it, yes explore and eventually you’ll settle on a flow and structure that works for you. Maybe identify a task and play around with the GPT Builder.

[open trench]

It's all greek to me. I'm already overwhelmed. I'm sure it's just my brain getting in the way of itself.

[dim cradle]

one step at a time...

https://chat.openai.com/g/g-1lzU4mdPk-art-prompter here's the thing i just threw together to get decent prompts i copy to the api. just enter your concept/theme in a few words. you can use the output for the api or tell it to visualize. but the output quality should be pretty good.

[open trench]

@dim cradle your big brother post reminded me to post this picture in the bonus section of: https://discord.com/channels/974519864045756446/1192809271189053541

I'm so out of league with this. I have no idea what you just told me. lol

[dim cradle]

We might need a Big Brother gallery. Might be something on dystopia already probably.

[late blade]

well penalty box

[olive sonnet]

Hi

[late blade]

yo

[olive sonnet]

Nice,I can't make a sci-fi ebook cover using AI image generators like Leonardo AI and Imagine AI. The results aren't what I want. Can you help me create a good cover for my sci-fi ebook and non-fiction ebooks using Dall-E (in Chatgpt)

[late blade]

what do you mean can't make?

what is your idea?

you have a concrete description ot Leonardo (assuming davinci)? The imagine Ai?

[olive sonnet]

I don't have chatgpt plus

[late blade]

ah

that's a challenge

[dense mesa]

bing image creator

[final compass]

almost there, but not quite 😄

[dim cradle]

Flawless.

[olive sonnet]

I have tried it 3-4 days ago but I can't use it for my ebook cover as it will damage intellectual property of Bing AI image generator's terms of use

[open trench]

@dim cradle did you see my sort of dystopian city invasion post from before? https://discord.com/channels/974519864045756446/1184310287344611500

[dim cradle]

oh no, will in a sec, buddy

[thick smelt]

chilling scarecrow taking a break from guarding farms, even them needs a vacay too 😆

[dim cradle]

just curious, any reason you don't have a chatgpt sub for your work?

the concept is neat