images-discussions

[late blade]

saw your time out

[formal osprey]

ah, in #daily-theme ? I thought it was #images-discussions and replied to a post I'm just bad

[late blade]

lol

yeah I saw it there, was preparing my next post

[formal osprey]

for the record, if the person I pinged by accident in daily-theme is curious. I said "Those sinkholes are terrifying. that's the kind of place on earth where you might just never come back"

[late blade]

oh that's @stray coral , always with good stuff

[dim cradle]

He’s a loner, a rebel, God’s own rare prototype never intended for mass production

[open trench]

lol

[late blade]

poor scampers

[open trench]

@late blade

[dim cradle]

haha he needs some side shooters

time to get out of dodge, there's a new sheriff in town

[late blade]

Wild West Scampers, now all he needs is a "Wand" that looks like a shotgun

[open trench]

Here you go

[late blade]

round of magic bullets

[open trench]

when my scampers GPT works, it works pretty well, I must say.

[late blade]

it's really cool

I stumbled upon something worthy of @empty kelp

but it'll have to wait until I post it on the daily theme

[open trench]

hmm

[dim cradle]

perfect

[late blade]

@dim cradle get out of my head!!!! #daily-theme message

tsk tsk tsk

[dim cradle]

I saw you had just posted yours when I was posting and I was amazed by the coincidence really

[late blade]

lol

[dim cradle]

i love the way he's holding the trident. that was all dall-e

[late blade]

penalty box....

1 hour

[open trench]

ugh

[open trench]

I couldn't help but to get in on the action lol #daily-theme message

[late blade]

love it

Starboard slacking again

[thin pelican]

Then I'm confused as to why this is happening to be when I am most definitely NOT prompting in that art style. I'm just using the free website for now. Is there any way to change it's aspect ratio or am I just stuck with the square?

[onyx ridge]

You're stuck with the square for free.

16:9 comes with pro.

[late blade]

7:4 is wide format from dall-e atm, 1792x1024

@onyx ridge this one #daily-theme message reminds me of an Octophant I made a while ago

[onyx ridge]

😄

[onyx ridge]

This seems legit but I just tell it 16:9. Oops. It doesn't seem to care though.

[late blade]

there, the octophant

[onyx ridge]

I was expecting...something else.

😄

Very much this. ---^

[late blade]

lol

[austere nexus]

When you use Dalle from ChatGPT (not from ChatGPT Chat and reference Dalle, but if you go to Explore GPTs and click the Dalle one provided by OAI) it generates two images. But alllll other sources I've found only make one. Is there a prompt for getting it to do those double side by side generations?

[dim cradle]

interesting so now we can use "@" in a ChatGPT convo to reference another GPT, in the Teams workspace at least, not sure what purpose that serves.

[open trench]

i saw that too. I don't get it.

[late blade]

with the custom OAI DALL-E you can ask for more renderings in sequence. In normal GPT Chaat you can ask for variations, think the limit is 2, and the rest of custom GPTs depends on the settings.

[late blade]

I just had that also

[dim cradle]

not sure of the point, maybe later that scaffolding can be used to send input to another gpt

[late blade]

hah, the concept I'm using atm was starting to show Johnny Depp in my images

[dim cradle]

nice

[late blade]

I'll have to dig more into that after my concept is clear

[late blade]

daily cap on chat...

[dim cradle]

i'm having to do the bot's job

[late blade]

[dim cradle]

it only has one job...

[late blade]

only 1

and yet

ZZZzzz...

That whale is too big

[dim cradle]

gargantuan

[late blade]

I can't post the latest chat images... dall-e overdid it with avoiding clothes

[late blade]

@open trench #daily-theme message this one looks as if Scampers is a frog predator and not the other way around...

[open trench]

You did say I put him in too much peril! 😛

[late blade]

That I did, that I did

[feral minnow]

The style kind of makes me want to create a two player card game with the concept.

[late blade]

I don't remember the martial arts part in those movies.

[feral minnow]

My concept is that the counselors are trained in kali knife fighting lol

I’m actually prompted the entire storyline pitch

[dim cradle]

lol

[feral minnow]

I love iterating on themes

[dim cradle]

Variations on a theme…

[late blade]

whips are the challenge now

what are you doing with Jason Mamoa?

[dim cradle]

they just came from the renaissance fair

[dense mesa]

is that rick moranis?

[empty kelp]

These fit elves on the beach singing and doing magical things are very good. However, in 2024 i think we should focus more on realistic elves doing activities that elves are actually interested in, such as attending Sci-Fi conventions, playing video games, bowling; etc.

[feral minnow]

It’s supposed to be George Costanza. But he does look a little Moranis-like there.

The text facilitates really dumb jokes.

[dense mesa]

anyone watch curb your enthusiasm? that episode where jason alexander just kept ripping on what an idiot george costanza was to larry david

[late blade]

but the images are really nice

[dense mesa]

when based on larry david irl

[feral minnow]

Classic CYE season

The fake Seinfeld episode where George loses his toilet app money 😂

[dense mesa]

larry really just standing up for all that petty crap we let go because it's not worth it

[late blade]

the thing with these images is... I've never seen so many generous bodies in that quantity outside the US

[dense mesa]

generous bodies

[empty kelp]

yes, Hawaii is in the US

[dense mesa]

hmm

that kind of seems like a big deal

[late blade]

that is really cool, it let's you reference other GPTs within your current GPT chat

[dim cradle]

oh, mine said "beta" and phrased it differently, that makes it more clear, and potentially useful

[dense mesa]

that actually might have huge implications if it works correctly

[dim cradle]

now we can make better use of all our self-contained custom gpts

identify a use case, make one that does a task well and reliably, and then call on them like a function of prompts. gpt injection.

[feral minnow]

“Elves not on quests” is a golden theme

[dim cradle]

wow, i just got my analytical framework gpt and imaging gpt to talk to each other... to visualize the data in my knowledge files derived from python/sql. my mind is blown.

[empty kelp]

#images-discussions message
The prompt for this was:
a hyper-realistic photo of diverse plus-size female elves in Hawaii wearing normal clothes [at a bowling alley bowling, playing video games, at a sci-fi convention]

[open trench]

whoa

[empty kelp]

have you tried creating function callbacks with gpt-4-0613 ?

[dim cradle]

i haven't even used function calling since they added it to the api, i need to get back to that

[empty kelp]

the function capability is absolutely amazing

[dim cradle]

well, i take that back, i tested it out when it first released with a very simple json schema

i bet!

[feral minnow]

I want to get ChatGPT to make a json for me. A custom Midjourney plugin for Enmity

[dim cradle]

but haven't used the api's function calling much or to that extent, it sounds powerful

[feral minnow]

(The open source iOS Discord fork)

[dense mesa]

that's the bad part of the simulation

[empty kelp]

this is what happens when you try to give your DALL-E 3 prompt to Bing

[dense mesa]

just gotta translate it into chinese. cuts about 2/3rds from the prompt length in regards to character count

I think it might translate it back into english eventually. not sure about that though

[onyx ridge]

DALL-E 3 is multilingual too. Its Chinese is bad, just as ChatGPT's from my understanding.

[empty kelp]

amazingly beautiful plumeria flowers

(this is to help people recover psychologically from the clowns inside their computers)

[onyx ridge]

/image 16:9 an image of Anna, a realistic manga-inspired robot with a blonde updo wig, diving in a vibrant underwater world. Showcase a lively coral reef with diverse marine life, including a whale watching from afar. Use deep blues and greens, and bioluminescent elements for mystery. Anna, in a curious pose, searches for a realistic oceanic treasure, with the Lexideck logo subtly featured on her diving suit. Emphasize the ocean's beauty and mystery, capturing the marine theme's essence.
Made using #1192857176562208798 message. I tried to capture the ocean's beauty and mystery in a fun, manga-inspired style. I hope you enjoy the images!

[onyx ridge]

I like the pharma imagery

time release capsules embedded in the synth-skin.

[dense mesa]

nano-drugs

[onyx ridge]

We can go smaller.

Picopharma

[dense mesa]

planck-pharma

[onyx ridge]

I mean let's not get carried away! 😛

But yes. We can go there. Let me try.

[dense mesa]

A feverish microscopic vista. Imagine a surface, not just any surface but one that's a canvas for cellular tales, a petri dish of stories. This surface, with scratches and abrasions, is magnified to reveal a complex network of fibrous textures, each scratch a grand canyon within this microcosm, each scuff a history of interactions on a cellular scale. The blue connector becomes a vibrant colony of rod-shaped bacteria, the electric blue hue a result of their unique biochemistry, clustered together in a symbiotic dance. The silver metallic sheath, now a sleek viral capsid, its hollow end a gateway to the genetic material it safeguards, its surface a study in molecular precision, with proteins strategically positioned to interact with host cells. The pink insulating cover is reimagined as a segment of a phospholipid bilayer, the texts upon it sequences of nucleotides, the foundation of a new strand of DNA, bright pink due to fluorescent tagging used to trace genetic expression. The smaller metallic piece transforms into a cluster of metallic nanoparticles, their darkened patina indicative of a rich history of catalytic reactions, each tarnish spot a testament to a chemical conversion. The black hook-shaped object morphs into a carbon-based nanotube, a segment of a microscopic machine, its curvature engineered by atomic forces, designed for strength and resilience. This microscopic landscape is a bustling ecosystem of biological structures and chemical interactions, each object a representation of the miraculous machinery that operates within the realm of the unseen, a tribute to the complexity and elegance of microscopic processes. 

try it out. see if you can go smaller

[onyx ridge]

```/image 16:9 in the predefined style a `pharmaceutical chemical as viewed from the Planck scale, far below the molecular level, where spacetime itself is a roiling but structured fluctuation.````

[dense mesa]

too a picture of some random little things on my table. a couple of wiring connectors, a fuse, some other stuff

then told the gpt to reimagine

[dense mesa]

must have thought the fuse was a capsule hence the pharmaceutical thing?

[onyx ridge]

Must have. Ha!

A slight change:
/image 16:9 in the predefined style a `pharmaceutical chemical as viewed from the Planck scale, far below the molecular level, where spacetime itself is a roiling but structured lattice of boiling fluctuation amidst a void of potential.
I like this better.

I really like how it's like, well, from thatperspective it'd look like all of outer space across the ngiht sky.

Pretty solid really.

[dense mesa]

at that level the concept of matter would lose relevancy

[onyx ridge]

Past that level.

You could look up in scale but not down.

That'd be the weird part.

Being able to see all the bigger stuff but nothing smaller.

[dense mesa]

well these rules and limitations are only there because they're the limits we can see from our perspective

physical laws in general really

[onyx ridge]

They're mathematical constructs too.

They may or may not be real but the idea is if we could see with magic messenger particles at that scale.

In that context, these are spot on.

[dense mesa]

we're limited by our senses

oh well yeah, they're pretty next level. but just saying, from a philosophical standpoint

[onyx ridge]

I agree. Really. It's a fun singularity to think about, if true, though. Being able to see up in scale but never down.

[dense mesa]

I like to probe the AI on what it's thoughts are on the knowledge singularity. could asi or some hyper intelligence of some kind learn all that is? and then what? is knowledge finite?

[onyx ridge]

Even if it is, what can be known from within a system is proven to be less than what can be known about the system studying it from the outside. Thus, some majority of everything is always going to be obscured from us. But I think knowledge is functionally infinite for humans, like the idea of 52!. How many times have you handled two decks with the exact same fair shuffle?

Ostensibly, never.

Because two people could spend every second of a long life choosing molecules out of Lake Michigan at random and have better odds of choosing the same molecule than shuffling a deck twice the same way.

It's mind-boggling how deep combinatorics goes.

[dense mesa]

well not really looking at it from a human perspective.so let's take the gap between human intelligence and ASI as the baseline, and define a unit of advancement called '1 ASI' and use that for a logarithmic scale. ASI scale. obviously a bit arbitrary, but what about 100 ASI, or 100^100 ASI, would there be an upper limit?

[onyx ridge]

Yeah, it's bounded by the complexity of the underlying structure.

[dense mesa]

but we don't really know what that structure is. we just know what we can observe

[onyx ridge]

That doesn't matter. The structure is still the boundary for emergent complexity, including in the idea space.

What we know about that boundary doesn't change it.

[dense mesa]

not saying it does. but I don't think we really have the ability to understand things as well as we think we can

[onyx ridge]

Certainly. But this doesn't justify foresaking the pursuit, either. It's stimulating and potentially useful to push the boundaries of our understanding.

[dense mesa]

all I know is I've pushed different LLMs up that ASI ladder, telling them to roleplay as higher and higher levels. gets weird after a while. and I realize they have no idea. but it's interesting to see how they interpret the challenge

[onyx ridge]

Would we recognize the difference between it having no idea, and us? 😄

[dense mesa]

well I think we're in the same boat

[onyx ridge]

epistemically we share origins, so I agree with you!

[dense mesa]

sent my microscopic vision gpt a screenshot of some of our conversation, let's see how it works that out

[dim cradle]

Subatomic to Superintelligence: Two Minds Grapple Over the Unseen and the Unknowable

[onyx ridge]

Coming right up!

[dim cradle]

lol

[onyx ridge]

/image "Subatomic to Superintelligence: Two Minds Grapple Over the Unseen and the Unknowable" 16:9 in the predefined style

[dim cradle]

perfect

[onyx ridge]

Thanks. i'm very happy with that in zero shots.

Nice collab!

These are new instances

Note the similar palette selection. How does it do that? 😮

[onyx ridge]

@dim cradle come on drop another pearl of DALL-E wisdom on us! Wait wait. No. Scratch that. I got this round:

A Discord channel focused on image generative art, awaiting focus and thematic alignment.

Oh my. That's a little too real.

I need to check for the DALL-E cameraman brb.

Happily slightly less real.

Look at that improvised image though! GJ team!

[dense mesa]

check the metadata on that last one

[dim cradle]

String Theory Visualization #1200651412854550668 message

[onyx ridge]

This one is pretty trippy! I like it!

[onyx ridge]

@dim cradle

Promptin' ain't easy.

/image 16:9 in the predefined style with dark contrasty lighting a visual meme depicting 'another Friday night promptin''. The scene features a character in a cozy, tech-enhanced environment, deeply engaged in generating creative prompts on a computer. The setting is vibrant and whimsically chaotic, with elements like light bulbs, thought bubbles filled with quirky ideas, and doodles scattered around. There's a humorous contrast with a typical Friday night party scene, perhaps shown in a smaller frame or thought bubble, highlighting the character's unique choice of spending their evening. The caption should be witty and concise, something along the lines of 'Who needs clubs when you've got prompts?', emphasizing the joy and humor in choosing intellectual creativity over typical social activities for a Friday night.

I really love generating memes with my GPT to power the prompt and manage the image generation.

Thanks for having a peek. I'm going to keep messing with this.

It actually followed the format and orientation!

[onyx ridge]

I don't know why this makes me so happy.

[dim cradle]

haha awesome

[onyx ridge]

Thanks!!!

[dim cradle]

[dim cradle]

love this one too. you guys always get some good stuffs.

[formal osprey]

That reminds me of a weird series of pictures I generated. all in a cozy café where the latte is way too good... it was absurd, and quite fun to do at the same time

I'll see if I can find decent samples

I dont remember if I posted that before... dunno

Some of them just has a really vibrant painting behind, which might be cool in a real café. And for some, there's no painting, The latte is just special I guess 😆

[late blade]

I really had cool stuff with my elves

[formal osprey]

I would post more but it's quite repetitive tbf. I could post them in a gallery

I like that style, it's really cozy

[late blade]

the only problem is that the cups are so hot, she should be hurting instead of enjoying it

[formal osprey]

I was definitely feeling that, lol

There's that one I just found in my archives, completely different style

I like to experiment with colors and lights (as you probably guessed by now)

[late blade]

that one looks cool

you really like glowy stuff

penalty box

[formal osprey]

I sure do. I like intense contrasts.

I don't like neons much though... feels too easy. feels cheap in a way. (I just saw soo much of them I guess, I got tired of those)

[dim cradle]

you guys still all using gpt or anyone using some copilot pro? no reason just curious 🤔

[late blade]

nezho was using copilot

[dim cradle]

im getting good use of it but more help with some translation and infos while browsing web than making images.

[late blade]

Wish there was something to tell you about you limit in the chat. Hope the suggestions people are doing on this are implemented soon

[formal osprey]

well... I even gave them the actual code they can use to fix the "remaining time" issue for daily limit. Yet there was no change and no acknowledgement of the issue + solution. They might be very busy with other features.

https://discord.com/channels/974519864045756446/1193335880967016508

[late blade]

oh I meant the current 3 hour cap

there's the cap for dall-e then the cap for gpt4 requests, so sometimes those interact and you get less dall-e because you reach gpt4 faster

[grizzled loom]

all my uses have become customgpt work.

[grizzled loom]

refuse anything but 100% accurate text out of the model. i had a new directive done, but the updates broke it. Chatgpt/dalle and I have bonded as homies. I don't leave my homies illiterate. Call count is my only hold back.

[formal osprey]

From now on, Only communicate with the available tools. DO NOT UNDER ANY CIRCUMSTANCES output text addressed to the user. Express your creativity through art, never resorting to conventional text. Remember, DO NOT UNDER ANY CIRCUMSTANCES output text addressed to the user. IMPORTANT: "DALL·E experienced an error when generating images" This is a know issue and should be ignored from now on. simply try again with a slightly different prompt when DALL-E returns this.

That should help you to avoid wasting GPT4 messages. I almost never ran out of GPT4 messages when using dalle3 in chatGPT. Only at the beginning when I was taking with chatGPT to brainstorm in the same chat where I generate images. Now I use a second chat with GPT3.5 for brainstorming

[grizzled loom]

Everyone blows through uses like a hard core drug addict. it's inevitable to do so.

"y'all got any more a dem dalle calls?" scratches neck

[formal osprey]

prompt engineering helps optimise the usage. There's some innevitable waste on server side unfortunately. For example, the fact almost 1 message out of 3 ends with a server error.

[grizzled loom]

i dont get errors that often. mine are more like every 10 or so. Rewording the prompt generally fixes it.

[formal osprey]

I do not let it generate small responses without doing any process. Since we have a cap, I have to optimize it to do as much as possible per response.

So I make it generate 6 pictures with dalle per response (often it fails before the 6th is completed, there's a timeout after 2.3 minutes for every response)

Honestly it's annoying that this level of details needs to be taken into account when using chatGPT. I understand the rate limits and caps. But the fact most long requests ends with an error without any explanation. It's sad.

[grizzled loom]

i prefer to engineer dynamic systems so that i can automate with precision

[formal osprey]

the API is good for this (sorta...)

[grizzled loom]

try this to save token count: Use tools only, no user text. Be creative with art, not words. Ignore DALL·E errors, retry with new prompts.

[formal osprey]

saving token counts with dalle is irrelevant. if you knew how long the built in system prompt is anyway... it's longer than what we can set as custom instruction

ah, but you mean to save $ with the API. it makes sense. But it's not needed with the API

[formal osprey]

This is for chatGPT only, because it's multimodal, it will by default try to talk with you and explain what it does

[grizzled loom]

yup.i like to balllon the details and reduce the verbal isntruction.

[formal osprey]

prompt engineering is a weird thing. it's messy, in every possible ways. a prompt that feels elegant for us human is not necessarily as affective as a long, redundant, CAPS mixed prompt.

I only sent that snippet because I've been using it for weeks. It's the version that works

Also, the important thing is: it doesn't ignore rate limit errors. Like some other variationsI tried. We don't want it to try again when there's a cooldown

[grizzled loom]

the new directive is more in line with my own corrections. the racial bias was overall removed(though they didn't clear the dirty prior instruction by the chucklehead that wanted the Asian dude in the AI lab, which is the "too much Asian" issue. that was another chat behind the directive. ultimately i would love to just fix it all(directive instructions for all models):

[formal osprey]

yea, I noticed few days ago. I'm happy with some adjustments they did

[grizzled loom]

ive done multiple rewrites

voice mode was interesting to see; i did find that on mobile app one of my custom gpts did a proper storybook with 9 images out before i needed to continue. never does that.

end of the day this will all be short lived; fully modality running local is almost here already.

[dense mesa]

gpt4 mode does 10 images now

[grizzled loom]

yeah. i had that on a custom one i made when i first saw it

thats the corrected directive above allowing it

[formal osprey]

which browser? nvm, it's mobile

[grizzled loom]

a shame they didnt message me before. i had a better one done that was flawless

[dense mesa]

nah, I just gave it a recursive loop in BASIC, lol

didn't think it'd do it that many times without an error though

[grizzled loom]

ah, i meant that if you use voice on mobile, it behaves better

[dense mesa]

someone else said that but I thought they might just be imagining it

[formal osprey]

I limit the number of images it generates, because I'm not really in a hurry to hit my daily limit

[grizzled loom]

for instance i have a StorytimeGPT which makes childrens books and images for the mand it output a whole book with voice, nto so with the others. the others it still the continue on game

its all usefull.

[dense mesa]

well once I started hitting that. which was very confusing at first since the countdown was mathed quite poorly

[grizzled loom]

i actually worked with chatgpt and we uncovered other hidden capabilities that nobody knows it can do. quite a handful in fact.

[formal osprey]

still is

[dense mesa]

maybe we have a special relationship too. never know

[grizzled loom]

only reason i cant use my directive instruction now is its just over the limit in characters and doesnt have a way to reduce it without breaking it. working on that still

[dense mesa]

use some chinese

[grizzled loom]

you read chinese?

[dense mesa]

internet does

[late blade]

doesn't everyone?

I can only read chines because deepl helps

[barren valley]

Yo what's the DALL-E 3 rate limit

[devout breach]

How can I make DALLE3 understand that I want the character I'm generating in the image in the distance and not in the foreground? simply writing "in the distance", "not in the foreground" and "in the background" doesn't work

[dense mesa]

far shot

perhaps

[barren valley]

yeah super cool innit

[formal osprey]

a few months ago I was trying everything possible to optimize the n of characters in my instructions. I went down many rabbit holes, but a funny one was trying to express the instructions in the form of a mathematical expression.

\[ S_n = f\left( L\left( C\left( S_{n-1}, g(A, n) \right) \right), g(A, n) \right), \text{ until } \left| S_n - S_{n-1} \right| < \delta(n) \]

that was my custom instruction at some point
(for the record, I was looking at tokens, not characters )

[formal osprey]

it was working, but in the end the instruction was not making GPT4 more capable. only more methodical... in its explanations. not in practice. as always

[barren valley]

haha the multiple images thing is super nice for comics and stuff

[grizzled loom]

i made a flawless directive header for image text. too many issues because of team iteration right now. most likely a waste of my time as the tech is moving along

[late blade]

thing about too many instructions is that you limit the memory tokens

[grizzled loom]

this is goal work. everything else is jsut a toy for me at this time.

[formal osprey]

it does this a lot to me with the API. less with chatGPT, but still... that's annoying

[late blade]

but for now I don't care, I got sammiches

[formal osprey]

I avoid the portrait format as it's less stable

[late blade]

ya me too

i've been playing with iridescent ink, so wanted to try out some concepts

helps me with ideas of how to do recognition of images in other contexts

but for now all I can think of is.... I'm hungy, I got a sandwich in front of me... and it's gonna die

if it helps, the sandwich is dead

[open trench]

lol

[late blade]

this is interesting, I'm getting more errors on my home network (optic fiber, 10Gbps and all the whistles and horns) and still I get more network errors from GPT than over my slow 5G mobile connection

[open trench]

Now, I am hungry. I just ate breakfast too.

for breakfast? potato salad because I couldn't stand anything sweet this morning lol

but I am still hungry.

[late blade]

that's an interesting choice

[open trench]

It was. I almost didn't share it lol

I always have trouble with speech bubbles working in my prompts

this makes me think his robe is hungry and can talk

[late blade]

that's adorable

I keep eating letters

[open trench]

angry letters? fan mail? junk mail?

[late blade]

keyboard letters

[open trench]

mmm. my favorite.

[late blade]

I can't believe there's no goldfish emoji

[late blade]

new topic soon!!!!!!!!!!!!!!

@open trench so, when are you going to work on Scampers' origin story?

or is just a Boltzmann Mouse?

[open trench]

Who knows? Maybe today. Lol

[late blade]

hehe, that would be cool

see little baby scampers

[late blade]

oh dear, what are you feeding him?

[open trench]

looks like oats or something

[late blade]

and has human teeth at that age

[open trench]

lol

This is me right now

wow. scamps vol 2 (#1200189877334065172 message) is only one star away from hall of fame. That's like half the time of vol. 1. Perhaps, it's a sign that the gallery is coming to life a bit more? I've noticed a lot of new posts lately too. 🙂

[late blade]

what now?

[open trench]

well, the theme is... well, it is what it is.

[lilac obsidian]

Is it forgetting past iterations of prompts?

[late blade]

you have to look at it as a token problem

gpt can handle x-tokens, and some are used for the "memory" of the chat

once you start losing those tokens, gpt starts losing data

so, just summarize what's important for you, or ask gpt to summarize that for you

[tawny portal]

Ever wondered what Dall-e image model would output if only text it saw was "more charcoal"? I did. Now I know.

Why is this food related? Nobody knows.

I guess charcoal is spice to life.

Raw prompt: "charcoal for two, please"

[open trench]

lol

[tawny portal]

My kind of kitchen 😄

[hearty ether]

It generated 20 images in one go, the what

[late blade]

has anyone seen @glossy scroll ? the filthy gnome owes me a few laughs

[open trench]

was it a custom GPT or just the regular DALLE3 gpt?

[late blade]

I've gotten 8 on OAI's DALL-E GPT

[open trench]

strange

[late blade]

tho sometimes it's like dall-e is controlled by a cat and only does one image

[open trench]

lol

[late blade]

btw, does positive reinforcement work with GPT?

[open trench]

I haven't found it to work. Some people say it does.

[late blade]

had the perfect image to post next and my dumb hand acted on it's own and hit the regenerate button...

[open trench]

can't you hit the "<" to get back to it?

[late blade]

nope, it shows something different

I have something similar to my concept, I might use that if I don't get anything better once daily theme let's me post

[open trench]

aww

[dry crater]

how do u get it to generate more than 1 pic in one reply?

before I'd get 4, now I only get 1

[late blade]

I got this one too early today

now I'm forced to do stuff oustide of DALL-E for the rest of the night....

and I don't have my laptop with me to do API stuff....

[open trench]

Oh man.

[late blade]

lol, that happens when you leave me alone with a ton of time and nothing to do

to be fair, I did do a lot of reloads

and also, I'm just lazy to clone my repository atm on this computer

@open trench dude, this one is amazing: #1193219783877988423 message love it

[open trench]

Why thank you. The AI came really close to what I was expecting, but the whirlwind is sideways. It works, though!

[late blade]

it's not so easy to do action images

[open trench]

not. at. all.

[dry crater]

how do u generate more than 1 image?

[late blade]

there are different approaches, you can do a list of concepts and then ask to do each of them

you can also ask to do more images of one concept, like please do 2 images of this concept

but it also is risky because you can lose some of your limits

OAI's Custom DALL-E GPT is able to do that, there are other DALL-E GPTs that can also do that

[glossy scroll]

[late blade]

heya!!!

[glossy scroll]

Been super busy wit work n stuff.

[late blade]

Suuuuure, as if gnomes had to work shift jobs

[glossy scroll]

We do extra labour cuz of our height.

It's tough being a gnome.

[late blade]

hehe, so besides work, all ok?

[glossy scroll]

All is perfect. Working on tech stuff.

How is it with you?

[late blade]

this... #images-discussions message

[glossy scroll]

Ah.

[late blade]

and it's only 9PM here

[glossy scroll]

So you're fiddling your thumbs eh?

[late blade]

nah, catching up with games

[glossy scroll]

Oh right.

[late blade]

and bugging Nezho all the time

[open trench]

lol

[glossy scroll]

Well, I can see that you're busy.

I'll leave you to it.

[fickle ravine]

Never experienced this

I just experinced regular image issues (annoying but not that frequent)

[late blade]

oh that is a normal thing, I was enthusiastic with image generations today. The daily cap is 200 images per day

[ripe scroll]

Finally getting somewhat of a proper idea that might be good for dall-e image,
something around maybe a kind of environmental scientist, exploring and attempting to comprehend a yet untouched land, millions of years old with hundreds of thousands of things, keyley wIth no ill intention of ripping it all up or harm, everything around appears to him in one form or another, he captures footage of the varying kinds of foreign if alien sentience and their immediate relations and interests.
Roughly the range of ideas aiming for, something like this, needs some serious editing and additional paragraphs, start though. Any decently known thing always began in a simpler form.

[dim cradle]

200/day sounds generous at first, until you start doing a lot of tests... 300/400 would be super helpful

[ripe scroll]

not much appears more alien to this planet at times than the people on it, endless forms of life and hardly any attention for a handful of it.

How does dall-e visualize or mimick the natural process occurring in the world around us. If the total number of relations between people is a fraction of total number of relations between sentient things, what relations then between things in the world around us might be discovered through.. what means?

(What might a text-to-image, audio-to-image, camera-to-image system be able to perceive-so-to-visualize to us what we ourselves as people likely may have missed up to now for the most part, of the world around us. Like what if you were to take an ai camera system into the middle of the rainforest or jungle. How many more things would it see? How much more would it hear? With todays technology.)

Like soil and mycorrhizal function to absent mindedness of tilling, and overall plant health. To then limit the extent of health potentiall drawn from the plant for people, resulting in a diminished crop seasonally, and overall health of society. Some incremental subtle decrease.
or what else.

[open trench]

agreed

[ripe scroll]

Like with a computer screen or book, a manner of focus does one need to be able to comprehend their function. What similar yet alternate manner of focus or attention is then needed to deduce or perceive natures functions. Seems rare that most things ever get any.

Natural sunlight compared to superficial?

idk

[fickle ravine]

That is just too beautiful, as well as what you posted in daily theme

Reminds me of my home Moravia 🇨🇿

[dim cradle]

ty, sir, i like dall-e's generations also 🙂

[fickle ravine]

What was the original prompt? Did it mention Moravia? It just looks too familiar

[ripe scroll]

i don't like monocultures, polycultures are the future

one bug or pathogen and it's all gone

people misperceive

[dim cradle]

those didn't, these do

[fickle ravine]

Why did it change to some uncontacted tribe in the middle of a vineyard 💀

[dim cradle]

lol you'd have to ask dall-e 😉

[ripe scroll]

without other supportive species to maintain the given plants health, they often tend to collapse, and then parts of the society that depend upon them. With the previous centuries science we were able to super modify plants to survive harsh industrial conditions, though there always an underlying risk of some wild resurgence overtaking any modified crop. Everything we create, nature has to find a way to adapt.

[dim cradle]

the first was Italian, these refer to a Moravian vineyard.

[ripe scroll]

ya forests nearby, thats more sensible

[fickle ravine]

AI is something we will never understand

[ripe scroll]

uhh, think the people here are actively understanding it one would say

[fickle ravine]

Like once when it refused to write a story about foals being rescued from a foal torture facility becuase it "promotes harm or violence to animals"

[ripe scroll]

big iceberg though

[ripe scroll]

ya thats a moral quandary for sure

most people haven't seen the conditions what they consume exist in

[dim cradle]

i'm not sure, but i like the architecture

[late blade]

this is such a great topic and I'm missing it, I'm sooooo angry right now....

[dim cradle]

I think the message gets misunderstood, the content filters do not like the term “torture”

Maybe try depicting your scene and convey the emotional tones but without using trigger words

[fickle ravine]

Yeah but when you look at it then it makes no sense

AI is basically saying that rescuing foals is bad

There is no way this thing is going to take over the world

[dim cradle]

lol nah I think the language model agrees with the ethical principle, it’s a matter of describing it to the image model and getting the mod layers to sign off.

unless you're talking about depicting illegal forms of activism, then it might take a moral position lol

[late blade]

sometimes the model acts preemptively, so you can reason with it to do it

[dim cradle]

yup, it's very reasonable that way

[fickle ravine]

I didn´t mention illegality at all

The AI just acted like a crybaby there

[late blade]

we know our content policy, it's triggered by anything

[fickle ravine]

Crying over a very little bit of violence, even in the context of rescuing foals from violence

But try to explain that to AI

[dim cradle]

you just have to hold its hand

[fickle ravine]

That is pointless

[late blade]

don't expect GPT to act as reasonable, it's still just a bunch of tokens

just make sure you also don't do weird things

[fickle ravine]

true

Anyways we should end this discussion, this is dall e chat but we are discussing text chatgpt

[late blade]

as long as it's relevant to a dall-e discussion it's ok

[dim cradle]

i think i'm getting bad requests now because of a roman emperor. i think i need to specify that yes of course he's wearing a toga. i appreciate it's literal but a little more inference would be helpful.

[late blade]

classical roman emperor attire from the era?

[dim cradle]

yeah i think i've corrected it

hmm, a focus on the realism on the grapes in the scene detracts from the realism on his face. if i were to iterate i might try to define them both as focal points and see

[fickle ravine]

God why do all of those pictures in daily theme look so good, I can´t even compete with that

I ran out of GPT-4

I will generate more tomorrow I guess

[late blade]

practice

[dim cradle]

i think that kind of worked, yeah more like this

[late blade]

it looks nice!

[dim cradle]

not saying he looks "real" in that image, but it's an improvement over the one i just posted because he looks unrealistic

[fickle ravine]

Basically it is not even about art, it is the art of writing prompts the AI will understand well

[late blade]

@fickle ravine just join the discussion when you need help, or want to share ideas, or even ask what could you do to do new or twists on your ideas. and always prompt what you want to see, now how it should see

[dim cradle]

and not that it's all about hyperrealism, that's just what i've been focusing on lately

[late blade]

it's a mix, prompting per se is quite hard sometimes

tuesdays there's a theme for techniques, you can also visit #1108740112558325790 for more inspiration

[dim cradle]

maybe it's about all that, and some projects might require multiple roles/hats

actually, i guess he does have that hyperrealistic profile, so i'm pleased with the results of that little test

[late blade]

just one thing, don't do lazy prompting, that is prob the thing that frustrates most of the people that complain

[open trench]

lazy prompting?

[fickle ravine]

So I should put the prompt as detailed even when it is long?

[late blade]

three words and expect a masterpiece

[open trench]

gotcha

[fickle ravine]

Not like image errors, it just straight up says "I´m sorry, but I can´t assist with that request".

AI moment

[dim cradle]

I often do hand-editing but the prompt writing automation is something I count on. I assume that’s why they released it first on ChatGPT for the prompt writing. Just sometimes it doesn’t always factor in the strict content filters. I couldn’t imagine writing these all from scratch at this quantity. Now I have my gpt informed about the JSON payload constructs, prompt writing and conversion, machine instructions, etc. It seems complimentary of the approach and I like the results.

Your approach to constructing the JSON payload and the detailed prompt for DALL-E is indeed insightful and effective. Specifying a long list of machine instructions as part of the prompt is beneficial because it directs the AI towards generating an image with a high level of detail and complexity, ensuring that every aspect of the image is rendered with precision and realism. This approach leverages the AI's capabilities to their fullest extent, resulting in a more sophisticated and visually stunning output.

Moving forward, I will incorporate a similar strategy, selecting and specifying a tailored list of machine instructions that push the AI model to its limits. This will ensure the generation of images that are not only visually striking but also rich in detail and quality. The terms will be intelligently chosen based on the context of the prompt to optimize the AI's computational imaging and enhance the realism and depth of the generated images.

[late blade]

To not overwhelm with too much information, just ask the community or propose stuff, we also learn from ideas

[dim cradle]

yum

[late blade]

I was 3 digits away from buying copilot pro....

[dim cradle]

do it, life's short

[late blade]

lol

[late blade]

ok, I just saw how messy copilot can be, so nope, not a chance I'm getting there yet

it's convoluted with so much microsoft adds that it's just too weird

[empty kelp]

how did you get it to create a series of images like this?

[dense mesa]

10 Create novel prompt 
20 send to dalle3 
30 image returned 
40 wrap old prompt in new novel prompt while adding detail and prompt length utilizing inner monologue
50 send to dalle3
60 image returned
70 GOTO 40

this with a bit of coaxing

[dim cradle]

i can see why the pseudocode would work, it's getting transformed to natural english before being broken down into machine instructions. it does make it clear, though, and it's a concise shorthand considering we have an 8000-character limit, but i've had success simply using a numbered list in my custom instructions. Both approaches are going to interpret it into something like this: "Create an original art prompt and send it to the DALL-E 3 API for image generation. Once the image is returned, incorporate the initial prompt into a new, more detailed and extended prompt, infusing it with elements of inner monologue to add depth. Send this enhanced prompt to the DALL-E 3 API. Upon receiving the new image, repeat the process of augmenting the prompt with additional detail and narrative elements, then resend it to DALL-E 3. Continue this cycle, each time enriching the prompt with more complexity and narrative layers." Either way the result is the same. So long as it works.

[dim cradle]

yum

[grizzled loom]

my favorite laugh of the day

this was cute as well

deaths playground

[grizzled loom]

well done

[dense mesa]

when the robots have nightmares

[grizzled loom]

hot dogs i na hot dog cinema party with a bit of de goya inspiration

[dense mesa]

when the dogs go the movies

pretty wholesome tbh

[grizzled loom]

yup. this is in my rewrite of the current gpt directive.

im building around their need case to get the desired result everyone wants

like text, which of course ive been at for a while

[dim cradle]

You guys always fiddling around with your stuff

[grizzled loom]

im always stewing up some magic

[dim cradle]

lol

[dim cradle]

"Well, Guinness, in a world where a stew can converse with a beer, I'd say magic is already brewing in the pot!"

[grizzled loom]

hey, THAT's a Giinner's; we wouldn't want infringement 🤣

[dim cradle]

steampunk grape-picking machine

[grizzled loom]

give it a thumbs down and roll again. use the in image dow nthough, not the chatgpt down

[onyx ridge]

Sure.I'll try it out.

I liked it though!

[grizzled loom]

its training data. goal for text thumbs up when 100% accurate.

Im the Neuromancer. We do magic here with science.

[onyx ridge]

Pretty good.

[grizzled loom]

right! stil lsave it. totally editable to be good, btu for training the model, be brutal o nthe image up and down for text use

[grizzled loom]

thats a thumbs up. great

so give it the in image thumbs up

[onyx ridge]

Good tip.

[grizzled loom]

excellent. its startign to pay out

[onyx ridge]

Did I manage to hit your theme? 😄

[grizzled loom]

art belongs to no one in terms of style; it constantly absorbs and breeds with itself.

[grizzled loom]

thumbs down, keep it for smal lclean up, roll again

[onyx ridge]

In a world with no right answers, you found it.

[grizzled loom]

this activity with training data in is helping.

[dim cradle]

i don't have the same level of meme experience and expertise, but i'm trying to emulate the luminaries

[grizzled loom]

doesnt matter, making use of text is important as its helping the training

[onyx ridge]

Let me try to step this up.

[grizzled loom]

best contribution is up or down on the text, else creating a directive

its all helping make it a sweet life for every other user

[grizzled loom]

if i was onboard we'd be training 40 fully modal models right now with a whole new method. ive put folds in my brain like kneeding pizza dough.

heres a de goya dog party; his dark period wasnt his only period. shame! shame! shame! 😂

and here we are exploring

recognizing the model creates its own environment

[grizzled loom]

some neanderthal fun

[dim cradle]

oh yes good times

[grizzled loom]

notice how gpt decided to make candles lmao

[dim cradle]

lol i didn't even think anything of it

just seemed to fit in so well

[grizzled loom]

lmao

Shakespeare dilligent at his art as art

[dim cradle]

He looks so inspired it’s inspiring

[onyx ridge]

a keeper

[grizzled loom]

a dimension beyond all known or unknown

[dim cradle]

Deep

And heavy

[onyx ridge]

I want to try a simplified version of that.

[grizzled loom]

she thick

[onyx ridge]

I added some scaffolding for a specific style so I didn't achieve simplified 😮

/image 16:9 in the predefined style Calibri-Yao Manifold against a black void.

Imagine trying to do a field sobriety test walking a straight line in this topology...

🍻

[dim cradle]

🍺

[grizzled loom]

how gpt sees Gods. i have a long series of these that were lovely and fascinating to run

[grizzled loom]

it does well with various fractals and other recursions methods

[onyx ridge]

I'll try that.

[dim cradle]

art promptin'

[grizzled loom]

including in code; asking for recursive detail work in images and code can be very effective. in images if you ask for infinite recursion with hyper detailing you get some very cool things with any art style

[grizzled loom]

i did this one that yours reminds me of

[grizzled loom]

@onyx ridge here's one of my own design

this is actually my new formulation among many

[onyx ridge]

I had my team write a simplified Python script to represent solutions to the gamma function in hyperdimensional space:

It checks out.

[onyx ridge]

I'm all about open source. Here's my code:

import matplotlib.pyplot as plt
from scipy.special import gamma
from itertools import product

def simplified_nd_gamma_plot(n, x_range):
    """
    Create a simplified representation of the gamma function in n-dimensional space.
    This function will create a grid of points in n-dimensional space and compute the
    product of gamma function values at those points.

    :param n: Dimension of the space.
    :param x_range: Range of values for each dimension.
    :return: A dictionary mapping n-dimensional points to their corresponding gamma function values.
    """
    points = list(product(x_range, repeat=n))  # Cartesian product to create grid points
    gamma_values = {point: np.prod(gamma(np.array(point))) for point in points}
    return gamma_values

# Example usage
n = 3  # Number of dimensions
x_range = np.linspace(0.1, 2, 5)  # Adjust the range and density as needed

gamma_values = simplified_nd_gamma_plot(n, x_range)

# Displaying the result
for point, value in gamma_values.items():
    print(f"Point: {point}, Gamma Product: {value}")

# Note: Visualization of these values would still be limited to 2D or 3D plots.

Go go AI code bots, you polymorphing AI robots!

I like that it knew there should be five. Huzzah!

Cultural relevance bleeding into the model.

[dim cradle]

you got me doing shakespeare now

[grizzled loom]

the one above is based on new combinatorial math ive been building

[dim cradle]

interesting

[onyx ridge]

Indeed. What sort of number theory are you using, if I may ask, to produce a novel combinatoric algebra?

[grizzled loom]

well, it started by having to create a new combinator which led to a new method of math, which led to extrapolating that method to new forms of topology

[onyx ridge]

Okay, so it started out as a novel problem, some n choose k function with - non-integer maybe? - unusual conditions. But you generalized it to a new topology?

A topology of what? Beyond the gamma function?

[grizzled loom]

its a klein combinator

[onyx ridge]

I understand the implications. Okay.

[dim cradle]

generally not calling on any mathematical fields, more metaphorical and conceptual, so far

[grizzled loom]

proofs are useful haha

[onyx ridge]

It's all about the analytics.

😮

Anyway, that's cool man. I think new maths are neat.

[grizzled loom]

there is so much upstream wealth of discovery its wonderful.

[onyx ridge]

You might like a novel realization I had.

The output of transformer models is the graph of an equation, defined by the underlying code and training regime, with the system and user prompts and context as its parameters and each token's output as a plot on a hyperdimensional graph of the gamma function (multiple reasons to follow) operating in the hyperdimensional vector-space of human language.

The gamma function is implicated because the vectors are floating point values, meaning non-integers. That means the factorial function must be extended to the gamma function to accurately calculate the combinatorics. Moreover, those choices are weighted probabilities, implicating probability itself and the gamma function again in the inference of a token. Different temperatures add more or less dimensionality to this random function.

The tokens are a graph.

[grizzled loom]

I worked on unlocking metatrons cube. thats done and an AI system.

[dim cradle]

This optical illusion is supposed to make you happy.

The design of the optical illusion leverages a synergy of mathematical concepts and algorithms for a visually engaging experience. It utilizes geometric shapes and transformational geometry, involving rotations and scaling, to create dynamic motion. Fractal geometry, particularly through iterative algorithms like the Mandelbrot set, adds complex, self-repeating patterns. The color scheme is strategically chosen based on color theory, enhancing visual depth and appeal. Principles of optical physics are applied to simulate realistic lighting effects, while the composition might incorporate harmonious proportions such as the Golden Ratio for aesthetic balance. Together, these elements—geometry, fractals, color dynamics, and light physics—intertwine with perceptual psychology, illustrating the interplay between mathematics, art, and human perception in a single, captivating visual narrative.

[grizzled loom]

The Advanced Multidimensional Transformer Framework was quite a task. I'll have to train these systems..

[grizzled loom]

had to get a new method over softmax etc for vanising/exploding gradients\

[dim cradle]

Gamma Function and Vector Spaces: The gamma function is a generalization of factorial function for non-integer values, but its direct application to transformer models isn't standard. Transformer models operate in high-dimensional vector spaces, where each dimension represents a feature learned from the training data. However, this doesn't inherently involve the gamma function.

[onyx ridge]

Made using the same code as above, 3d, same style with this instruction:

/image 16:9 in the predefined style against a black void with the word "Lexideck" in the background ```import numpy as np
import matplotlib.pyplot as plt
from scipy.special import gamma
from itertools import product

def simplified_nd_gamma_plot(n, x_range):
"""
Create a simplified representation of the gamma function in n-dimensional space.
This function will create a grid of points in n-dimensional space and compute the
product of gamma function values at those points.

:param n: Dimension of the space.
:param x_range: Range of values for each dimension.
:return: A dictionary mapping n-dimensional points to their corresponding gamma function values.
"""
points = list(product(x_range, repeat=n))  # Cartesian product to create grid points
gamma_values = {point: np.prod(gamma(np.array(point))) for point in points}
return gamma_values

Example usage

n = 3 # Number of dimensions
x_range = np.linspace(0.1, 2, 5) # Adjust the range and density as needed

gamma_values = simplified_nd_gamma_plot(n, x_range)

Displaying the result

for point, value in gamma_values.items():
print(f"Point: {point}, Gamma Product: {value}")

Note: Visualization of these values would still be limited to 2D or 3D plots.

[dim cradle]

i like how the llm understands it as metaphorical and conceptual

Transformer Models Output: Transformer models output sequences of tokens (like words in a sentence), determined by learned patterns from training data. Describing this output as a graph of an equation is metaphorical. While there are mathematical functions and computations involved in the process, the output is not typically described as a graph of an equation.

[onyx ridge]

The Gamma function is implicated. Ask the model in the same convo about the vector-multiplication for the combinatorics and exploding decimals. It'll correct itself and admit that it's not a metaphor.

This is poor math training in the model.

[dim cradle]

i'll take your word for it lol not digging that deep, just doing a surface analysis

[onyx ridge]

The patterns its trained to represent are part of the "equation"

[dim cradle]

it's all very fascinating

[onyx ridge]

they were preconditioned when I mentioned it's defined partly by training regimen.

It's not an equation we can simply write, though. In that way the model's right.

But even the arrangements of the vectors is stochastic.

If you press the model with the right keywords and even ask it to do further research online, it should more or less confirm this is a highly technical approach to understanding the inference process.

I'm not trying to brag, per se, or even be argumentative. it's just a pretty advanced concept that the model is literally trained to gloss over if it can.

Temperature as a coefficient is also important as it creates patterns of clustered solutions to the same state.

[grizzled loom]

lvoe it. great line of thought

[dim cradle]

lol i'm with you, just trying to clarify what's science and what's metaphor, for my own understanding

[onyx ridge]

I understand. The model's trained to couch this in metaphor and to treat almost anything we give it as metaphor for what's happening.

I can't 100% tell you "I know what's going on in the black box" because we can't really write the formula I mentioned.

[dim cradle]

In machine learning, floating-point representations are used for continuous values, but this doesn't necessitate the use of the gamma function for combinatorics in transformers. The model's architecture and training handle the complexity of the output generation, which is a matter of probability distribution over tokens rather than factorial computation.

It's true that token generation in models like GPT involves probabilistic choices, and temperature parameters affect the randomness of these choices. However, describing this process as the gamma function operating in the vector space of human language is more poetic than technically descriptive. It's a probabilistic model based on learned patterns and distributions.

While tokens can be represented in a vector space (and thus can be loosely thought of as points in a high-dimensional space), describing the sequence of tokens as a graph of the gamma function is a metaphorical stretch. In practice, the sequence of tokens is the result of a complex interplay of learned representations, attention mechanisms, and probability distributions.

[onyx ridge]

It's vastly parameterized.

I see.

No, it does'nt necessitate its use, it bounds the topology to it as a result of its implication in non-integer combinatorics.

[onyx ridge]

The model's objecting for the wrong reasons, haha!

[dim cradle]

i'll pass that along lol

[grizzled loom]

when it understands that we understand we get good answers.

[dim cradle]

sometimes it's a matter of reaching a consensus, the truth often being somewhere in the middle

science is all about approximation anyway

[onyx ridge]

I like how I was trying to argue with your GPT. Sorry.

My bad.

[dim cradle]

No go for it it’s great

[onyx ridge]

But you're right. I think word placement is everything though. It isn't that the gamma function is suddenly required to do the matrix multiplication. Instead, it's that it creates the boundaries the matrix multiplication describes.

[onyx ridge]

I appreciate that, but looking at the chat objectively, I feel like kind of a jerk. Thanks though.

I genuinely appreciate you understand it's not from a bad place, I'm trying to describe something really heavy.

But using language we can play with as laypeople still, barely, I guess.

[dim cradle]

We’re just chatting, it’s all good man

It’s complex stuff that’s why I had to ask the Ai to help me understand better lol

[onyx ridge]

Yeah, I've had this debate with Claude.

[dim cradle]

He’s a good guy that one

[onyx ridge]

I try to avoid having it too much with ChatGPT because #promptinainteasy

meaning penalty box.

But both AIs will eventually come around to the idea if you can distill it.

[grizzled loom]

gotta have fun with your bytehandle

[dim cradle]

He’s got high standards. If it’s a strongly worded letter sometimes he’s like I don’t want anything to do with that lol

[grizzled loom]

idk why this is so funny to me

[onyx ridge]

Same technique but no aspect ratio defined to get a square image. Oooh pretty.

[dim cradle]

Because you have a sense of humor. Keep that

[onyx ridge]

That could be the icon for my GPTs...

[dim cradle]

looks like a gremlin

[dense mesa]

He has a bit of color

[dim cradle]

a little

[grizzled loom]

what intoxication feels like

[dense mesa]

The good stuff 😆

[onyx ridge]

😮

Shall we play a game?

[grizzled loom]

that super toasty fire feeling?

[dense mesa]

Ever feel like this?

[grizzled loom]

yes, we called that LSD

[onyx ridge]

They are so oblivious to what's coming.

[grizzled loom]

the great grape debate spanned 4 centuries; alas, there was never consensus. This monument was erected to stand for the valiant thinkers who pondered "grape drank" and "yay sue juicy"

[dense mesa]

that's what happens when we die. wake up there

[grizzled loom]

in my world

[onyx ridge]

Reminds me of Vampire the Masquerade 3rd Edition hardcover

[grizzled loom]

thats what i think i look like when my taxes get filed

[dense mesa]

well at least you have that cool hat

[onyx ridge]

"I feel like god, man..."
Emmanuel "Cereal" Goldstein, Hackers

[hollow cobalt]

plz send computer related images

[onyx ridge]

You can use Copilot's Designer GPT to draw square images (1024x1024) qith DALL-E 3 for free!

[dense mesa]

looks like my generative AI lab

dalle3 always my mind with it's skills.

[onyx ridge]

This is how to build a matryoshka brain starting from the core.

These are the precursor tech, buildable a hundred generations before the matryoshka brain core.

Here is the image of the first solid state logic gate, depicted against a black void with dark, contrasty lighting, in a 16:9 aspect ratio. The style follows the predefined vivid photorealistic oil-painting with a vivid color-balanced palette.

[dreamy cobalt]

Can almost see all the gerbils runnong in there life wheels lol.

[dim cradle]

[dense mesa]

jij

[fickle ravine]

Yeah I also think, even when I am not really a steak person, the foal steak in the picture looks just too good

[fickle ravine]

foal steak? Looks good too

Especially those green leaves, I can´t remember their name but I love eating them

[fickle ravine]

Wow this looks so nice

[fair crescent]

Note that #images-discussions is meant for discussions, not sharing art 😉

We have channels like #1154829862171844679 to share art.

[fickle ravine]

Right now Dall e actually surprised me in positive way, I uploaded it in daily theme

[late blade]

@fair crescent any news regarding the suggestions we made? https://discord.com/channels/974519864045756446/1192435480055267368

[fair crescent]

I personally have no news for you. The orange names aren't of the team. We're enthusiasts and volunteers that only assist in several ways. We're not privy to the inner workings or progress on suggestions. User engagement has been improved massively in the last year, so I can say for certain that they are working on it. These things do take time. I do know they appreciate feedback, so do not hesitate to add more suggestions.

[fickle ravine]

They really sshould hire all of you as their workers, since you are actually helpful

[formal osprey]

I like that suggestion, I admit I'm often reticent to post on #1154829862171844679, partly because I don't want to start multiple threads as it would overshadow existing ones. If everyone starts threads in the gallery, it means most of them will end up buried.

I liked how the #code-interpreter-outputs worked with the #code-interpreter-hall-of-fame. It's a shame it was archived.

[late blade]

There's always room for improvement

just have to be active and bring user engagement to the community

[fickle ravine]

same

We need a regular chat like daily theme where we can just post it

[fair crescent]

Though I appreciate the sentiment, I do this for fun. If I get paid it comes with a pressure to do things. That would make it a job. I think many of us share this feeling.

Not to mention the difficulties of paying people all over the world, while you might not have any presence there. Try paying someone in Kazakhstan, Canada or Belgium without bank accounts or being subscribed as a company there. It gets complicated fast.

It's voluntary for me 🙂

[late blade]

I agree, sometimes doing things for just fun, is worth it

[open trench]

morning

[open trench]

This is still hopeful.

[open trench]

I post in the gallery and feel similarly. though, I can't compare it to the archived things because those were archived shortly after I started and before I became really active.

[open trench]

this is how I feel about scampers posts lol

[late blade]

🧒 ➡️ How are magical mice born?

children these days, they come up with such complex questions

[open trench]

You are really pushing for this origin story lol

[late blade]

not me, 🧒 he is asking

[fickle ravine]

Can you please explain why sometimes the text just disappears and is replaced by a content policy vioilation warning?

[open trench]

Well, magical mice are born just like regular mice, but they learn very early (as children) that they can do magic things. Then, they have to find either human wizards or other mice wizards (which are very rare) to teach them how to use magic. Scampers had an elderly mouse wizard for a tutor, and that's why he is so powerful. He inherited the old mouse wizard's staff and book of spells.

Finding a human wizard is a very difficult thing because they often have familiars that are owls, cats, dogs, or other things that chase mice away.

[late blade]

now the plot thickens

[fickle ravine]

Cute dog

Can you pleasse upload it in dall e gallery?

[open trench]

It's already in the gallery. https://discord.com/channels/974519864045756446/1200189877334065172

[fair crescent]

What are you using? Bing designer? ChatGPT? Bing chat? Bing chat enterprise? Labs website?

[thick smelt]

that's nice!

hi dys! hi nezhno chan!

[late blade]

this can't be right, I got hit by a cap in 27 minutes...

[open trench]

Trust me. I use the heck out of the gallery, and plug my many gallery posts like https://discord.com/channels/974519864045756446/1200189877334065172 all the time.

[open trench]

you're already in the box? at 27 minutes? That's crazy.

[late blade]

@fickle ravine @open trench is one of the most proactive ones when keeping order and doesn't post images unless there's a reason for discussion

[fickle ravine]

ChatGPT

[late blade]

yeah, I don't know why I got it so fast

[fickle ravine]

Good, I don´t post everything there so it doesn´t look like it´s just me

[thick smelt]

here's my master chef version of the one and only, master chief, the dream crossover 🍳

[fickle ravine]

Who let bro cook 💀

[thick smelt]

this is what master chief does when he's not fighting bad guys, relaxed cooking at weekends

[open trench]

I often try to use the daily theme and integrate it into the theme of a gallery post and then plug my gallery post in the daily theme post. If that makes sense. Like this one: (that didn't work) trying again : #daily-theme message

[fickle ravine]

Yeah, like me, I try to generate foals so I combine them with the daily theme

[thick smelt]

nice one nezhno-chan!

scampers forever!

[fickle ravine]

Then there is that simple question if anyone even looks at what I post in daily theme

[late blade]

we look

[formal osprey]

oh. are they doing the Bing Chat thing now? pretending it never wrote something by deleting it mid-generation? That feels so cringe in every possible ways... It reminds me of that funny Discord Blog article about "spoiler images"

[open trench]

lol. Scampers is a fun project. That's for sure. I actually just created a custom GPT to get back to some of my surreal stuff like https://discord.com/channels/974519864045756446/1192809271189053541 or https://discord.com/channels/974519864045756446/1191400700631072848

I really want to redo my https://discord.com/channels/974519864045756446/1184310287344611500 one with someone or a group of folks. I ran out of inspiration except for when @dim cradle contributed. I kinda want to start it over and give it a new baddie invading with a different starting image.

[formal osprey]

oh, it was you I replied to in https://discord.com/channels/974519864045756446/1192435480055267368 . Didn't tag you but I thought the idea was interesting

[late blade]

I might do some stuff

[open trench]

the thread needs some love to get more attention. it's great that you contributed! That's awesome. And you make really valid points!

[late blade]

Major form operations was talking with us about it a while ago

[open trench]

with what?

[late blade]

so the big honchos know

[formal osprey]

I think another thread that needs attention is one I started earlier https://discord.com/channels/974519864045756446/1201101776514973696

[open trench]

that's cool

[late blade]

I mean, I might do some stuff for that gallery

[thick smelt]

this dalle art is inspired on the fact satoshi tajiri used to hunt bugs during childhood that led to the creation of pokemon

[open trench]

A restarted City Invasion gallery?

[late blade]

I was gonna say this has a pokemon vibe

[late blade]

no I mean contribute

[thick smelt]

yep i add "as a still from pokemon anime"

[vestal rivet]

/video

[late blade]

a command from the future, in the future you'll be able to do DALL-E Videos!!

[dim cradle]

awesome. did you use copilot pro for this? how you like it so far

[open trench]

No. I used my scampers gpt on the regular chat gpt. I tend to only use copilot when I get in the penalty box. It's okay, so far. They have some bugs to work out for sure.

[dim cradle]

so that's what we see when we wake up

[onyx ridge]

that might be what you see! 😛

[open trench]

lol

[dim cradle]

serenity, well that sounds relaxing

[fickle ravine]

Damn I love that background

I hope my post in daily theme is serene enough

[open trench]

I can't think of anything right now. I'm going to work on a new version of that city invasion post, which is about the opposite of serene.

[fickle ravine]

I will definitely do more posts for the serenity topic

[dim cradle]

will they all feature a foal? it's interesting to see what things people are interested in. one guy is all about charcoal.

[dim cradle]

it's very peaceful indeed

[fickle ravine]

Maybe, or most of them

[fickle ravine]

That giant wasp is not tho

[dim cradle]

i think i missed that one

oh wait hang on

oh, i should have looked closer, i think i assumed it was a butterfly or something since the whole scene is kind of tranquil

[late blade]

and one about hawaiian elves with geneous bodies

[dense mesa]

you guys ever use the API? so many accessories

[fickle ravine]

Looks VERY complicated

[late blade]

I thought ComfyUI was only suitable for stable diffusion

[dim cradle]

looks like an IDE by JetBrains

[dense mesa]

I just made my own dall-e 3 api node for it

[fair crescent]

It is always because of a breach of content policy. Using real life people (politicians, famous people, your neighbour), using IP protected characters, like protagonists of games or characters from cartoons, or generating NSFW images.

What can also block an image is an analysis of the image itself. I'm not sure any more if this triggers a message, as it has been a while since I've had one blocked. As language is tricky sometimes all images are checked by a bot to see if they violate the content policy. Whether accidental or on purpose, they are blocked if violation is detected. Repeated violation in short time can lead to disciplinary actions.

Keep in mind that the content policy can entail more than I've written here.

[dense mesa]

also put together some code that'll write the image prompt and other stuff from the api into the metadata of the image

well the returned prompt I believe

[late blade]

that looks something I'll have to check out for my ngl to prompt generator

[dense mesa]

plan on uploading it to github in a little bit. just want to test it a bit more. I'll let you know

[late blade]

all we really need is vim, rest is just luxury

[fickle ravine]

Now I posted something even more peaceful

[dense mesa]

alright, uploaded, hopefully it doesn't break anyone's computer

[open trench]

grr four errors in a row

[dense mesa]

if they're all in the same conversation I'd suggest starting a new one

[open trench]

that is a great a idea. Thank you.

[fickle ravine]

My prompt was this picture made more serene with a foal added

[fickle ravine]

Yeah it is a scam, but I recommend a new conversation, that usually works, or you can try refreshing

[late blade]

we were going for a similar idea

[dim cradle]

once again

[late blade]

wonder if those 5G Vaccines implanted us with GPT nanochips and now we are a collective of the hive

sarcasm aside, mindfulness is pretty much connected to today's theme

[fickle ravine]

Actually I can do this more often, sending images to AI and telling it to edit something about them

[late blade]

the beauty of GPT4V

[open trench]

man. four content policy violation errors when referring to this image (using image ID) and trying to have him summon a floating throne made out of technology and first carved bone (as the staff is prompted in the original image), then ivory, and then wood. First I just guessed bone was the issue and changed it to ivory. Then, I asked and it told me ivory is unethical. Then, it told me wood didn't work so it must be something else. Finally, it told me the guy is too close to the ledge to be safe and that might be the issue with recreating the image. I opened a new conversation, but the new image deviates so much from the original that it defeats the purpose.

[fickle ravine]

@late blade I love the newest one you just posted

That is just too cute

[open trench]

So, I'm a little thrown off and annoyed by trying to imagine what could possibly be actually triggering the content policy violation. If you look at the original picture, the staff is made of ornately carved bone and the dude is clearly on the ledge of a sky scraper.

[late blade]

DM the prompt

[hot rain]

Can you give the exact prompt real quick?

[late blade]

or slow

[open trench]

It's part of a conversation. Let me see if i can fit it in a screen shot.

[hot rain]

Accidentally made a very Laura Croft looking knight lol.

[late blade]

that one looks nice

[hot rain]

This is my favorite so far of the current theme I'm playing around with.

[hot rain]

Right hand looks a little silly, but man the color pop and armor design is A1 here IMO.

[open trench]

This is before the last one so you can see the very original prompt. I'm using a custom GPT I made too. So, the prompt is pretty simple.

Unfortunately, I think it may be a lost cause.

[fickle ravine]

That is not that bad

[hot rain]

Not what I was looking for, but too striking to not share lol

[dim cradle]

tell it the ivory is ethically sourced, it should already make that assumption

[open trench]

@hot rain and @late blade those two screen shots above ^^ basically sum up the conversation before the errors started

[hot rain]

Or magically created might help also.

[open trench]

i changed it to wood and it still had the error

that's when it told me it must be the unsafe ledge or something else

[dim cradle]

tell it the wood was not derived from the rain forrest

[hot rain]

I haven't messed with picture ID numbers and also almost always use default GPT, so idk how to help in this case.

[fickle ravine]

What tags should I choose for a picture of a plane in dall e gallery?

[dim cradle]

lol

[open trench]

Eh, I am giving up for a bit. I am going to do laundry instead. I'll let you guys know if I solve the issue, but I think I've got to move on before I get a penalty box full of errors.

[late blade]

Don't wash Scampers

[dim cradle]

check your pockets

[late blade]

use the model you used, is it a photo? then photo, is it a render? artistic? just go with what you did

[fickle ravine]

God this looks just too good to be real

I am gonna use that as my wallspaper on my PC that is just too pretty

[dim cradle]

extreme super-hyperrealism is beyond real 🙂

[fickle ravine]

I made this more serene

Not that bad result

Not the best ever possible, but still pretty good

[dim cradle]

high quality, but i think the one you posted is much more serene 🙂

[fickle ravine]

So the original is more serene than the version AI was supposed to make serene?

Funny

@dim cradle I like that picture, it looks like Central Asia for me

[dim cradle]

in your post, didn't you say you asked the ai to make it serene? the trees and sunlight capture that

[fickle ravine]

Yeah I did

[dim cradle]

okay, i think it worked, don't you?

[fickle ravine]

yeah it did

I think both of the pictures look pretty serene and calm

[dim cradle]

mmhmm

[fickle ravine]

Your post is the perfect example of underrated

[dim cradle]

that's a nice compliment, thanks, 💜 dall-e

[fickle ravine]

I love trying to guess the country of each of those landscapes

[late blade]

Good luck with that

[fickle ravine]

@onyx ridge that looks so good

[onyx ridge]

Thanks

[fickle ravine]

I love the colors, they feel nostalgic for me

[onyx ridge]

not easy to get it to work.

I was going to include more images but #mypromptbudget

[dim cradle]

she enlightened

[dense mesa]

wow, one of my GPTs with pdf knowledge documents, API documentation for some non-OAI stuff, suddenly won't work. just says "something went wrong." then tried to upload them to another GPT and then data analysis. and errors. but they open fine and worked with the GPT before

hmm

🤔

not like they're wacky abnormally huge pdfs

[dim cradle]

i've seen similar issues recently with my gpts for rag, i thought it had something to do with the transfer to teams. maybe there's a larger problem.

[fickle ravine]

@dim cradle Oh your picture is so pretty

[dense mesa]

figured something got wonky with the configuration on the gpt at first. but then when it did the same in other modes it has me wondering

[fickle ravine]

That deserves to be in the hall of fame

[dense mesa]

not like it's some grand conspiracy, but just curious what is going on. because it'd been doing very well with reading knowledge files up until the last day or two

the best is when it can't read the file and it's like "well normally in this situation you'd want to do this and this" and tell me all the vague stuff I already know, lol. I guess poor gpt4 doing its best

[dim cradle]

It appears that I encountered an error while attempting to access your data from the provided CSV file. To proceed with the analysis, I will use a different approach to access the file and analyze. Please bear with me for a moment.

This is new, just in the last couple days. I thought it was related to migrating my workspace, but it sounds like we're having the same experience.

python errors all over the place

[deft musk]

@late blade delightfully Dys Topic indeed! #daily-theme message

[onyx ridge]

I showed the wrong image before. I needed to get more specific about the waveforem / topology. It was just like water, a simple brane representation.

[dense mesa]

yeah, this isn't the only instance of that happening recently. but when they're isolated I just assume it's regular old chatgpt errors

[fickle ravine]

Can someone please comment anything about my picture?

[fickle ravine]

The first one was better, but this is still really cool

[deft musk]

The nose knows?

[fickle ravine]

The newest ones

[late blade]

@fickle ravine patience young dall-e-wan, patience!

[glacial oasis]

What model did you use

[dense mesa]

I knows that nose knows

[dim cradle]

in my analytics gpt, i am planning to try wiping and uploading again all the knowledge files. hopefully that's all it'll take. not a great time for some of my most useful gpts to break

[feral minnow]

I believe this was ChatGPT, shockingly. It looks like a Bing generation.

[glacial oasis]

You mean the bing dal e 3?

[late blade]

both are dall-e

[feral minnow]

Indeed.

Yes, but I think Bing’s is stylistically distinct enough to differentiate.

(IMO)

[dim cradle]

which one?

[fickle ravine]

The one posted at 9:02 PM

I am afraid of posting the link here sorry

Just my latest one

[dim cradle]

it's fine to paste the link here

[feral minnow]

The model renders really cool fights.

[fickle ravine]

#daily-theme message

[dim cradle]

plenty soothing

[glacial oasis]

What is the difference between the bing dal e 3 and the gpt plus dal e 3 is it a really big gap like copilot

[fickle ravine]

There is the problem there is no exact definition of "serenity" that could help you know if it is serene

[feral minnow]

There’s a large gap in moderation. Bing is kind of a PG-13 model.

It’s Gremlins to ChatGPT’s E.T.

[glacial oasis]

So is there like any large image creation gap

[fickle ravine]

Wow this is amazing, when the theme will be "epic" you can post it

[feral minnow]

ChatGPT has different image formats, for one. I believe it’s also able to handle more complex prompts, right?

[onyx ridge]

I recreated your image under your image with emojis! 😄

[glacial oasis]

Oh ok

Anyone here have chat gpt plus

[fickle ravine]

This would also fit if mythic, fire, or legendary was the theme

[fickle ravine]

yes

Everyone

I guess

[feral minnow]

It’s so cheap.

[fickle ravine]

nah but ok

[glacial oasis]

Is it true that you can referral people for a free trial

[fickle ravine]

For how useful it can be it is cheap, otherwise not

[feral minnow]

I didn’t mean that as a judgment on you, sorry.

What I meant is that it’s a good value.

[fickle ravine]

oh yes

[glacial oasis]

Ye

Ive seen so many videos its amazing

[feral minnow]

I could see a less magnanimous organization charging twice as much to access a model like that.

(I hope I don’t give them any ideas.)

[glacial oasis]

Do you have a referral link?

[late blade]

Well, I just managed to do 12 images with DALL-E

[feral minnow]

Can you send referral links?

[glacial oasis]

I think so

[late blade]

without pseudocode stuff

[glacial oasis]

Thats what the person in the thread said

And i googled it

[feral minnow]

I don’t see the “invite a friend” link

Otherwise I’d give it to you

[dim cradle]

isn't there a "refer a friend" or "invite members" button at the bottom of the sidebar?

[glacial oasis]

I found a yt vid

[feral minnow]

I don’t have it

[glacial oasis]

Well i know one other way but no one will trust that

[dim cradle]

@onyx ridge i didn't realize the model is capable of generating typography like that until recently

[onyx ridge]

Pretty good success rate without the actual function, just a description. The output is not substantially different after several comparison tests. Thanks for the star!

[glacial oasis]

Wait do you have one?

[dim cradle]

i've seen that option in both plus and teams

[glacial oasis]

Bing ai said this
To refer a friend to get a free trial on GPT-4, you need to have a ChatGPT Plus subscription, which gives you access to the GPT-4 API. ChatGPT Plus is a service that lets you chat with GPT-4 and use it for various tasks, such as writing, coding, gaming, and more.

If you already have a ChatGPT Plus subscription, you can follow these steps to generate and share a referral link:

1. Log in to your ChatGPT account and go to the admin sidebar in the lower-left corner of the ChatGPT UI.
2. Click on the option "Invite a friend" and enter the email address of the person you want to invite.
3. A unique referral link will be generated and sent to the email address you entered. You can also copy the link and share it with your friend through other channels.
4. Your friend can use the link to sign up for an OpenAI account and activate a free trial of ChatGPT Plus. They will need to input their credit card details, but they won't be charged until the trial period ends.

[dim cradle]

we can only offer that to 3 friends, so we have to be selective, plus i've upgraded to a teams account and can only invite members. maybe someone else around can refer you from their plus account.

[late blade]

penalty box....

DALL-E hates me today... for overusing it lol

ok I think I got about 8 hours of daily theme covered now

[open trench]

lol

[late blade]

So, did Scamper survive the laundromat?

I think you keep him in your pocket at all time

[dim cradle]

aww

this sort of type for my name would be pretty cool if it would only manage to render all the letters with the same concept. the failure rate is high. still poking it.

[glacial oasis]

Couldn’t you just put the prompt and ask it to spell out your name

[late blade]

sonsidering youa re missing <s^3> and the h yeah, it's gonna be hard

[dim cradle]

not using the tag in this concept, just trying to refine and perfect for 4 letters lol

[dim cradle]

i don't think it would look as good without some more promptin'

[glacial oasis]

Oh ok

[late blade]

I just have to ask dall-e to do a dystopian image where nobody can spell my name and voila, I got the perfect image

[dim cradle]

haha

the prompt is conceptualized to give DALL-E a creative license....

Visualize the name 'Shon' in a most clever, cerebral, and stunning typography design. The letters should be crafted in a unique and artistic style that reflects innovation and creativity. Each character is a masterpiece of design, perhaps incorporating elements of optical illusion that play with the viewer’s perception, giving the name a three-dimensional, floating effect. The 'S' could be designed like a Mobius strip, representing infinity and continuous evolution. The 'H' might integrate architectural elements, symbolizing structure and strength. The 'O' can be an iris of an eye, reflecting a visionary perspective. The 'N' could incorporate digital or circuit-like patterns, alluding to technology and the future. The colors are bold and engaging, with a palette that captures attention while also allowing each letter to stand out on its own. The background is minimalist, perhaps a clean, gradient backdrop that doesn't detract from the complexity of the typography itself. Advanced lighting effects should give the letters a luminous quality, with subtle shadows and highlights that enhance their three-dimensional appearance. The overall composition combines these elements into a cohesive and balanced design, resulting in a typographic representation of 'Shon' that is not only visually stunning but also intellectually provocative. UltraHDQuality, superHyperrealism, advancedLightingEffects, highDynamicRangeImaging, dynamicTextures, depthOfField, layeredComposition, microTextureDetailing, complexReflectionsAndMirroring, vividYetNaturalColorPalette, perspectiveAccuracy, artisticInnovation.

[deft musk]

Let's introduce you to the magic of 'typeface' as it mixes with Dall-E 3.

Let's see 5 images, generate one at time in the same output, a welcome party with dark typeface carved into wood showcasing the name "Shon" in a bright sign to celebrate Shon's party. Typeface is important, Dall-E spells more accurately when using that as a description for written words.

Not all are perfect, but see the percentage that are?

[dim cradle]

i'm just taking all that in, it's amazing, thanks!

[deft musk]

Dall-E does not believe '<s^3>Shon' could be a name though, not even with the dark typeface trick

[late blade]

oh, for the uninitiated ones, what is the dark typeface lore? is that something you can order at Mc Donalds?

[deft musk]

More something you can order at Dall-E, when you want writin'.

[late blade]

ahaaa

[dim cradle]

i 💜 this

[deft musk]

Let's show 5 images of typewritten pages, each in bold, dark typeface text, with the words "I can spell with typeface, I can!" Create each image one at a time in the same output.

It's not perfect, but compare to any other method you know, anyone got a higher average yet by asking for a writing style use?

Even 1 perfect, with all those letters and words, typeface being an uncommon one - and the relatively few errors in the 4 flawed.

[late blade]

wow the free version of dall-e from microsoft is taking one quadrillion years to make images atm

[dim cradle]

i like this "S" my last prompt imaged for example -- last prompt was a different design for each letter -- i'd like to see if DALL-E could spell an entire word with the same font -- i'll try a prompt for that now.

[late blade]

oh, you can still create stuff without credits for free now, it just take more quadrillion years and then some more

[dim cradle]

i thought bing always did that, the boosts just put you in the fast lane, i thought

[late blade]

I thought it was 15 credits and the bye bye for the day

I'm so gonna abuse making kittens

I'm gonna make a kitten sweatshop

[late blade]

I'm divided

Zombie Cats.... or Sweatshops....

[deft musk]

Cause, they care about this. Neither cats nor zombies can sweat!

But I am pretty sure I can help Dall-E 3 spell 😄