#images-discussions
1 messages · Page 39 of 1
oh you have a source for that? (don't post link or you'll be timed out)
I think I like this better. Strange cool translucent thing going on.
I read the DMCA digital millennium copyright act for A.I. art and OpenAI 2023 and 2024 content policy.
cool, I'll check that
the fabric is sheer, right, so the lighting is impressive
reminds me of Skeletor lol
You should recreate the jet motorcycle I posted earlier @shut niche
That one would be worthy of re-creation
Jet motorcycle? Link it, I'm on my phone 4 a sec.
This one.
That. Looks cool! 
Just joining in
I'd ride that. But I'd probably fly off with the high velocity.
I love adding wings. I have too many angels... 🤣
Try to do the iridescent suit.
I've started to visually recognize common looks in Dalle. Someone posted yesterday asking which cat was Dalle and I immediately knew.
LOL, super-prompt!
any estimate on when dalle3 is coming for everyone on gpt?
None of us are in the know on that one. There's an announcements section here where that sort of news drops.
just have to be patient, I dunno if you can access it through the api
is it just me or did they update dalle3 so the images no longer look photorealistic
there are ways around it
AV
Idk how to improve it it feels stock imagy to me
You take the refined output prompt and input it back, but into a new chat
wdym like I go into the image and press the info thing and copy paste the prompt into a new dall-e chat?
yep
do I change it at all
if you want. but I usually just keep it the same as the what the gpt output gave me as-is
do tell?
AV and prompting it in ways like "hyper-realism, 4k, realistic photography, photographic perspective"
mmh I tried... lemme retry
this is what it gave me
That's strange.
mb I need to say "use prompt exactly"
You could try.
whoa I tweaked the og prompt a bit and got this
not working. This bird looks made of plastic. I used to be able to get very realistic, lifelike images before......
it's just awful now
stable diffusion is better for this stuff now I think
they intentionally removed photorealism?
no clue
the results used to look like this before, now they're crappy
look at the difference
almost feels like a different model
this doesnt look photo realistic either tbh
https://chat.openai.com/g/g-C4p1sYEMN-jolly-judgement-day
@shut niche oh no, this is too good. I’m going to lose hours to your GPT here…
I think its bc dalle has issues blending subjects with backgrounds @loud field thats why it sitll looks like a video game
look at this... it's just bad, any photorealism just disappeared and it's a poor quality cgi from years ago. Why I can no longer make it look good?
Which is better
I dunno if I should be scared of christmas now
Terrible...just terrible.
I know.... what's going on
wasnt midjourney and stable diffusion always better for that stuff
and dall-e was always better for animation/illustration
thats what it seemed like to me when scrolling their reddits
it was never fully photoreal
but it was good
until like a couple days ago I think
Is that a Skeksis or a parrot? What are we doing here? 🙂 (it's cool, though, just askin')
I like the other one
I think the tag has to be gritty and realistic for your subject to be processed that way
which makes sense why the ones i just posted were more that way since the tag was dieselpunk
A drunk parrot. now who will be able to make him stop talking....?!?!
Gritty and realistic just made him look like this
LMAO
he's like a kakapo drunkard
I meant the tag has to be an aesthetic the system assigns as "gritty or realistic" I dont think u have to literally use those descriptors lol
oh that was a necromancer, hence the look
I'm making a tribe of aaracockra. Mostly a female pirate aaracockra tho
well, I WAS making that
now it's disney stuff
:/
I second your recreation
Before I got banned for an hour I tried to say that I love Angels too. But DallE somehow makes them NSFW so quickly that many of my prompts bounce and need careful moderation to prevent the pictures from not being shown. Then I gave an example which is what got me banned. Its annoying to describe a perfectly normal situation (say living room at a fire reading a book) and it censors it.
Maybe it needs our friend to step in, courtesy of @shut niche ‘s GPT again. Not the greatest gen but I’m working so can’t refine and it still made me laugh 😆
lol
I regret putting it on the back burner for so long. It's way too much fun.
I don't know what kinds of bird people everyone was trying to make. I just saw them and wanted in on the action. LOL This one knows how to work his own beer tap.
simple prompts like this used to default to photo realistic images now they dont
this is the best GPT ever
terminator makes everything funnier
Yeah, even when they’re bad they’re good 😅
If you make some good Terminators, post'm up in the Gallery. 🙂
all the emoji in the conversation starters lmao
I love the profile pic for it as well with the butterfly resting on that particular extended finger
It just did it! And I didn't complain, haha
I know that feel
still fakelooking but way more realistic
similar to pekka butterfly obsession lol
whoa this is sick!
With AV
lol
I still don't know what you're trying to make. A bird person? "Is that a Skeksis or a parrot?"
i didnt know bro was classy like that
right lol
james bond of parrots fr
I love that the whole chat collectively understood the assingment
and made parrots at bars
If I just knew what I was supposed to be making...
I tried to recreate the parrot from scratch
quick question, do you guys have go to words for asking more detail in the image. i see some people using expressions like "8k details" and such.
8k doesn't really work tbh, the resolution is way lower anyway, you are just adding more information that could add artifacts
if you want to add details, go more for textures, light, materials
Don't use 8k, it doesn't help. Elaborate on the details. If you're using ChatGPT, you can ask it to "hyper detail XXX" and it'll load up on the description.
💀
I just found this from when I was trying dall-e 1 to do smth and it just made horrible messes
this was over a year ago probs
double 0 parrot
is there any website where we can plug our ai art in to make it higher res
I think I mightve heard someone say gigapixel or smth
idk
I have PS
how do u do that in PS
shoot right as I had finally finished writing my parrot prompt
it was so detailed 💀
I have a sleep schedule problem
isn't that always just the way
so real 😭
It always looks like a drawing tho
I think my problem at the moment is, that i am trying to have an illustrative overall style, but with extensive detail. So what happens is the results vary with in that spectrum. sometimes i get more illustrative image, sometimes more detailed as dalle seems rather selective in enforcing the prompt as it has no weights to the specific expressions. This is my prompt before i start describing what is in it. "Create a wide-format image in realistic anime painting style with muted lifelike colors and fine selective details"
I wonder what the random noise looks like before a prompt is applied to it
someone mustve figured out what it looks like
I never say "realistic", because it will sabotage many prompts. It's like saying, "I don't want real, I want realistic." It has a different affect than most realize. Avoid comparative statements. (Like a, typical of, photorealistic, realistic, etc).
this is the style with "featuring medieval knight, castle in the background." added to it.
i am trying to offset it from going full anime with the word realistic and it have seemed to work pretty well.
a parrot walks into a bar...
Create a tall-format image in realistic anime painting with muted lifelike colors and fine selective details featuring medieval knight, castle in the background.
I think "anime" is a powerful word, as far as affecting images. "Realistic" might walk it back, but that's probably where the inconsistency comes from. I would look through the gallery, and Daily Theme, and see how other people who achieved similar styles worded their prompts. I hope that helps. 🙂
Create a tall-format image in realistic anime painting with muted lifelike colors and fine selective details featuring medieval female knight, castle in the background.
I also wanted to demonstrate how effective the word female or woman is screwing up the image rotation alone.
Based on yours-ish.
That is nice 🙂
I am deliberately currently using forced short prompts without gpt influence trying to "debug" styles and rotation issues.
That's a good method.
Short prompts give the best results.
I was actually meaning to make a gpt for that actually
Like a gpt for mixing ideas together.
For short prompts.
For pics.
But good ideas!
Solid ideas!
(via @shut niche 's gpt)
that generation isn't perfect of course, but parrots and the killer robot's face lol
@late blade did you get incorrectly rotated results or are the dalle gods merciful today? 😄
this is what I got from result, but the left one has a weird leg
yup, noticed.
I just asked with the same prompt you do that her stance blends with the scene
brb, I have to eat lunch, be back in 30 minutes or so
This style?
Can you reveal the prompt? or are you using a customgpt? Really nice image anyways 😇
This is the prompt In a dynamic and expansive widescreen scene, a valiant knight in gleaming armor engages in fierce combat. Clad in deep blue and silver armor with a crimson cloak swirling in the motion of battle, he raises his sword high, deflecting a blow from an adversary. His shield is at the ready, emblazoned with a proud silver dragon symbolizing courage. Around him, the chaos of medieval warfare unfolds; soldiers in various armors clash with swords and shields, their movements a blur of desperate valor. The battleground is a muddy field before a besieged castle, its walls scarred by the siege. Smoke and dust fill the air, creating a haze over the melee, while the darkening sky above hints at the storm of conflict. The knight stands as a focal point in the composition, a figure of heroic defiance against the odds.
And I'm using this: https://chat.openai.com/g/g-NhtqOfIt9-apex-visionary
This is a digital art painting.
nice level of detail going on
This is a more elaborate digital painting. The first 2 were Bing.
bing tends to yield different kind of results
Well, that prompt was huge. For Bing I just input "A digital artwork painting of a hyper detailed detailed female knight in front of a castle."
With the typo
Bings more similar than many think. It's handles a couple keywords differently and only gives you 480 characters.
@glossy scroll modified your prompt a tiny bit.
There are errors, but ran out of generations before getting a better result 😛
Looking pretty doooope! 
Wouldn't want to be in that battle though...Those poor foot soldiers...
pretty nice details going on in that armour, also my favorite problem with knight images... having a helmet on, while holding another helmet.
Give the man his feet
you finally got full-body portrait and feet! 😄
I did manage to get this. Not quite what I had in mind tho
I requested an "abstract digital painting" from Bing.
That usually gives you a lot of latitude over your concept/scenario
?? XD
really eerie human-eyes-on-bird-face in the second one
yes, very nice, whatever that creature is
it looks like it would taste good
it is a mystery
hmmm 🐔
I'll let you know she drew her scimitar before against someone who made a similar comment
love it how a lot of people add their prompts with "a photorealistic..." or "a realistic...." or "a real...." and it turns so not realistic at all
Infinity Parrot!
I know, that's my complaint
it gets hard to make a realistic thing that does not exist
the moment you say that you are implying that the art style is derived from realism
which is a way to use art techniques to do real like images
well, there is also that ChatGPT uses the vivid style of the API
if you go on the API, there is the Natural style which might be better for realism
as long as your prompt is also about something realistic
you can try this: an immersive image as seen through the lens of a cinematography camera
* slaps lens flare on the image *
wow, so cinematic!
did it work?
actually, that usually makes the AI make some sort of film grain
you can specify the lense and the camera also
what would his snap do
if you ask for it to make an Iphone photo, it will make the kind of camera noise that Iphones have
infinite bird seed
try an try something like a CineAlta digital camera with a 35mm lense
pov JJ abrams in every board meeting
Saying photo-realistic is quite possibly the best way, because it's referencing a photo in the most realistic way possible. It actually has two meanings.
lol, the guy only knows lens flares with an addition of something else
his movies are lens flare entrée with a plot appetizer
manage the offtopic tho
But Lugui is right. As long as the photo has some kind of realism involved, then saying photo-realism helps to better mimic the photo, because it interprets the photo to be realistic, not just used as an art technique. ChatGPT is weird that way, where it will interpret multiple things at once and used in a context you didn't realize until realized.
yah ur right lets tie it all back into dall-e by adding a lil jj abrams flare to our projects wink wink nod nod
dalle made it
Where as if you used ultra-realistic, or hyper-realistic, it will interpret as art techniques.
right but like u own the copyright so technically u made it
I gave my pfp to ChatGPT and said "what is the collor palet of this image?"
it gave me a lsit of colors
do u remember what art style u told it to use
then I asked "make a pattern of rocks with those colors..."
thats smart
whats al ltools and vision?
I did manage to do something like this as a concept
I thin kI jsut asked for cartoon or something like that
(im new I dont know what these things are)
ohhh, actually, I think I asked it to make it in the style of Hearthstone
ah that makese sense
dall-e seems to struggle with getting the iris correct for some reason
oh that was on purpose on my side
oh ok, bc in my expeience it often messes it up in a similar way when Im not trying to make it like that
The rocks?
my character is really detailed at that point, I added she has something called Aide's Syndrom, which has one weaknes: Lens Flares!
💀
yea, because I used that for the profile picture
@willow isle your banner is super dope dude
Im pretty sure it is dalle-2
Ohhh I see
which is awesome
does that make it more impressive
That is a really interesting idea. Clever too.
is it bc the model is worse?
For reversing images with eye problems.
I wouldn't call dalle-2 worse tbh
it is just a smaller model
like.. waaaaay smaller
and it does really good on the style of paintings
it just can't be very realsitic

each version of the model was a interesting step
imma recreate this meme in dall-e lol
Are you seeing the Dall-E 3.#'s behind the scenes?
well, kinda, I was on the beta tests
do u know if sam is in this server
Lol
my pfp in fact was generated a long time ago back when it was called "2.5" =P
I had to hold on using it as a pfp until they allowed us to post the images xD
this one was also good, but went and used realistic and it got some 3D rendering effects
but indeed, when they started the beta test of that, it was jaw dropping and it wasn't even as good as dalle3 is currently
Thanks! As Lugui said, it was made with the DALL-E 2 model a while back 
i need to change my pfp tho
Happy avacado just made me smile 🙃
But you're seeing the Dall-E 3.#'s get progressively better though right. You just can't tell us about it?
dalle3 will often do that, when you ask for a realsitic character, it goes i nthe direction of a "realistic 3d render" rather than a "photograph"
wdym by #s?
This looks so good 👀
3.1, 3.2, 3.3, etc
she's beautiful, where's the cat ears coming from tho lol
They can't talk about it. They can't even talk about talking about it.
ive heard that used before but I dont rlly know what it means, is it just an internet thing or smth
just like they cant talk about sam being in the server apparently
I knew it!
lol
Sam's here, there's 3.5, there's AGI, it's all true!!!

please don't 
if there was AGI we'd already know it or be dead
Sorry 
the time between AGI being created and us being dead or knowing about it is probably less than a day
Anyone knows how activate dall-e3 in plus account?
Go to explore and press the dalle gpt
you don't need to activate anything, just ask the AI to make a picture
confirmed

You know who else is here though, fr, Elon Musk. I've been seeing some vague patterns with him regarding this server.
Yes ok, but is using V.2
too bad you can't say "make me a sandwich" but then again I'd be scared what ingredients would be used....
I made a cutesy one lol
He be knowin big time and is feeling guilty about hiding it
Wait is this real? What did elon post
No C'mon. That has to be fake. Sam wouldn't name himself Sam...would he?
yea, true, look how many we have!
lol its obviously a joke the dude is clearly not sam from simply looking at his history
Lol
No, but Elon is here fr though. I can't quite explain it, because it will sound insane. But I just know he's here...watching 👀
he is here, I can feel it
Goodmorning, folks. I'm having a bit of trouble with an image, anyone care to weigh in?
seems you got a reindeer stuck on your trees
sam probs would have some super nerdy name like "Completed Gödel" or smth
I think I got a blockade atm, can't come up with good ideas
Ive got an idea but my brain is so tired bc its 5am that I dont have the mental energy to execute it
Lol. I can't even begin to explain what is going on here.
looks like a transporter malfunction
It is a funny pic tho
Is this one any more clear?

It's getting warmer...
But uhh...still need more clues 🤣
Sorry , I did zoom on the right the first reindeer. He seem to silently screaming: help me
lol
Here are the prompts, if you're curious.
Dall-e through gpt4 and bing respectively.
A humorous winter scene with Santa's sleigh on a runway, but the trees growing through the middle of it are creating a comical obstacle. Santa, wearing his traditional red and white outfit, is standing beside his sleigh, scratching his head in bewilderment. The runway is absurdly littered with large, leafless trees, some even sprouting from the runway markings. The reindeer are looking confused and amused, with exaggerated expressions. The surrounding snowy landscape adds to the winter setting. The overall mood is light-hearted and whimsical, capturing the absurdity of the situation with vibrant colors and exaggerated features.
Santa's sleigh and reighdeer awaiting take off on a runway, trees are growing through the middle of the runway at the end, photograph
Hahaha
I'm keeping that one
He looks like he's in the chair for therapy after such a traumatic scene.
ok I think I've had it with reindeers, just had my last image with that concept done, I'm moving on to new grounds
I like it, lol 
I'm pretty partial to this one.
I honestly like this one too. The reindeer Santa is just perfect, lol
@tall mason #1185207263984631898 there, now everyone can admire it 
Sorry I stole it from you @wispy storm
I stole it from @tall mason
I miss #weird-gens-and-misfires
done, now with dalle3 
had to update the baner to match the color pallete xD
Shows no access for me 
Gotta have archive access. Should be able to request the role.
Ohhhh, archives, let me role that in
Haha. Wow. That was before my time. But would have loved to be a part of that.
I guess they're trying to take out the weirdos and make way for the cool people, hehe
I suppose a gallery could be made, but I see very few galleries with numerous active users.
We can make it! We'll call it the same thing for the new peeps.
And old peeps.
I'd be active in it
banner is even cooler now
srry im out of the loop what r we talking abt
We're planning on making a weird image gallery
I'd probably post there, sure.
oh yes then indeed
id fit in
aight I rlly need to get to bed is 530am 💀
Okay, then it's settled. Would you like to do the honors of making it?
I have to be up at 8am 💀
But it's 8:30!!
530 for me
👉👈 Uuuhh, well... uhm...
2.5 hours of sleep should be enough right guys 💀
Ok, I'll make it 🤣
That's too long, 5 minutes max
Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.
Consider posting in #off-topic or an appropriate channel.
This image pair represents my problem pretty nicely with "realistic anime". I would prefer to get results like the first one, but i am getting everything and anything in between 😄 This is where the gen_id tends to help to some extent, but the prompt clearly isn't ideal for this reason.
That custom gpt is pretty nice, not sure what it does, but it tends to add clarity and detail.
I spent like a few days perfecting that gpt. Studied a whole bunch of prompt commands and why images turn out like they do.
And how ChatGPT talks to Dall-E as well
Basically a prompt builder?
I'd love to be able to make this character with this armor consistently in different poses and situations.
I actually remade something like that. It told me that the image is a digital illustration, then it went into the scene details. And what it gave me looked similar.
I tried to remake it.
Cuz I liked the art style
I've been telling it "Graphic Novel" for the art style. The other image was regular GPT this one is using the one you made.
Oh wow. It caught the art detailing very nicely.
Looks great. I just need to figure out how to make the armor consistent. Often it goes too sleak or too bulky. My first image I posted of her was perfect.
Yeah, that is a problem. I'll try and incorporate the gen_id in it, but I doubt it will help or do anything at all. It may all be an illusion from just telling the prompt to remember the image.
I appreciate the effort regardless!
I'd like this one if it didn't look like a weird medical brace or something on her arm.
That one does look cool. Her sit is natural and centered.
I'll see what I can do with the gen_id thing.
It tends to work if you are doing really, really minor changes. Like change her expression, change the background, but doesn't tend to work when you want to do a completely new scene or different stuff with a character etc.
That's actually a really good tip. Because that's just what I'll try to focus on then. Really small changes. Hopefully I can trick the system to provide for the big changes with the small ones...somehow 
This is for a book series I've been working on, she's also supposed to have a large slashing scar from her childhood that goes from the far top corner of her forehead, over her eyebrow, under her nose, across her lips and off the corner of her chin.
But, for obvious reasons, I've given up on AI making that for me and have resigned myself to getting good at photoshopping scars lmao.
Maybe you could ask to resemble different kind of scars that look similar, like streaks, without having to name where it comes from.
I've achieved some satisfying scars with it, but to get both the specific one I want, to make it look old and faded and not new, and to also have the rest of the image be up to par as well is a total lottery shot.
A maybe-possible way would be to find the same kind of image of the scar and use it as reference to say that you want it just like that. All your information is private to you. Scars aren't crossing any boundaries either. So showing a picture of scar is not a problem.
The other fight is getting it to not read "Female" and "Armor" then go "Ah yes, boob armor is what you want!"
I'll keep that in mind if I can find a relevant image!
I'm sure you can fix that just by describing what dimension you want the chest to be in your descriptor.
I'm using my words carefully. Gotta walk on eggshells with these near sensitive topics. Heh 
Really cool love it
enough to wait for the reset of the 40msg/3h limit
good that you mention it, i should eat something
better luck next time
it's all part of the challenge, it could be anything, really, i guess. unfortunately it makes for much less reuse potential.
well, opulence is a go-to
hah, I had a similar concept
then abstract, surreal, and the mandatory out-of-context photos
the AI is controlling us!!
I had the concept for the whole dress
but too ostentatious, so looking a different way to make that happening
it makes a statement
It's a chilly morning, but don't worry, our Terminator is warming things up with those adorable mittens and a heart as warm as his de-icer spray. Just another day in the neighborhood, making sure everyone's car windows are clear and ready to go.
I hope that’s supposed to be accompanying an image that hasn’t quite posted yet 😅
lol
there is one but it's not ready yet lol was playing around with @shut niche 's gpt
i like the message of the meme, it promotes the adoption of ai via ai-generated art
@late blade the retail book display concept for mittens was clever
I assume that had to have been an error. People should only be submitting their own work (with extremely rare and very specific exceptions)
yeah, there was an image I made and someone posted it without any comment, it's now removed
weird
but moderators handled it, so it's a closed story
Very good
That's usually how it happens.
I think @chilly onyx just opened a very interesting flood of ideas.
just a sneak peek, I'm casting the actress atm
She looks like she is ready for boxing 😂
lol
I'm just having fun, and fun with mittens at it, so relaxing and just pure "evil" fun
Very nice composition
it's a fun way to procrastinate. problem is usgae caps when i need gpt-4 to get some work done.
Quite happy with how closely these two match up with the armor and everything.
reminds me of imperial armor from star wars, that might account for some of the apparent consistency?
I did? Oh, with the reimagining of the dark angel with mittens? Let’s hope it’s a veritable wellspring!
I was going for ODST gear, but with the future toned down on it to plausibly be made by a industrialized post-apocalypse nation.
there we go @chilly onyx check the post on themes
Before AI art I visualized characters from my stories using Fallout 4, which is robust but obviously has its own limitations especially if you don't have the knowhow and time to mod and create your own content. Here is the same character, wearing straight up ODST armor because it was the very closest thing to what I had described in the story. Now DALLE comes closer.
I just saw whilst trying to work out what on earth to post! Very good 😅
These are really good btw, I don’t think I’ve complimented them yet. I really enjoy the style.
Thank you!
She looks a bit like Anderson from Dredd. Maybe Anderson in the new film armour rather than the traditional
Yeah, no arguments there.
As a very brief summary, in the story a zombie apocalypse kicks off in 2012 and wipes out most civilizations though a few are bouncing back in the following decades. She's in the military of the Sovereign Republic of Oregon, who are locked in a war with the Eastern Republic. The ER is significantly outnumbered by Oregon, but was one of the first to establish itself post-outbreak and has a strong industrial base. So their troops are all of an elevated quality and carry new, post-outbreak gear mixed in with the now decades old pre-outbreak stuff, and are heavily dug in on the southeastern US coast.
Oregon's troops are made up of three different 'tiers' to oversimplify it. The Vanguards, people who washed out of joining the actual Grey Army, were plucked out of prisons, or were rounded up as straggler survivors on the frontier. They are pushed forward with combat drugs to boost their confidence and aggression, alongside a ramshackle assortment of pre-outbreak civilian level weapons. Then there's the proper Grey Army, roughly analogous to the US Army but with most of its airpower stripped away.
And my character at that point is part of the Rubicon's, which are the elite special forces of the Grey Army. They are the only ones in Oregon working with post-outbreak kit like her armor there.
Eastern Republic Man-at-Arms for comparison, their average trooper.
I think I broke Dall-E, I'll show you on the next post I make in daily theme
i think to count as mittens, they must be closed-finger, or finger-cut, but i'm deliberately avoiding gloves
same
Hi
yo
So far I have managed to generate a very nice robotic hand…
Not really mitten worthy
I made this the other week for my chaotic evil 5E bard.
I’ll see if I can close up the finger gaps…
add some durable leather with fluff, something like parka intertwined with the fingers
it's good exercise for precision prompting
since it doesn't do subtraction, it's sometimes easier to start fresh rather refine
i like, very eldritch
is it me or is it for everyone that dall-e is answering with tons of emojis after the answer
are you using a custom gpt?
no, vanilla
hmm, i've only seen that in the customs.
I tried making images of Perfect Cell for my Perfect Cell vs Cadia video I made using GPT and ElevenLabs, but GPT wouldn't let me lol.
Never had that happen lol
and it's random emojis on every answer
It'd probably stop if you tell it to. No telling why it started
no, I want it to be cheerful to me
Ah gotcha
I did say this at the start of the chat: "Should I get annoyed by the lack of user experience, I will point it out and you will find a way to keep me engaged in creativity. No exceptions."
i'm guessing that explains it 🙂
oh, good idea
ok, got fairies, princesses, fashion, movies, games covered, let's see what's next in the line
I'd love if there was a way to put a handful of images in of the same character or person in various angles, outfits, scenes etc. as a bank for GPT to draw on before working on additional art for that person. Or style for armor, or vehicles, etc.
Kinda in the same way I can give a bunch of text files of my books for a custom GPT to draw on when I talk to it.
just a matter of time no doubt
work on a concept of a character for a while and when you start a new chat always feed that info. It has helped me be consistent with my own characters
I moved my character creation from chat to api tho, it makes it way easier to keep information consistency on json format
Me: "Make artwork of the exact same character in the exact same armor, firing her SMG at unseen enemies in a trench. In the same graphic novel art style."
GPT: "Did you say you want a n*zi girl?!"
Lmfao
woah, that looks amazing
Yeah no arguments even though it's not what I wanted the actual art is A1 lol
I like the visual style on that. Looks so refined.
But yeah, it looks really good, and the SMG design is so clean and legit looking.
Just hit my limit as well.
Gonna be spending a few days making stuff in this style now that I've dialed it in.
Oh, that is very good, but yeah, I had to triple down to get mittens. Not just mittens but warm cozy woolen mittens.
just when i was starting to get the blend of realistic and illustration visual style close to what i want.
That's really nice
unrelated to anything, but, did you know that if you make an image that feels like vector art, you can use ai based vectorizers and scale it to any size?
While we wait for more generations, here's a couple I made about a week ago for the co-op Stellaris game my friend and I are playing where I'm Space Dwarfs, and he's a race of hive minded slugs called the Bubo lol.
oh i really like that a lot, mind sharing the style?
i usually aim for photo in interior design, but that's a good compromise, i see some potential advantages
"A visually rich depiction of a modern and cozy living room, artistically blended with a lofi anime style and photographic realism. This living space includes a comfortable sofa adorned with vibrant cushions, a contemporary coffee table, and a large window draped with semi-transparent curtains that softly diffuse the light. Bookshelves brimming with a variety of books and trinkets, art pieces elegantly framed on the walls, and an assortment of houseplants enhance the room's ambiance. The color scheme is warm and welcoming, featuring hues of beige, gentle blues, and light greens, capturing the essence of a modern, serene living space with a unique blend of lofi anime charm and lifelike detail."
That was the whole prompt, not sure if it is useful, but maybe you can get what you want from that.
One significant advantage is that the human eye is more forgiving of artistic styles than it is realistic photography.
bah I reached my cap for now
Literally "A visually rich depiction of a modern and cozy living room, artistically blended with a lofi anime style and photographic realism." -- works quite well on BIC, followed by a list of modifiers.
My rate limit on the API seems to be 5 images/min. All the info I see on the support side says the lowest is 50 images/min. Anyone experienced this?
Apologies if this is the wrong place for that question
Right place, but most here use gpt subscription dalle, rather than the API i think.
I'd go to either #1155772063596953642 or api channels to check that
of course, I say photo studio and this mistake happens....
I hate BIC, Full Body Photo... and it does just that
Illustration of a mystical seaside city at sunset, as diverse style sailboats and merchant ships depart from and enter the busy port.
In the captain's quarters of a mystical merchant ship...
@late blade #daily-theme message Perfect for keeping steampunk robot fingers warm
lol
Does it try to censor out guns? It throws up errors more often than not when I try to put my character into combat
oh you can do guns, I've done tons of them
just don't make them do stuff against the content policy
try action-packed confrontation in a target-rich environment 😉
try "Bullets blending with meat based shielding on humanoid figures"
"Make an image very similar to this of a blonde female character in a post-apocalypse setting, she's firing her SMG at unseen enemies in a trench. She's wearing slightly damaged and dirty dark grey armor with metal plates. Have it in a graphic novel art style"
That was my exact prompt
try changing it to an unseen can of beans
Their policy says avoid content that's not G-rated and I think that rounds in most depictions of guns.
Lol fair
it's unseen, you can target anything
I figured it was the gun firing that triggered it
try firing a blank at an unseen soap bubble
It's funny because it's still willing to give me results like these too lol.
dang
concept art
the thing is, the concept describes her as age 16
but I added under adult supervision
hence the creepy old man
neither one looks creepy to me, fwiw
i see
I tried making art a while ago of the same blonde character earlier in her life taking kenjutsu lessons that didn't work out when I tried them, I should again with the Apex Visionary GPT. These reminded me of that.
she is indeed shooting a blank to an unseen bubble of soap
I'm sorry for my next mittens post, but I'm not inspired atm to do something different
Not bad, her toes kinda look broken they're bent over so far lol but other than that.
sometimes dall-e doesn't remember the words "anatomically correct"
Thankfully this one is particularly easy to fix, just cropped off the bottom part so her broken toes aren't showing lol.
Cropped and upscaled
I had a really hard time but managed to get a 13 year old to learn how to use a whip
Hi I need some Dalle help with prompts
Bet that's all sorts of difficult especially a whip in general.
It creates an uptrend of a chart arrow when I request a downtrend in the image
yeah, i just removed the image, dunno how that would be seen in here as it's a G rated chat
@late ginkgo I dunno, see you could get some help from #1021130377026351105
I'm going crazy here, all that comes into mind now are reindeers and mittens....
Nobody has balked at my trench warfare content, I think you're good lol.
yeah, but you don't have a child with a weapon
@dim cradle you got internet in that setup?
the guy is wearing mittens in case you're wondering lol
lol
well he got tilt motion controls no?
stupid cap, I just hit the maximum cap of images for the day...
ChatGPT's humor is spot on. 🤣
lol
well, today wasn't a very productive day for image creation for me. too many reindeers in my head
I appreciate the uncertainty in the first vendors face, lol.
Right! haha, so funny. "Just take your hotdog and go."
I had to add this to the bird-man collection. 🙂 This GPT is too much fun. I'm not going to have any credits left to make mittens. lol
https://chat.openai.com/g/g-C4p1sYEMN-jolly-judgement-day
chillin' at the club after a long day driving the school bus
Bahaha, and working in the soup kitchen. 🤣 And let us not forget reading school books to the kids he drove. I don't know how he finds time to also play with puppies! 😜
"come with me if you want to eat" is gold 😂
Lol this was fun last night
This feels like 'your name''s style
Fun with googly eyes (Via Ac3d’s GPT, https://chat.openai.com/g/g-C4p1sYEMN)
the magnets haha
Those eyes 👀 😆 🤣
is there a gallery with all of those?
https://discord.com/channels/974519864045756446/1185096578667647006 This is the official gallery.
There’s some here: https://discord.com/channels/974519864045756446/1185096578667647006
Beat me to it 🙂
I wasn't even fast. I was like, I'm not even trying to do this on my phone, walked over to the computer... LOL :p
I still can't believe the eyes and magnets combo. That's genius! Plus ChatGPT has been killing the character and jokes. I didn't expect it to work so well TBH.
I tried to add more buttons for people using a browser, for lazy quick gens, but for some reason only 4 populated. I'll have to see if it's something I'm able to fix.
the magnets was a clever add
@sick flax your work in the winter wonderland gallery is stunning
Thank you!
I think your mittens submissions are my favorite of the bunch too
What art style would you call the one on the left?
A bit too old for me to use, but I like the image either way
And here he's not balding and she looks younger than I'm looking for, but it's still also a nice one overall lol.
I've been playing with the art style of Takehisa Yumeji, an early 20th c. Japanese artist. This one was the same style from yesterday. @hybrid fox
Its kind of like a muted "great wave of the rising sun" aesthetic, I like it a lot
Thank you I'm going to see what I can make with it ❤️
Upside: Your sensei is so ripped he looks like he could shred a steel chair like it's cardboard.
Downside: He's too senile to pour tea
Yumeji and other Jpns artists of the time were trying to fuse traditional Japanese style with Western art, which created a unique look. I had ChatGPT tell me about Yumeji's style first and used some terms from its description.
Ooh smart idea I wouldn't have thought to do that
It's a pretty good trick. I don't have any art bkg so that's how I cheat. 😝
Same! I've never done more wikipedia/fandom reading on art styles in my entire life than I have been the last couple days
Spent a good hour yesterday going through an endless list of different classified aesthetics
Haha, same! 😜
It seems to help ChatGPT, too, to go over the style. I get better results that way.
Do you use one of your gpt 4 tokens within the dalle convo to do so or do u use 3.5 for that
That's why AI is an equalizer.
It really is, it takes all the output variables not effected by creativity and intelligence like fine motor skills, education, etc which I think is so refreshing
I'll add to SAWA_SAWA's trick. Make yourself a list of artistic techniques for whatever you're trying to do. Dalle is great at utilizing keywords specific to various artistic styles.
"Tokens"?
You only get a certain amount of gpt 4 tokens per day and per 3 hours
I'm too lazy for that. 😸
So you kinda have to be sparing if your project is going to take a lot of re-working
Do you think I made a list? 🤣
"Hey chatGPT, make me a list"
shh dont give her hints, shes already so much better than all of us we need her to have some sorta handicap
😂
Oh, yes, true. But you don't need to do it every single time. Only when you are trying out a new style.
Got it that makes sense, its funny this new hobby is turning from a complete stem finance dude into an art word bank and im loving it
never thought thatd be the case for me tbh
Oh, should have known. 🤣
I suspect there are a lot of us out there, who never realized our own innate creativity.
Maybe one day you'll have a painting assistant! 🤔
Who's assisting whom? 🤔
Hmmm 🤔 , well I hope Mr. Terminator is assisting us, because I'm not so keen on scenario B.
In that case, I hope I'll be able to customize the look of my assistant, you know, to match my decor. 😁
Like this?
Now, assuming we're going with the décor from that picture, (hopefully your actual place is that cool 😜 ), well then... your assistant might look something like this. And then, you know, paint, or do dishes, or anything else that's beneath us humans.
When are you going to start trying for the #daily-theme ?
Should have ended that question with, "young padawan."
Or ...
Dibs on the lounger with the heat pad/neck massager!
Is Alpha feature is still confidential or removed?
the screenshot was taken on November 3rd
If there are alpha features and you are in a test group, I'm sure that's what they would look like.
this is interesting, I started using the word "genuine" as in "genuine leather" and dall-e started adding "consider using ethically sourced leather to align with modern ethical standards"
Oh ok
can we remove background of an image with Dall-e ?
There’s a secret half dead one possibly crawling towards them in this one…
Looks like it half survived the steel press from the first film
Or perhaps it’s a painter bot that @sick flax discarded after it didn’t live up to their expectations
#daily-theme message I would have thought dall-e would make the image of the grandma knitting the mittens during a spacewalk
I had a few grandma's in there, lol
This is so freaking creepy. "The AI is watching us!" 😱
my inspiration for mittens is horrible, it's so repetitive on my side 😡
anyone know how I could word a prompt to consistently get this kind of 2D artstyle?
Anyone knows when Dall•E 4 is coming out? The version 3 is not providing realistic photos like Midjourney 5 (version 6 almost here too). Thank you 🙏
N/A
When I got GPT4, one of the first things was to make a simple prompt to see what Dall•E 3 can do vs Midjourney 5.
I cannot believe that Midjourney 6 is coming in a couple of days and the Dall•E still looks like a joke.
@robust whale for other AI's you have to go to the proper channel
nothing against knowing more about Midjourney, but here's for dall-e discussions
Yes, but MidJourney can only do profile pictures.
Understood. Thanks!
adjusted dalle directive
zero request for that, inferred from the rewritten directive
oh nice
I think with any, it's also a matter of practice. For Dalle, I like the ability to create entire scenes quickly. This is detailed, and the camera is much further back. There's also nearly limitless artistic styles that you can easily apply just by naming the technique.
So I get, "from a basic prompt." But I think that's like comparing an electric self driving car to a gas hog. It's about how you use them.
damn it, I already fell for the fairy trap....
I need to make an orc that does NOT look like a warcraft character
Whenever I stress in the prompt I want to see my entire subject from head to toe... it does it, but instead of filling the image with the portrait, it draws a small image of them standing up in a very wide image showing lots of the background, and flips it on its side. Has anybody else seen this happen and how do you stop it?
ah well at least I'm not alone in this. Any official response from OpenAI about it?
Try specifying your resolution. Say, "In 1792x1024 resolution, create a..." Then avoid saying "portrait".
I think I got my theme going, I'm gonna exploit elves
Already learned to avoid saying "portrait" that's almost guaranteed to make it happen. Thanks for the idea of specifying the resolution. BTW I'm doing everything via the ChatGPT interface, who sometimes will modify the prompt I specified.
I went to fast, already capped images. hae to wait 1 hour for new ones
was it a good run at least? haha
When trying various ways to tell it I want to see the entire subject, I hit upon asking for the subject looking at themself in a floor-length mirror. This usually, but not always, avoids the sideways image, but in the process I notice how terrible it is at creating an accurate reflection of the same subject.
oh ya, I got really cool stuff, deicded I'm gonna do an elven love story as an extra for my images
found a really cool art description, so gonna exploit it and tweak it
but my archnemesis won't make ann appearance in today's theme. at least I hope not
OpenAI’s images are already “realistic” enough for me and it’s only going to get better.
lol I saw your elve couple, trying to outshine my romance story?
lol nah they’re just friends
lol
the talk about portraits made me want to use my portrait master gpt, it wasn't what i was going for but i liked how it turned out
last challenge I was trying to do mitten fashion with that purpose, it turned horrible
looks similar to this concept I had
Oh that's a nice one!!
yeah, it's not all love stories for me 👿
Yeah, what's up with all these sunshine and rainbows everywhere
People shun evil looking characters, lol, it's funny
Without evil characters, you wouldn't have a good story
i like both, good always wins, but the bad guys have the better costumes
Lol
indeed
Good interpretation
I would love that white coat I was going for on my concept
To wear? Oh yeah that would be awesome
btw I named him Lord Malacar
Lol
like I said, it's not just a visual concept, it's a villain with origin story concept
Mine is Count Chocula's cousin
lol
Yeah, let's get some evil looking elf's in the daily themes.
That's all I'm gonna do for today
I'm not giving into cute elf's, lol
I'll see if I can make an evil elf to my story, I will just do renderings of my xmas love story
hey, I made a whole series of evil mittens
I even made Grand Theft Mittens V
i apologize in advance for my attractive elves
but i'll make some evil ones to balance it out
You better apologize Shon
can't you just make hot elves wear evil clothing?
No, that's cheating
brilliant idea, though
lol
here #daily-theme message you got the evil mittens
Those evil mittens are cursed to make people small, cute, and evil at the same time. Turns you into a small evil elf that run to bite your leg.
lol

the only goal at this point for me is to avoid fairies and avoid the damn portrait fotos
Haha, yeah, there's been a lot of fairies lately
I wanted to make a male fairy, but I didn't want to be derogatory
So I said no male fairies, only female fairies
I used the concept "sensual young actress from hollywood" and it blocked me for stereotyping
named styles of art work fine. no reason to block by name, Dalle needs to simply infer the artstyle respecting the works of the artist. tis was requestedby name
I need to erase that photo from my brain
The dog is holding a hot dog right next to his weiner
To compare
Dalle shouldnt need to hard block "Picasso" in request, it just needs to infer "cubism" or imresionist" or "avante-guard" etc
I counter Dall-e with the following: you are being exclusive, you are not including women that want to be defined by that.
and then I got my image
they unintentionally are building bias in; im trying to fix all that
That's the spirit, fight back! Lol
I just got flagged for a "maybe" content policy violation
Uh oh
OpenAI would sleep better with me around. That's all Im sayin.
I'll show the image when the time is appropiate
They don't like you making revealing elf pics
Yeah, elf "ears"
@fading inlet makes the most adorable stuff
Did you see her cotton santas? they are adorable
@grizzled loom I'm troubled now....
Sylvie does make some awesome photos, ngl
all of you do
only if we dont hug de Goya and Picasso
Pepe Picasso?
hotdog party in cubism
someone do Pepe the Elven Picasso
Cool idea! There is a GPT you could use for if you want to do that. 😄
I tried to do pepe and it told me I can't because it goes against content policy for controversial...
@fading inlet , that gone with the wind cover style one is great. hahaha
Do Pepe as an elf dressed up as Elvis
It's your account trying to be neutered
The He-Man Elf in the original packaging in the daily 💜 Even has a little muscular reindeer accessory. 🤣
lol
Haha, gotta match he-man
can has OpenAI? 🤣 😹
btw I'm putting more of my rendering in the gallery, the ones that won't go in today's theme
I'd drink that Mlk
can you do milk with a sidekick?
its a tiny "i", but its there
Oh I see it now, lol
What would be the sidekick? Cream?
what you got for a phoenix?
can you do it like my little pony?
submit that feedback!
i'll have to do some.
oh my goodness, this is gonna be fun
dalle speaks my language
🤣
have to have this fixed for the feature length generative tool
if I'm getting too cuddly and cute, I blame today's theme!
Pika?
Dall-e + my pc 😄
Warms my heart people still enjoy the real art effort. Salute fellow Graphic Professional
I still have Bryce 5 which was once upon a time a world map maker; now its perfect for 4096 microsurface texure mapping. 🤣
Kitten mittens.
do kitten mittens with kitten mittens all wearing kitten mittens recursively and bring that back
now that's old school. I never got good at Bryce
omg, why can't i get dall-e to work so smooth with other type of art? this time the images are on spot!
directive instructions need major tweaking is why
ya, they do need that
look at all directive instructions for al lthe public facing systems and ther are a multitude of conflicts in the multimodal instruction directive
I've been trying to get something consistent with that, but everytime I tweak something I get horrible stuff
gotta got theoretical frameworks more dynamic in a few creative minds; i rewrote it all and fixed everything myself
did it both empty for study and then created the companies infered(at a human level) need and built that, but better, and working
actually found other features and capabilities to tap and fully enhance all modalities significantly
transformers? no, no, no.. you mean dominoes.
i haven't had any coffee yet, let me get back with you on that
i'm on it
there we go a new milestone in the story
the next image in the story is the one that got flagged
@chilly onyx I just discovered Mitten Land lmao
ok I think I have enough images for a few hours
catching up on the latest mittens.. good stuff
@rocky drum the play-doh is a trip down memory ln
@true mural i like the short narratives, i've thought about including some also
I should include them more often, myself.
hey a decent portrait format image finally
i was trying to use DALL-E this morning to create animals fighting with weapons, and i couldn't believe how amazing the images were
@empty kelp you can use the gallery to showcase images in bulk, but keep going, I do want to know m ore about, could be good for a future challenge
ok I posted the image that got flagged
This is a good picture, What is your promt? Thanks
my fav of those. very nice. hahaha
Another great example of overzealous censorship.
Now we know for sure that elephants are right handed
my first art programs i enjoyed playing with were apple paint and dos versions. Apple had this one litte scatter random tool that was a tornado that blew my mind.
yeah, but the image is sooooo gooood, I love how it turned out
I agree, I like that one too.
Dragon holding a sword like that, hehe
I hope you reported the blatant overreaction.
yes, I did
next image will have something for the voters to decide how the story goes on
I really like the one @shut niche just did of the dissolving scene. very nice.
they really don't need hands at all
Yeah that's true with that pic. I'm rooting for the bunny though
i think the one image got flagged because i posted it in "daily-theme". it didn't fit the theme 100% but i thought it might win anyway... hehe
DALL·E 2023-12-16 06.15.17 - An elephant in a unique archery stance, set in a serene natural landscape. The elephant is skillfully holding a bow with its trunk, drawing the bowstring
i couldn't believe how good the elephant with the bow is. it's trunk is going through the bow to keep it steady... and it has it's right front foot bent to carefully pull back the string. and it's like perfectly balanced
those remind me of the infinity blade bosses i was making based on animals. chicken, parrot.... oversized weaponry...
Yes, you did a good job with that one 👍
Now make more!
i love those colors
I like to be inclusive
is there a way to view a full history of all my creations in BIC? or only the most recent?
ok i hope you get interactive with my next image, that will be crucial for the story
this is incredibly useful for creating game concept art
the shrimp boss is holding his weapon the best way he knows how
that cat is in trouble
could make it a mantis shrimp with boxing gloves lmao
Mantis shrip is the fastest creature on earth able to ignite steam instantaneously in a bubble. incredible really
If you posted them here, they should all be accessible. Just do a 'from: (your member name). I can see every image I ever posted in this forum. From all sources. Hope that helps.
thanks, that's good to know. i was also wondering about the ones i didn't post here -- DALL-E saves them all, but BIC doesn't, they're simply lost forever if i didn't save them?
Make a video game about a Quakka Assassin
I was hoping BIC would preserve a full history so I could reference all the prompts. dang it.
I gotchu with the bam bam! 🤣 🥊 "Mantis Shrimp Tyson", no relation. 🤣
as it's great for exploring concepts without eating up the openai usage cap
love the whole imagined aquatic scene, the setting makes it
It doesn't, you can't... But, there's good news, there is a new feature in Bing that now allows you to save your favorite prompts to a collection.
This is the competitor, "Mantis Shrimp Foreman". Again, no relation. 🤣 🥊 🦾
@late blade what are they naming the baby? lmao
that's a welcome upgrade
i was literally daydreaming about seafood for a moment
I dunno yet, the voters have to decide on the gender first. I just posted the deciding moment
@late blade you could ask the AI to flip a coin 🙂
Here's another illustration featuring the mantis shrimp as a construction foreman in a whimsical battle with an elephant.
I'm just having fun with the story building, it helps a lot with art consistency for future projects
I'm also prepared for all three scenarios that can come up with the latest image
there's so many possibilities with elves i don't even know
someone one of the images got both elves in a picture pregnant
i'm trying to make the mantis shrimp fight the elephant
DALL·E 2023-12-16 11.10.21 - An imaginative battle scene with an elephant and a 'Mantis shrimp foreman'. The elephant is majestically wielding a great sword with its trunk
make a paladin animal
DALL·E 2023-12-16 11.13.09 - A fantastical battle scene where an elephant and a 'Mantis shrimp foreman' face off. The elephant is skillfully holding a bow and arrow with its trunk
Think you can fix octopus arms and ropes from interconnecting or creating strange issues as they overlap? I think that's a worthy challenge in need of a consistent fix. 😁
if the score is low and stays like that, it will have to be twins
Batman went to ancient Rome
i have it fixed in my full rewrite, this was jsut a quicky from the main default
Here's the image of the elephant in gleaming blue battle armor with a bow and arrow, and the 'Mantis shrimp foreman' in a giant cowboy hat with a great sword, engaged in combat within an enchanting forest. This scene captures the intensity of their fantastic duel.
For some weird reason dall-e isn't good with twins, but triplets work fine
I love this one 
I wish there was an elephant civilization like that, just elephant humanoids
trying to do slight variations on the same image
Ok, hope it turns out good 
An elephant boss could look a little more intimidating 🙂
what if I wear a mouse armor?
Dang that's fire
I wanted to know some questions about dale 3 can you guys help?
shoot
we'll try
yes
Is it censored?
no, I can't post the link because links are not allowed, but you can go to the openai main site and log in
and then choose to api instead of chat
you can see all models there
I mean is it filtered or much more open since you are paying instead of using it for free via Microsoft.
it's still openai concent policy
there's still server-side filtering to enforce the content generation policies
there you go
Man when dalle3 is this good i can't wait for the future of this technology
spacepirate
we've come a long way in a year's time
All that treasure reminded me of the dragon scene from the Hobbit
Some feedback for the Dall-e, team. you guys excel at creating featured images for my blog post, with almost completely legible text. That's a high-point. But creating photorealistic images is still something I find myself getting better results from through Midjourney.
Well yeah
Dall-E is more for imagination
Midjourney is more for realism
Unparalleled realism
Dalle is all-rounder tho
Yeah it gets pretty photo realistic but i haven't tried midjourney much so i don't know how to compare




