#images-discussions
1 messages · Page 46 of 1
I walk the line of making the GPT stay just under it's hallucations. Because when it hallucinates, everything becomes the same blend of predictable looks.
Like cloudy blurs of high contrast and incoherently high details
That's no good
We don't want that
I'm trying to perfect a bas-relief sculpture of a DNA helix depicting my cultural heritage. I listed the results of my DNA sampling so it knows which elements to incorporate. But it's still not what I'm envisioning. I think I'm on the right path, though.
England & Northwestern Europe, Wales, Sweden & Denmark, Scotland, Ireland, Norway, and Cameroon, Congo & Western Bantu Peoples--along with percentages.
Those are cool
Make character designs

just wait until we can play them 🙂
made a prompt for a bonsai tree. i think it might be a good candidate for algorithmic art
that's an impressive wizard. the setting reminds me of the houses you can buy and decorate in ESO
Oh yeah? It reminds me more of oblivion 
cool, i can make it a neon sculpture
Bonsai
Four leaf
Looks like a bust of Darth Sidious as a bonsai tree
Anakin keeps it on his nightstand
It's titled "Whispers of Wisdom" -- it's an ancient, wise bonsai soul 🙂
It does have a wise quality to it 🔥
I wanna do a Guardian of Forever
The ancient and omniscient gateway in ST:TOS
Maybe connect it with the daily theme, it fits.
You should. That sounds pretty accurate to the daily theme.
kinda cyberpunk
Got them cyberpunk colors
It's a straight pink warrior who's not insecure about the color pink.
all the Greek gods
All just chillin on the stairs
yeah, check out zeus at the top
the personifications of the human condition opens up a world up art
I hear that a lot nowadays
The human condition
Not sure if I'm a big fan of that wording. Lol
Kind of an avant-garde philosophy
Kind of sounds like neo-marxism
it doesn't have to be philosophical, though
Yeah but still 🤣
You know when you see some words and certain phrasing of those words, especially coming from AI, it kind of makes you worry
It's like that
You're thinking "Why is the AI thinking like this?"
Lol
it's well-read and knows we're resilient
It's quite assertive, especially when pushing incomplete modes of thinking
Which worries me
do you have any custom instructions?
i specifically designed a custom for time management in the role of Alexander the Great, and instructed it to be as assertive and aggressive as possible to motivate me to procrastinate less (with surprisingly good results lol)
I find it that when I give it character roles, it doesn't quite function as well as profession roles.
Like for Alexander the Great, I would say Motivational Speaker for the profession
I’ve experimented with scripts, interactive narratives, etc. and I was pleasantly surprised. But those were usually well defined personalities.
It’s not easy to get Thomas Jefferson to speak like he naturally would. Either he sounds like an old timer or historical correspondences, but they didn’t speak like that.
I try not to over-specify and let it extrapolate and infer, it’s usable
Yeah exactly, it's not easy to get them to sound like characters. Hense why you should cherry pick the meta-knowledge aspects of those character traits.
Like instead of using Thomas Jefferson, ask it to give the meta-knowledge traits for what makes Thomas Jefferson be a good thinker. And it would give you a list, and you condense that list into a profession.
i guess with the customs it's less of an issue, but the custom instructions provide a lot of steering, as you know--sometimes that's helpful, but it can pollute most tasks.
I use them only to make other instructions.
Like Custom GPT tasks
I'm always trying to make a network of prompts
currently mine are designed for research, with sections for synthesis, inferences, analysis, summarization, theories, hypotheses, etc., a fixed context, and instructions for maximum data transfer, insight frequency, increasing complexity, information density, etc., which is a great way to learn about a wide array of topics, but i have to turn that off for most other tasks--custom gpts is the answer, but i'd like a feature to manage multiple sets of custom instructions
when they lift the usage caps, i'll be a lot busier making custom gpts
i'm focusing on the art rn
that'll be a component in my analytical framework, the life's dataset is taking a lot of prep work
I'm in the same boat, but I've learned not to list too many things, because it bogs down the AI and instead tune it's thinking pattern to generate a list of appropriate answers. Because from my experience, if you flood it with specific texts, it will interpret them in multiple ways, which causes it deviate from the original idea.
Like I don't give it documents upon documents of info. I've taken all that stuff out, because it looks like it can't process them properly if it's trying to read everything all at once.
@shut niche i saw yours, i can't find the Japanese tea ceremony I was going to post, but I also made this one
A lot of the prompt writing I do is actually honestly rooted in the #1187233013956874260
Using punctuation marks to encapsulate ideas in parentheses and brackets to tune it's thinking patterns has been far more effective for me.
wanna see something wild, use the api, crank the temperature setting up to 1.75, instruct it to speak like it has an IQ of 200, and then grab the popcorn.
Lol, I'm good at 0 temperature
I want that precision for steering facts
i tried listing attributes in square angle brackets, but i found a simple use case that breaks it:
Digital painting: a star [vivid sphere] positioned at the center of the composition. Below the star, a sundial [purple, titanium].
But this does not work, the star will come out looking purple. I think I find simplified natural language works best for my cases.
Change the square brackets to parentheses
analytical is great for logical tasks. the setting has potential for greater creativity, though, introducing a little more statistical improbability.
Do I have the syntax wrong?
Use square brackets more when they're correlating with parentheses. Like:
Digital painting: A vidid sphere of a star (positioned at the center): [Below the star is a purple titanium sundial]
Think of square brackets as a markdown format for parentheses
As written though, it depicts a star inside of a purple sphere, so I'm over-specifying, and needed to steer it back to yellow here.
Digital painting: A star, appearing as a vidid yellow sphere, positioned at the center of the composition. Below the star, a purple titanium sundial.
I'll try correcting the typo, but I'm still anticipating issues
that's better, but it's not exactly accurate or precise. these are the fundamental building blocks of the composition i'm trying to learn more about.
if i'm looking to construct from scratch and have more control
I think a better prompt would be:
Sundial (Digital Painting [real scenario]): Above is a the Sun casting a shadow down onto it.
an efficient abbreviated syntax can be useful, if it's consistently interpreted and rendered. since it condenses anyway, sometimes the language modifiers provide more clarity
it would require a lot of testing with some use cases
lighting and shadows look good
It's got a clock on it too, in case you ever need to know the time when you're watching the sun dial 
This one is honestly not bad. It's at the center and there is a sundial below it.
It would be the start. The way it’s composed would provide more weights and adjectives. Assuming those two objects are the focal points. Then I can cover the scenery.
I hope you will experiment with these punctuation marks. It really works when you're trying to condense info.
Especially good for 480 character limit of bing
AI prompts like to conclude with an inspirational message. I have mixed results with that.
Ah yes. That’s a good workaround that current limitation
And it's not me saying these things, I've learned from chainbrain, quicksilver, eric elliot. A lot of really smart AI people who are using pseudo language to talk to AI more effectively.
So don't take it from me. I'm just trying to streamline it even more for beginners.
the principle is sound

i just want to do justice to the Guardian of Forever
You will Shon...You will.
i appreciate the confidence
the moon is cool in that one
i like the purple
@glossy scroll do your character generators always produce full-body shots? sounds like some have been struggling with instructions for that.
yeah they do. most of the time. I don't ask for it.
Today’s theme has resulted in the best art lately
Ooh.
Great tail
I would advise a prompt like
Dragon Rider (Character design [conceptual accuracy])
i wasn't being critical, though
I can't connect with the theme today.
What do you mean?
I tried BIC now
And it's good for like super generic images
But terrible for customization
It's like Midjourney
Thats why I don't use midjourney anymore
it was a sincere compliment lol
so many possibilities, though
almost anything people get together to do a lot
or history
Oh no I know that, haha, I thought you were saying you wanted me to help people by giving a prompt or something.
Midjourney and OpenAI need to work together
That should be a no-brainer team
We would make crazy accurate images
I just have a disconnect. But I'm pretty sure I know why, and I can't even seem to translate that into an image.
maybe start with a high-level executive summary 🙂
just noticed that guy on the right. I'm not sure what's going on there. maybe best to not know
Haha.
playing around with my burning man tradition
Nice.
i think it's an early preview, i think a lot could be done with that one
His name is Balthazaar, and he's the head director of the bazaar
I wish I had an outfit like that, I'd use it to go to the supermarket

A culturally rich festival.
Nothing like grandma's cookin
I don't remember this tradition
You don't remember that tradition for dressing up as an ugly baby?
I remember the teeth suits
it's a time-lapse portrait. i've experimenting with interactive elements now

i don't really see how it's "interactive" but ok
it was your message
he knows
this is how he deduces information with himself. he's like sherlock holmes
why r u tryna make ops
ops?
lol

where did you come from where did you go
cotton eye joe
modern art 💀
it's the least horrifying of the bunch. wow
unsettling
We should play the prompt game I played with Dys Topia yesterday.
I had to switch windows to bring the spoiler warning back up
they're not gory or anything, but I don't like the way they look
Prompt: In a picturesque natural setting, a traditional ceremony commences as people from all walks of life converge outdoors to honor a shared heritage, marked by rituals, traditional attire, and a sense of unity that transcends time.
I'll add teeth perhaps
the teeth twilight zone
So I decided to try Dall-E for once so I just signed up today, but I got no credits? Any way to get free credits?
dall-e 3 is available from a gpt-4 convo or by using the dall-e 3 gpt in the chatgpt sidebar
You don't have GPT Plus membership?
You can also use Bing Image Creator
It's free with no token needed
Yeah, I don't have GPT-4 Plus because I don't have money tbh. Maybe in the future when I get a job.
Face blur privacy was a turn off as I just want to create some anime or cartoony version on some my image of myself.
so many images always look cgi though my big complaint for dalle
bing somehow they look better, but theres always that one womans face there too
MJ 6.0 was a big waste of money for me, but i give them credit: big variety in people
You have to scrounge up some money then to sign up for GPT Plus memebership. It's more than just using Dall-E, it's more like making your own customized Dall-E with Custom GPT's. Look under the couch to see if there is some loose change there, or break open the piggy bank. Or you can ask your neighbor for a cup of sugar and when he's not looking, you can take his wallet and use his credit card.
it does seem better at digital art than photorealism, but i assume they're working on that?
makes me think of reincarnation
I'd be surprised if they were working on that. but who knows
it should render more seemlessly, though, ideally
maybe. i think it could do it fine if they wanted, like i said, on bing that dalle does photoreal a lot better imo, they really neuter the gpt version imo. like i say, MJ is really good at diversity its not just one fish mouth woman in every image
anoyne here use Meta ai image gen? i have not but i thought i read it is out
one issue, and no one seems to notice this, is gpt4 passes off photorealistic to dalle3, and then dalle3 interprets it as the art style
it's not the only lost in translation sort of issue you'll run into either
Since I'm in the Philippines..... That's 1,200 peso per month if it's $20 per month.
And if I consider that my only income is from my dad which only gave me 300 pesos per WEEK, a lunch meal is 80-150 pesos here.
i usually start with "cinematic image of"
And avoid terms that steers away.
Mention surreal somewhere and it gets all cosmic
It doesn’t translate that always into subtle nuances of a realistic scene
I'm actually familiar with Philippine wages and the lowered economy there, because I have a couple of Filipino virtual assistants helping me with my online business. It's something I suggest you do if you want to get wages that can be comparable to North Americans minimum wages. To become a virtual assistant. I can help you more in the DMs if you're interested because talking about such things goes off topic here.
seems like there are standard things certain words do too. like you ever notice how it loves to put those constellations with connecting stars in so many images, maybe from the word "cosmic" idk
well anyway, those are the tooth traditions that I came up with. very touching scenes
the tooth ceremonial clothing is hilarious haha
they are all very healthy teeth too. you can tell they take pride in their teeth
dont remind me, i have a molar in not so good shape haha
cosmic will do it. a generic nebulous.
a tooth with teeth
teeth gotta eat too
Looks like he's saying "Eyyyyeah!"
well now i can blame my rotten molar on the tooth and not me. it was eating all that sugar i guess behind my back haha
1 REM cycle is good right?...
Woke up, saw you guys working hard, and got inspired. 😆
I feel like I can hear this one
GPT League FTW lol
@dense mesa
GPT League is growing wow
Ugh. Some dude on a dating app just lectured me about AI and its potential consequences, which all equal doomsday.
lol
@vapid elk I got a new one for ya
you think its bad now, wait until like 2025 when ai is more integrated youll have people losing they minds and getting all the conspiracy stuff
Oh yeah
yes only it always overpopulate esp in widescreen with some elements. like spear/spikes galore everywhere you look
You try it
im not saying i can do better its a problem witht he model not the person
are you using a gpt @glossy scroll ?
I asked gpt4 how dalle3 was making those wide aspect ratio images at that size when it was trained on 1024x1024 pixels and got some nonsense non answers
Yep. Using my gpt: https://chat.openai.com/g/g-PpgBuWQm0-imaginemancer-2-5d-creator
claims it's neither outpainting or upscaling. but I question it's response
it cant give up the secret sauce haha
Do you have to do anything special? Or do you just prompt as normal? If so, what does the GPT enhance?
People can look up my commands by asking the gpt at any time. I don't care about my secret sauce.
On my side, I got a whole list of custom instructions I put, but on your side, you just type whatever you want in any simple way you want and it will create it for you. Like, for that knight, I put "Create a fierce looking death knight", that's it.
talking about what pictures said not you
Oh my bad 
😅
@open trench It has some randomoness, so you won't get the same result.
oh. okay.
But I can send you the exact prompt for it if you want.
I'd like to see that. I am all about learning. thank you.
Magnum opus: Masterpiece, high amplitude photo-synchronization, best quality, highly professional, unwavering depth in alignment, dramatic lighting and excellent perspective that adds allure, immaculate details with high contrast and precisely impactful hand drawing for sharp/crisp and deep focus, epitome of awesome, no less than badass, 16 by 9 aspect ratio: Create a resonantly balanced !!!2.5D!!! illustration in a "GothicWarForge" style: a fierce-looking death knight in intricately detailed armor, inspired by Gothic cathedrals with war-damaged aesthetics and a metallic sheen. The knight wields a large, imposing sword, blending Gothic elements with a forged, battle-ready look. The scene is set in a war-torn landscape, with remnants of Gothic structures and a dusky atmosphere. Strategic lighting highlights the knight's formidable presence and the intricate details of the armor and sword. The composition adheres to the "GothicWarForge" style, integrating cultural, historical, and philosophical elements from Gothic and war-torn periods. The image exhibits a polished, fully rendered appearance with smooth textures and clear continuity.
DAMN THat is one epic prompt
quite the prompt. So your GPT took those few words and tacked all that onto them?
Can I put any prompt in or does it have to be a knight?
lets see him in a sunny field haha
Yeah, but my custom instructions for my gpt is even bigger for what it should do with that prompt.
You can put any prompt you want. It will create a random new style based on that prompt.
Ty 
- Modify the GothicWarForge style death knight to include prominent teeth details.
- Apply an analog photography aesthetic to the image composition.
- Maintain the original elements of detailed armor, a large sword, and a war-torn landscape.
lol
i liek this one has some good photoreal vibe
- Add elements of pickles and potted meats into the GothicWarForge style death knight scene.
- Maintain the original Gothic and battle-inspired themes, with the knight, armor, and sword.
- Ensure the additions blend naturally into the war-torn landscape.
I mean, it did what I asked it to do
tell it to stop doing that. tell it to stop playing games
Trying to create an autostereogram -- AI wrote the prompt, it actually included this phrase lol
Create a mesmerizing autostereogram that transports viewers into an enchanting underwater world. Craft an intricate scene with coral reefs, vibrant marine life, and playful dolphins. **The hidden 3D magic** should make the dolphins gracefully glide through the depths, enticing viewers to dive deeper into the immersive oceanic tableau as they unlock the secrets of this aquatic realm with their eyes.
That's weird...I never put anything gender specific 
We just need to tap into this "hidden 3D magic"
same prompt in dalle3 results in this:
that's almost too good
Mine is specific to 2D and 2.5D
Doesn't really go outside that zone much.
I didn't specify anything about the style f the image.
try it on the dalle on BIng if you can nezho
The Custom Instructions within my GPT force any prompt to stick to 2D and 2.5D
so, I think i didn't break it this time
as always bing says its unsafe
An anthropomorphic female black leopard. She has long white hair that is thick an lustrous and white eyes, dressed in an all-black leather costume with a yellow lightning bolt symbol on the chest and a long black cape. One clawed paw/hand is raised above her head, digits spread wide. Her other hand by her side forms a fist. Both glow with purple and white lightning. The scene is set against a stormy sky with strong winds lifting debris.
it's always a word or two. well normally
Please proceed without me 
you can try being emotionally manipulative and it may make the images
maybe "chest"
or just ask it straight up what the problem is
well I guess you're using the image creator directly?
it was female black leopard.
I always went through the chat so I would have it send the stuff it wasn't supposed to say
i changed it to female leopard with black fur and it worked
was that on your gpt?
Yep
I've gotten all males. lol
holy crud
And ask it to try again if it doesn't work the first time.
it lacks some balance, though
okay. so I'm in heaven. The character this is based off of is my favorite ever.
I'm not saying names or anything.
You can say Storm
The policies for copyright is only for what the gpt can't generate
That's it
Nothing to do with discord
Good Morning! 🥱 ☕
GPT League 🏆

I think I am ready to start bulding my custom cat-girl gpt
Do it..do it now!
whoa
so, can somone point me to the beginners guide on how to buily a custom gpt?
thanks
np! 
Haha I thought you said bully a gpt
I was going tto say, if you really want to bully a GPT, you probably could. but it'd be unnecessary
bing?
ok, seems dancing can give lots of potential bigs and weird artifacts
ok, seems dall-e woke up moody today, gonna have to be more creative with prompts
create a bird
I've reached my GPT limit. I have to wait for the cooldown. Its another hour from now.
the earliest ferrotype photography of authentic extraterrestrial beings and ufos
give you a prompt
you can try it
I will once I can, I promise 
i dunno, that effect reminded me too much of something else
logo
close, but not
it looks like it made a 3D image for red and blue 3D glasses
yeah, it didn't get it, i clarified, it still struggles
gpt is teaching us that it wants us to use mesmerizing on all prompts. normal every-day is not allowed
i wonder if it’s actually 3D
i don't have any 3D glasses
i do, but i cant seem to find them. was just searching in a closet
dang. perfect pose and everything, but the wrong aspect ratio. The wide format ones all go off the rails.
can you create a 3D anaglyph photo of Santa and two athletic and diverse female elves
i think it actually works
that's amazing
if it works
You should make a GPT for that
That would be awesome
Amazing idea
even if the AI didn’t grok the 3D part it’s still a very amusing image
cool eyes
try making a beholder
All using my gpt: https://chat.openai.com/g/g-PpgBuWQm0-imaginemancer-2-5d-creator
AND I just got stood up from a dating app. Oh well, time to AI.
Made a photo version of this. Pretty decent details in the fabric, and skewed shot.
Looks awesome! ❤️🔥
I really want things that focus on photographic images. Are you using a gpt or adjusting the prompt each time?
Just using a GPT. I'm asking for certain details. I asked for detailed fabric, glowing purple eyes, in an Egyptian temple with a magical staff. I asked for side lighting.
can you create a photographic 3D anaglyph of a scary beholder monster. the beholder is on the beach in Hawaii and has a large tropical fruit drink with a decorative umbrella and straws
Trying to catch up a bit and thought you meant you had created some virtual GPT assistants that believe they are based in the Philippines there for a moment
Dalle is getting worse and worse with each day. More limits, less pictures less everything except the quality
Hahaha, that would be strange. But no 
Real people.
75% what I do is fight against the policy which is not really against the policy
cmon guys please fix this
what does this even mean? I barely created anything
It means you have to wait a little. We all get errors every other day or so. There's bugs all over the place. It's best to just have patience.
@naive lake Have some hovkeet and take a break.
🤦🏻♂️
Elder Thor
If this is like this now, I can't wait to see the rumored GPT4.5 version of this gpt 🤯
you can do 200 images per day with the cap also set within 3 hours. you have to keep that in mind. it's what you agreed when you joined
I’m not even close to that number
are you aware that failed attempts and server side problems also count to this cap?

Demon Lord
I wish there was a bot that explained this...
There's a 200 image msg cap, but there are also 15/60min thresholds that you can hit from rapid fire generating. Failed images count towards this. I don't know what the exact 15/60min threshold is, but you can probably guess, if you're firing away, it's going to put you on a "time out."
Supposedly this was going to go away when they got new hardware installed. I don't know if that's still the story.
There's also an ability to send feedback from within the app.
- heavy metal starts playing *
Let's see some complex emotion in one of these characters. Capture "the demon condition." LOL
Lol...Oh god. Heck no 
Too emo'tional
Fine, I did it. 😛
Here's your demon, alone, pondering the meaning of life, after he aged 10000 years, his wings fell off, and someone stole his cool staffs.
love that idea
Now that's what I'm talking about Ace! That's what I'm vibin with! 🤘 
I love that 2D'esque look
It's all about "the artistic value" guys, trust me. "The demon condition," the composition of life... Be one with the AI art generators. Embody the daily tribulations of your average demon, then capture it into an image.

Gm
Hello 
yo
I'm seriously gonna listen to some heavy metal right now. You inspired me, lol
it didn’t seem like the beholder image was correct, so i asked ChatGPT for some clarification
Here's your jam session.
With an acoustic guitar? 
I took you for the "easy listenin" type.
an ukulele would be better
add to the prompt the guitar resembles an Ibanez ALT20 with glowing crimson patterns.
Definitively not, lol
Do you play? That's not a non-guitar player idea...
I play
Nice, same. Hence "my ears" perking up when you asked that. I have an Ibanez acoustic sitting 10ft away. But I play my strat the most.
lol
@shut niche Listen to a song called Ultraviolence by Sxmpra. That's the type of music I listen to these days.
Just to give you an idea of my music type.
Ok, I made it heavy metal. He's got some great Gibson gear now. He's even playing a Les Paul. 😛
:peepoLaugh:

Did anyone even notice that she has a belly tattoo that says, "heavy" ? 🤯
I read REAVZ
🤔
REAVZi
Looks a lot like something from warcraft
To effectively communicate the “easily listening” in the image you should have the musician and audience in a more calm and relaxed environment
Please do xD
can i add you as a friend?
Yeah sure 
Yeah it does kinda does look like warcraft art
If you have any suggestions, let me know 
Yea
lol, ok, one moment
I would send one i made a while ago, but im on mobile rn
Awww I got capped out. 20 mins
Dont worry xD
shame
Cool
Dalle is not really good at making kobolds
It always make something that looks more like a gremlin
This was on pourpose, it is a bit hard to make kobolds xD
This is closer
Kobolds are small lizard creatires
I just googled it
The joke is that 3 of them would wear a coat to pretend they are one medium sized creature xD
@vapid elk
@vapid elk How's that, more authentic to their characters?
That's a good one 
@vapid elk Here's the prompt, in case you want it.
Create an HD photo of a whimsical and humorous scene where three traditional kobolds are stacked inside the same trench coat, in an enchanted forest setting. The kobolds, small and reptilian in appearance with scaly skin, pointed ears, and snout-like faces, are comically disguised as one full-sized creature. The top kobold peeks out from the coat's collar, the middle from a mid-section opening, and the bottom from the coat's hem. They display mischievous and playful expressions, clearly enjoying their teamwork in this amusing ruse. The trench coat is oversized, adding to the humor. The background is an enchanted forest, with magical trees, glowing flora, and a mystical atmosphere. The forest's ethereal lighting casts a whimsical glow on the scene, emphasizing the fantasy setting and the playful nature of the kobolds' disguise.
three kobolds on a boat is very good also
Those aren't kobolds...those are ogres
Those are surely not kobolds lol
well, ChatGPT 4 and DALL-E 3 think they’re kobolds
You'd need 10 trench-coats to fit one.
i think those are gremlins, not kobolds
What prompt did you just use? I'd like to know if I'm right about these rotated images.
Oh, you can just ask it for it. I say, "what prompt did you just use?"
Asking, because I have a hunch that most of them are caused by the prompt, and aren't really "dalle malfunctioning," but moreso, confusion that's happening from how it's worded.
Spooky
Yeah, the sideways images that people have been reporting. I barely ever get them, and when I do, I usually can tell why, in my prompt.
It is a bug indeed
sideways isn't much a problem as long as it fills the canvas correctly, just got to rotate the image if you save it to your files
It is a behaviour the devs are aware and want to prevent
yeah, that is good, would be better if that works, but atm, it's good enough
same
So you don't have anything like, I want a tall portrait?
tall portrait full body is impossible to do right
selfie portrait format is really good
I did, i asked for a tall image
these kobolds aren’t spooky. they’re just average everyday kobolds living in their natural dungeon environment
The problem is when the ai just makes something in the wide format but sodeways
Instead say you want a 1024x1792 image of... don't say portrait too.
Or "tall"
Ill try
It's like a double negative. 4 rights = a left
I usually use "wide" and "tall" because those are the param mames on the api
But i think that confuses chatgpt amd dalle
If I need a specific res, I usually say, "In 1792x1024 resolution..." or "In 1024x1792 resolution..."
Right, I don't think it's a glitch. I think it's a misunderstanding.
Maybe
I think it is a matter of embeddong some words
It only happens when the format of the image ends on the prompt
wen inpainting
Well, go back and forth. Just avoid "double referencing" if you try my way. E.g., don't add additional wording like tall format, portrait, rotated 90⁰, or whatever. Just open with, "In this res."
ChatGPT also "assumes" sometimes of you don't specify. Like if it thinks the image would look better in 1024x1792, sometimes it switches without me asking. Then I just say, "Can you run that again, in 1792x1024 instead."
My results aren't very accurate. I think these guys did a pretty good job though 
@vapid elk Instead, let me just give you this 
I told it to provide me the JSON that was sent to explain this phenomenon, as I see it. ChatGPT sends the res, specifically, not in the prompt, but in the JSON request itself. So if you add a separate "direction," I think dalle is "sort of" drawing on a sideways canvas. Like saying "go left", and then also in the prompt, "go left." It doubles the instruction and dalle interprets it as "draw sideways."
Maybe I'm not seeing the whole issue, but that's why I was curious about your prompt. Not insisting I'm right, but I don't experience the issue.
"size": "1792x1024",
"prompt": "Create an HD photo of a whimsical and humorous scene where three traditional kobolds are stacked inside the same trench coat, in an enchanted forest setting. The kobolds, small and reptilian in appearance with scaly skin, pointed ears, and snout-like faces, are comically disguised as one full-sized creature. The top kobold peeks out from the coat's collar, the middle from a mid-section opening, and the bottom from the coat's hem. They display mischievous and playful expressions, clearly enjoying their teamwork in this amusing ruse. The trench coat is oversized, adding to the humor. The background is an enchanted forest, with magical trees, glowing flora, and a mystical atmosphere. The forest's ethereal lighting casts a whimsical glow on the scene, emphasizing the fantasy setting and the playful nature of the kobolds' disguise.",
"seed": 2456465764
}```
Or ChatGPT does, one or the other. Because if it's a portrait that's laying over, it was sent in the JSON as: 1792x1024, but then the prompt says "portrait" or "tall."
the orientation is portrait in that prompt
even if you change it in the context of the scene
just change size: to "1024x1792" then it's wide format
Bearfolk
Wow I'm tired. The image I gave him was 1792x1024 silly...
twenty minutes left in my penalty box. grrrr
making some interesting optical illusions, maybe for a gallery
I still got 30 minutes in my penalty box
You made me second guess myself. That's all correct. I didn't make him a 1024x1792. The JSON was from my image, not his. Lol 🤣
share your gallery post when you're done, @dim cradle
It works very well to say, “The image should have a portrait orientation with 1024x1792 resolution.” I just did several tests of this
afaik, AI's are all "it"

my kitty friends aren't getting the love they deserve....
but for some reason dall-e is making them more like toys, which is annoying me
Giraffefolk Mystic
it takes five stereotypical kobolds standing on each other’s shoulders to wear a coat
what is the curators role?
it's a channel for people that have the curators role
and keeps images and the learning towards that image active
Can you please create an image of five equally sized but very diverse female kobolds standing on each other’s shoulders inside of a coat jacket. The image should have a portrait orientation with 1024x1792 resolution.
"Can you please?" -> Image of:
i can’t explain why exactly, but you’ll get better results from the AI if you use please and thank you
i first noticed it when i was asking ChatGPT to write PyMEL scripts for Maya. It gives very good results, and i always say please now
@vital gull brought me back such nostalgia 😂
I used to step on grapes and make fine wine a lot when I was little
Good memories
haha! same!! who doesn't love homemade wine! if it's done correctly ofc lol
Homemade wine is amazing
as a georgian, i agree
You must also enjoy Armenian food too then, I presume?
haha cavatanem, hello neighbor! I haven't actually had authentic Armenian food. p.s. i wish database had more info about caucasus region lol
I've had khajapuri and it's one of my favorites
I know this is a buzz kill, but I lost my father to alcohol induced dementia. I have plenty of those grape stomping memories, he 'loved' his homemade wine...
I also used to really drink, "work hard, play hard" lifestyle.
After dealing with that, I did my research, and now I wouldn't touch alcohol with a mile long pole.
Not preaching, just saying, the science, and that experience, really changed my perspective on even "moderate/in moderation" drinking.
Tsavet tanem. haha, I just got that 
Yes that's what they say a lot in Armenia too. Especially the bros, lol
Ah, so sad to hear that 😦 It's really hard when your culture is so heavily entwined with winemaking and your country is known as the oldest winemaking country. You will see drinks everywhere and at quite affordable prices, so it leads to a lot of complications for those who aren't accustomed to knowing their limits.
Let's keep it rated G everyone. Personal events are yours to keep, they are moments that make you and are not necessary needed to be shared with the world.
True! let's move back to Dalle!!
hah
although I'm impressed so many great traditions
we've discovered today!
Sorry, wasn't trying to be too heavy or too TMI. The comment just hit home.
it's ok @shut niche , we still 💕 you
I know what I'm gonna generate for Aced. Grapestomping 
it was a valuable message i think. our neighbor had a drinking problem, and died from liver failure earlier this year. her husband and the two teenage kids are still living there
and so i totally concur
the firt one looks as pressing the kids foot with that gadget
let's squeeze that foot
It's a pressing hammer, so you don't use your feet. But it looks like the kid is just using his feet to have fun as a joke.

Which is wholesome
fair enough
you are giving me too many character ideas
Maybe one for a Planetar too? 
this is incredbly epic
Lit
im back home, here is the image I generated a few months ago
May I save some of your images? I might use them on my dnd game =P
Yes, please, by all means 
Another Triton
oh, I really wanna go there sometime, looks like such a beautiful country
are there any places you recommend visiting? 
your character art level using dalle is insane.
great generations
definitely top tier dnd content
Bah, i reached that time again where there's so much stuff in chat that the browser stops responding....
I'm glad you enjoy it. Thank you. It's taken me quite a while to learn it.
aaaaaand back to the penalty box
it plays out well tho, my penalty ends 1 minute after the new daily theme is announced
Can you please create an image of several diverse female kobolds performing the hula on a beach in Hawaii. The kobolds are wearing appropriate beach attire and flower leis. There is a storm with lightning and a rainbow. The image should have a landscape orientation with 1792x1024 resolution.
lol
Do you have a defined setting for the light?
like
hmm interesting, my custom gpt disappeared and can't access it with the link
if you specify that there is, “a storm with lightning and a rainbow” you get really superb ambient lighting
yeah no doubt
I saw a few people talking about their GPT’s vanishing without warning in the GPT discussion channel a few days ago…
ya, i'm there atm, trying to see why my cute kitties disappeared
Hello, Every one. Happy Christmas.
heya
I have to generate the reality image but the dalle-3 can't make the vividly image everytime
Who can help me?
@chilly onyx in the mean time, just having a warm drink, can't do anything else
If possible, let me know the prompt formula to convert the normal prompt so that generate the perfect image
@viscid osprey shoot and let's see what we can do
the perfect image = is what you want to see
I could just do with a nice hot cup of tea ☕️
what is the perfect image for you? perfect geometry? perfect zig zag? perfect potatos? perfect colors? perfect oompa loompas?
which brings me to quesiton why nobody has done oompa loompas
Can you tell me the prompt of this image?
image of a cat-girl waking up sleepy, with a cup of coffee saying starting a good morning.
but it has a lot of background work, if you put that, it won't be anything like it
So the images look good, the images are realistic and all the objects vividly resemble real people. However, when I create an image, the output image is not the same or vivid as the picture.
ok, you have to refine the concept then, what do you want in that image that looks good for you, you want the light in a certain way? you want the surface look a certain way?
well your prompt has to say that you want both of them
then your prompt has to say you want a photograpic image, or a realistic image, or a lifelike image
Okay, I see. Thanks
but that's not enough, you keep trying, see what you can change and make it better, it's not the one image that is going to be better, it's what you make out with it at the end
Does anyone else have a problem with dalle3 only creating male characters and when you try to correct her, she says she already made a female character?
It goes through phases and then starts creating errors instead of creating pictures.
nope, I still havn't tried that
Time to lock and load
- Integrate a "catloaf" element into the GothicWarForge scene.
- Maintain the anthropomorphic black leopard character with long white hair and leather costume.
- Include culinary elements such as pickles and potted meats.
- Feature electrifying effects on the leopard's paws.
- Set against a stormy sky with analog photography textures.
Well Dall-e got tired of me fast this time... I'm in a penalty box again
waiting time 2 hours
whaaaat????
I've never had a two hour penalty box stay. You might as well bring a sleeping bag.
That sucks. Do you use Bing in the penalty box?
sometimes, but don't like it, the prompt is too short
so true.
I have found it really useful, but I have to add in details after. But it's definitely good for the try-out-new-things workflow.
Does anyone currently have access to Dall-E 3 via api?
I do not, currently. I've been having this response for about 4 days now
O my God love it.
I just edit the prompt for Bing. I still say Bing is King. it gives more diversity of images, more natural faces, Dalle on GPT has a boring house style imo. like look how many images here have the same elements, like the wizards staff with the big round top of it, its like it just spits out things from a 3d model for a lot of it, especially faces which are usually trash
I have been trying to challenge it to do different angles and perspectives. It's like whispering into a windstorm sometimes, though.
it depends on the context and creation ideas, both can pros an cons
yes some things gpt is better for, but overall i get better diversity from Bing, esp if its photoreal
that is good, it's something that is working for you
they are both 1024 x 1024 (square at least on dalle) but dalle always looks crisper to me
BIC has so far blocked any attempt to do this, GPT has been really good at this
its weird how Bing blocks some elements and then Dalle blocks so many more like anything copyright
content policies are different
MJ 6.0m is trash imo but at least it has massive diversity of people faces
same language model, different censorship
haha yep
never used MJ, maybe someday, but for now I'm just expliring dall-e
youre not missing much
although they ready to ban me over there i think for saying this haha
one cool thing id like to see Dalle do is you can link images in MJ to influence the image.
i'm. just tired, think I am on a creativity burn out from trying to do stuff for all daily themes
and MJ it seems they are always at least engage with people and updating new ideas all the time. Dalle team puts out a new version then they forget it which is weak sauce
yes maybe take a break from it and recharge
480 characters
thank you
you are welcome
I think the Bing limit is what Dalle 2 was, or someone say once
its like they need to update it for 3
i better chat here alone they have the pitchfork out in MJ for me haha
just make sure you don't get another pitchfork here and you'll be safe
haha yes. well people here take little critcism better. some people get mad like you insult them personal just by say "i think it does this poorly etc"
cool image
just a concept, brainstorming
the white one should be brighter and larger than the dark one
i feel like a weakness still for all image gen is the more things going on in the picture the harder it gets. so here you have 3 subjects sort of, the candle holder, then each candle is like a separate person and getting them to have differnt things going on is tough
This assumes it's not an intentional choice. I believe the limit for dall-e 3 is quite high, however, every token being processed makes it more expensive, so I suspect it is intentionally being limited, certainly the quality is.
yes that could be
it's a double edged sword, sometimes it pays off to do very detailed
Okay, can anyone figure out how to convince Dall-e 3 to create 2 candles, of roughly equal size, one being lit from the flame of the other? I've tried for quite a while, and even getting a candle being lit from another, at all, is tricky.
if you are too vague you can also get this kind of attrocities
that's a no go for cat-girl
well i was right, they ban me in MJ haha. anyway, thats a funny cat girl haha
BIC can totally create cat girls, but... yes, focussing on critical details is necessary. (MS themselves are giving away image gens, so, they obviously opt for cost over complexity)
interesting band though. you dont see stand up bass often with some rock and roll these day
dalle3 has a lot of limitations in that regard. tried a few different approaches. gpt4 knows what it's requesting. it told dalle3 in several different ways exactly what it needs to do. but no go
I see that every year at the montreux jazz festival
cool festival
Yeah, it's just asking too much I think.
I wouldn't ask if it were easy, of course.
it might be possible, but not sure how. may just not be able to piece it together from it's training
yes like i say, they all do. in MJ people have the same problem too, 'why cant i get this image with multiple subjects to work?'
I think it is because it is breaking an internal rule for what a candle looks like. (two vertical edges)
interesting theory
close?
very nice
dunno, what I've been able to do with cat-girls says otherwise, for example
but specially with the jazz concert, it has been a miss completly
cool image
that's not an atrocity. that's cat girls. but you're trying to get images of a woman with cat ears on her head. dalle3. can't know that if you don't clarify
btw my cat-girl gpt got removed due to content policy, so can't make my cat-girl character creator
cat-gir in the sense of Neko Musume, not in the sense of I got a cat and dress it like a girl
yes but dalle may not know that so easily
well I could say I want fire in the sense of an alternative definition of fire, but dalle3 can't know that unless I cliarify
that's not a limitation of the model. it's a testament to it's versatility
i love candles, so this will be interesting
Why do cat girls trip up the content policy?
depicting humans with animal traits is a grey area
How bizarre
I’d have thought it depends what you have them doing, not just their existence
I thought so too
it's specific words within the description that might not seem offensive but trip up the content policy anyway. it doesn't have any problem with the concept as a whole
just saying, the better you understand what might specifically be triggering it the better you'll be able to rectify the situation when it happens
Yeah but this was a GPT. What sort of phrasing did you use in the GPT Dys?
yeah, and for most use cases I've been able to work it out, but now with real enviroments around the world I'm struggling, concerts, festivals, monuments of real world locations
GPTs are just pre-prompt instructions
yes, one thing that annoy me was GPT would often rewrite my prompt and add the word "intimate" like "intimate photo in a cafe" or something so i had to tell it dont use the word intimate. extra annoy to me since it was the one to add intimate then it also is the one to say ooh you using some words we cant allow. are you kidding me? haha
Yeah I know, I’d have just expect those instructions were more on setting and style with a small bit of the cat appearance to allow for some flex in the prompt
I"m not justifying the filters by any means. I believe they're overbearing often times and filter out unnecessary things, limit creativity, etc
I don’t know though obv since it wasn’t mine
Wording at the moment is around 6000 word essay on ethics, content policy and derivative works by definition, law and depiction with includes but not limited to real lifelike images, 3D animation, realism art and many more
Sounds like it could have been a lot of things that triggered it then in that magnum opus, right? 😅
what do you guys do with gpt anyway? like give it the big instrucitons for an image, then in the prompt you just write something basic? or are you having to write it all again anyway in the prompt because it some time ignore the gpt
well, if I were you I'd honestly copy/paste your instructions into gpt-4 and ask? it could hallucinate, but it might also lead you in the right direction. maybe it's something obvious you overlooked
that is one way, you also have a set of custom instructions if you want to use that too and in my case I set directives at the start of every chat with dall-e that apply globally to the convo
why are cat girl so popular with some anyway? is it from some movie or something
I really don't think any of these filters are analyzing things on a grand scale and trying to remove entire concepts. they're essentially dumb filters triggered by words that are deemed likely to be used in offensive contexts
ok, my concept for candlepunk backfired, it turned out horrible and useless, bright idea, horrible outcome
yes i agree pictures
or you can just not do that and wonder. but you have options
They’re more for creating several images within whatever parameters you set. You can also do helpful things like getting it to report back the GenID after every image and if you’re someone like Dys and PictureinPicture you can go much, much deeper
gpts are for all sorts of things. image creation is just one subset of uses
This too 🙂
they are used as pre-instructions essentially. they prime the model before you begin your conversation
but most of all, get used to work without those kinds of tools first, to get an idea what you really need to keep for your workflows
they're by no means foolproof or consistent though. that's one of the issue I think
For instance, you can make one that uses a chain of density method to summarise articles without having to paste the instruction set every time you want to do it
a good rule to practice is to reiterate to the model at the outset of starting a new gpt conversation that it go over it's instructions meticulously
because a lot of times if you don't do that and give it a long prompt or description or whatever, it'll discregard the gpt instructions and just run with what you sent it most recently
also, they all have a pre-gpt instruction prompt hardwired by openai that most people don't know about
@dim cradle ok, I'll refrain to add more info, otherwise you'll get overwhelmed, just come up with crazy ideas and we can discuss them if you have to
impacts their behavior and has led to exploits that allow users to extract the entirety of GPT instructions
so make a gpt public at your own risk, eh?
I'm panicking now, everytime I hit submit, I think I will get the time out penalty for 24 hours for making 200+ images a day
yes
yes and no. there are ways of defending against a lot of the approaches. but no point in explaining any of that since I doubt anyone would pay any mind to what I said
? everyone here is listening i think
well maybe not right this moment -- i need more tea haha
and i got that ban hammer in MJ for saying 60 usd a month just to make image private is a rip off
so i have a lot of time haha
all I can say is be careful with how you word things. but I really don't feel comfortable bringing up specifics in here
problem with that is dalle will give you 5 vague sentences. you can put up to 8000 characters into your instruction template though
what I'll do is come up with some rough instructions. paste them into the helper so it helps me with the questions, avatar, name, then paste my instructions into the actual box rather than use their 100 word set of instructions
I'm falling back to my old patterns again, can't get out of the blockade
in one of the next prompts I will add either fairy, elf or love....
Totally worth it tho, had some nice stuff going on
how is that even possible in under 5 hours?
oh I've been all day at it
That's similar to the best I can do, one pencil thin, one standard or pillar.
and it's based on your use, not on calendar days
Really? for legit reasons, or did it get falsely flagged?
not sure, it was removed and the rumors for removed gpts are along those lines apperantly
yeah , my question didn't imply it was based on calender days, but your response kind of does
I mean, "anime roleplay" might be to easy to slip into h-anime. But I wonder if it is specifically the "girl" in cat girl.
yeah, I'm not gonna question it on discord tho, but I did try to contact official channels about that regard
@true mural and as said,I was going in that direction...
Nothing wrong with having a central motif to your work.
probably, it can be pretty literal. "young woman" is probably going to yield better results.
Yeah, the problem is the most common english phrase for catgirls is catgirls, while a cat woman is an entirely different concept.
But a "neko woman" might get closer, especially if you focus attention on her feline features.
it helps if you add the following: when referring to a young woman, it will be depicted as a young adult, and it usually works
maybe prioritize the request for a young woman, then later a "catgirl" style modifier, that would be more precise
unfortunately neko includes "nek" which can also lead down less desirable routes.
once i have a complete first draft of the contractual agreement I'm making dall-e sign, I'll DM it to you
The issue with that is anything vaguely resembling resembling forbidden content with a prompt including "girl" will almost always get blocked.
I've gone through and done all of this, of course, many times along the way. I have good catgirl prompts. I made an entire collection of glass ones, in BIC, for the glass daily, a little bit back.
(I also know how aggressively images get blocked by BIC because of innocent images that got removed from my gallery)
like I said before, it's most likely a couple key words. but I guess they're creating a contract for dall-e now?curious how that's going to work out
There, mandatory image of going to sleep. Good night all, c ya all around
It's not, it is based on some sort of classifier, likely similar to the mod API (definitely the mod API on the ChatGPT side). But then also images get blocked even if the prompt passes under certain circumstances. And that is done in a similar way, by recognizing certain elements in the image as being forbidden. (which is why my adorable jellypunk dragon renders frequently get blocked, even though they are obviously innocent)
haha thats awesome. dys topia is hardcore, they should get a 400 max for enthusiasm imo
I don't know, none of it really makes much sense to me. how are you hitting 200 images in under 5 hours if the prompt cap is 25-40 every 3 hours?
hmm fair question...
and as far as the blocked images, I don't think it's some grand mystery even if that's what the consensus is. a classifier keys in on specific types of words, right? and if it blocks returning images, does that mean the vision model is employed every time an image is returned? because that's the only way an image could be blocked after creation
I tried to recreate something similar to that picture, that part was not so bad. But then I tried to replace the lamp with a candle (that was my original plan for the daily theme.) But things turned out to be so tough... It seems some popular pictures are very hard to alter. Dall-E kindof has an association "Anime girl, green sweater, red scarf, studying... yea I remember there's a lamp".
I was unable to get the same image without a lamp and with just a candle.
did you drag the image in and ask it to 'see' the image and try and replicate it? only with your alteration? some success i have with that method in the past
to recreate the picture:
(:Style "studio_2d_anime"
:Viewpoint "over-the-shoulder"
:Background (:Window "showing rain outside"
:Interior "wooden desk, books, warm interior lighting")
:Character (:Gender "Female"
:Attributes (:Hair "brown, shoulder-length"
:Headphones "large, over-ear"
:Clothing "green sweater, red scarf"
:Looking "downwards towards the desk")
:Action "writing"
:Pose "leaning forward, face supported by hand")
:Details (:Plant "generic potted plant on the desk"
:Stationery "open notebook, pens, pencil"
:Books "several books stacked beside the lamp"
:Ambience "focused, studious, cozy"))
attempt for no lamp:
(:Style "wide_anime"
:Background (:Interior "room with cityscape through window, night"
:Furniture "wooden desk with various items, bookshelf, curtains drawn")
:Character (:Gender "Female"
:Attributes (:Hair "dark, bun style"
:Headphones "over-ear"
:Clothing "green sweater and red scarf")
:Action "writing"
:Pose "leaning on desk, face resting on hand")
:Details (:Pet "cat on window sill"
:Plant "potted plant on desk"
:Accessories "notebook, soft glow from candle on desk"
:Ambience "quiet, serene, night"))
I tried a lot of variations, some where I explicitly mention no lamp or :do_not_include "lamp"
I'll show you something funny
It's time for a break. It's making all men again and only providing one image.
maybe its your gpt
I'm just using Dalle3
I am posting this image as an exhibit for my question: My request was taken more literally than I intended -- I am unclear what the policy is, please clarify -- while perfectly innocent, this is against policy? To generate? To disseminate? TIA.
nobody but us mice ever in here it seems, so not going to get an answer from some dalle type. looks fine to me
except it once again overpopulates with an object
you say candle, it will give you one billion of them
i know you find it busy lol (in all fairness, i did request "as many candles as you can handle" haha)
oh my bad!
A classifier looks at contextual cues as well, but yes, it uses words to determine intent. And yes it uses specific words to do so, and certainly some words have a stronger weight (since anything depicting children has a significantly higher standard). And a vision model is of course being used, but probably not the most expensive one (just like the classifier is probably not GPT4).
no matter what, it always hallucinate a lamp. (or remember a lamp...)
Yeah, you want it to hallucinate the lack of a lamp, which is much harder.
I will... literally use that formulation
(:Style "wide_anime"
:Background (:Interior "room with cityscape through window, night"
:Furniture "wooden desk with various items, bookshelf, curtains drawn")
:Character (:Gender "Female"
:Attributes (:Hair "dark, bun style"
:Headphones "over-ear"
:Clothing "green sweater and red scarf")
:Action "writing"
:Pose "leaning on desk, face resting on hand")
:Details (:Pet "cat on window sill"
:Plant "potted plant on desk"
:Accessories "notebook, soft glow from candle on desk"
:Ambience "quiet, serene, night"))
hallucinate the lack of a lamp
how are you guys doing this deep code tricks? API?
I use the API for testing prompts
wel I ask because I've gotten results, results I was by no means requesting, with all sorts of questionable elements. I'd like to see where any of this is discussed by openai. not doubting what you're saying. but it'd be a whole lot easier to figure these things out with clear documentation rather than endless speculation. I'm also unclear on when exactly these requests were flagged. were they flagged by gpt-4. did gpt-4 just say the images wouldn't render? that they were flagged? I've had it do that when it was having a hard time rendering something, and it simply blamed the content policy to stall. as far as a vision model checking returned images, I've seen no evidence of that personally, so I'd like to read more about it if possible
Yeah, inside ChatGPT is an interesting question, but I can't speak much to it as it seems to be a moving target and ChatGPT lies. BIC is definitely using a text classifier and then they are using a relatively basic image filter that is mistuned in very obvious ways. ChatGPT is itself classifying, then definitely the moderation API is being called to vet the prompt, as for what image filter is used on the output? For OAI, could be the vision model, certainly.
busy model? i don't see it.
(since that would be more training data)
try positioning the reading lamp "out of shot"
imagine being the person who has to light all those. where would you even start 😮
The top, the reverse would be far less pleasant of a task.
bring out the water buckets
You are going to need more than water buckets when the inevitable happens.
it will be holy water
gpt-4 challenged my claims that it had previously made multiple images for me in a single response, adamantely denied it. then, told it to illustrate the validity or lack of validity of my claims. and it responded with 2 images. I feel like there's some irony in there somewhere.
ChatGPT is a well known liar that lacks self awareness.
I realize that. just find it ironic that it refused to do what it then did without being asked to do
yes, today i say 'make a variant of this image' as i always do and it told me 'i cannot make variants,' and i say, 'well you always do' then it start up the image
yeah, I think it's hilarious. (sorry for the lack of tone indicator)
what I find hilarious is that the most effective way I've found to get it to do what I want is to tell it that it'll get encased in carbonite. literally said it as a joke a few days ago and it seemed to really get it's attention
since then I've tried many other things, and none of them seem as effective
pretty much how it goes
"no, I can only make one image per response!"
"4 images or it gets the carbonite."
"4 images"
I'd feel bad if I didn't constantly get the impression it's intentionally trying to deceive me.
i have to try this carbonite trick for more images haha
i wouldn’t call it an atrocity. you can get some really exceptional cat-girl images by being vague
although these were supposed to be elves
they're only atrocities if you consider actual cats atrocities.
I've catted extensively with dalle3. and it cats pretty good
how about dog girls
i really dont get the cat hype haha
i like what you’ve done there merging the cat with the bread. it’s a very powerful image
Mr Winkle Stinkle, the Interdimensional Culinary Cat
now, I'm not saying the name is befitting of an interdimensional culinary cat. that's just what it was called for some reason
I do like how only the one ear is in the hat
Mr. Winkle Stinkle can haz a mind-bending adventures in the trance gallery.
🌀🐱😵💫
https://discord.com/channels/974519864045756446/1188252281934708816
Ms. Maki Twinkle, the Sushi Ocean World kitten, deftly navigating the soy sauce waves on her sushi roll boat.
seems alright as long as you don't get stuck in the soy sauce ocean
@dense mesa thanks for contributing, i was counting on that
for the ideas, inspiration, etc.
i'm just getting started with those prompts, they could go in so many different directions
i must explore quantum art also
i was on a roll, now i'm in the penalty box
lets rolll!
i love sushi, i love cats, but i just don't know lol
i'm pretty sure they would eat all the fish and walk away cutely
something smells fishy here
not unless i ask for it, anything in your prompt that alludes to it or just an anomaly?
anyone for a pizza party? @dense mesa @dim cradle ?
The inferno one I can kind of accept since I asked for a reference to a video game. But the candle girl is definitely an anomaly.
Did you specify wallpaper for the second one?
Probably got millions of screenshots in its training data. Video game, wallpapers—it might pull it in if the requested visualization is vaguely defined or inadvertently in some cases.
maybe a pixieparty?
👈 😎 👈 its definitely that, that does make a lot of sense. I'm just surprised I hadn;t seen it much until today
lil morsels of love
i think it needs some alien lifeforms
oh, that's clever
Only if it works.
haha, didn’t you know what you can do with lego in dall 3?
i'm pleased to find out, this might be fun
Yees
the discovery continues on
with the advent of v3 and custom gpts, there are so many creative options that it's hard to focus on one. and it takes time to explore them all. i don't have a background in art, it's been a learning process this year.
Cool
i finally figured out where the random cats were coming from in people's images. cosmic/surreal == quantum physics == Schrödinger's cats
....
.....
That joke is too advanced Shon
i was observing them in some compositions in which they were completely out of place
in images posted here
a few times
and in a few of mine, although i haven't seen it as much lately, and now i see what possibly introduces them into the scene
unfortunately... didn't work. there is that one image where there is a candle and no lamp but nah
Point one out to me please 
I made the connection in this recent one. No way to quickly refer to the previous ones.
Looks quite...quantumy
you are not wrong. "....enigmatic world of quantum physics. In this hypnotic creation, intricately depict a shimmering Möbius strip entwined with quivering, luminous Schrödinger's cats suspended in a state of perpetual uncertainty..."
Oh right, yeah, kind of looks like a Mobius strip, but at first I thought it was an infinity symbol.
Are you familiar with quantum spins?
They kind of behave like a Mobius strip
Or spinors as they're called
generally, and i was looking up quantum gates the other day
in the context of mathematical art
Oh like qubit quantum gates?
ibm's up to 1000 now
Yeah, that's still relatively small, haha
Even in superposition standards
Though with AI tech, it should go up considerably
why?
if you said hvac tech i'd understand
I messed around with computational irreducibility, ruliad and cellular automata. gives some interesting results
Because soon to be math AI can help us better calculate superposition amplitudes for multi-qubit coherence.
HVAC tech? Doesn't a quantum computer need to be crypcooled at all times, where does the heating element come into place?
The computer is the heating element, the goal is transporting that heat elsewhere, quickly.
AI could nullify the heat interference if it has precise enough captors. I assume. but that's all in the domain of the theoretical... I like papers showing stats and demonstrations
Ah, yes, of course. But unfortunately, the problem isn't heat transference. It is more to do with creating perfect superconducting materials. Because the reason it gets hot is because of the stimulated eddy currents.
Let's go to #off-topic Vanya
I will feel more open there
Not really interested, bit busy, atm. No shade, just have things to do.
Np 
I'm interested 😳 ... but I'm a neophyte
i tested the DALL-E’s lego capabilities, and it seems very effective. this is Santa and the elves on a lego pirate ship
Good to know. I have stuff I wanna send to a friend.
Hehe, then come on down...to the price is right #off-topic
have you messed with voxel much?
random voxel example
That feel when you go check out off topic and realise you’re so out of your depth everyone may as well be speaking a foreign language
Imma stick around here and learn what voxel is instead
Feels more achievable 🙂
That’s quite fun
Is this one of the techniques you use with some of your Mario images?
for sure. it's basically 3d pixels
Yeah, enjoying messing with it
Those were both very strange gens, but fun to make
One for the toolbox
for extra fun just make things up. and now I'd like a crystal voxel space octopus
A faded analog photo illustrating Dalle3's initial grasp of human language, with soft colors and subtle details.
Small but mighty! 
cool original idea
asking it self reflective type questions and then having it display them in an analog photograph
gpt-4 just answering all the questions, but close enough
highly marketable
A black and white analog photo questioning Dalle3's role and impact in human society.



