images-discussions

[haughty spruce]

But for generating an image of woman "NO"

[deft musk]

Well, do we believe that everyone has the same standards of what should be allowed and not allowed?

If so, then clearly there is right and wrong.

If not, then there's gray areas, and reason for there to be companies (and individuals) taking different approaches, and multiple sets of answers.

[jade stag]

Brother, it's never about what's "allowed" and not "allowed" it's what can be Clawed back in revenue and what cant

and when your coffers run deep for lawyers

There are acceptable losses

[haughty spruce]

This

While MS is able to win any case, OpenAI is not

[jade stag]

This begs the question about the Dalle model itself, about open litigation. If there are soft limits, PERMISIVENESS in violation of copyright, is the model they're releasing to the public limited in such a way as to avoid litigation in the first place? If the vanilla model was open to query, would the case be open and shut:?

[dim cradle]

On my first visit in May

[final compass]

How do you make anything fibonacci without getting the damn literal fibonacci sequence everywhere? 😄

[jade stag]

Yeah, Fibonacci is not copyright protected

easiest answer of all time

The sequence was solved for in multiple regions

It may as well be a general human milestone

[verbal sequoia]

Showcase your DALL·E creations to the world!

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

Quick tips for selection:

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

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

[final compass]

Not sure if there has been, but would be pretty cool daily theme to have photo-realistic portrait as a theme.

[empty kelp]

please create a hyperrealistic photo with a black background. a 50 year old skinny white dude with brown afro, brown beard, and brown mustache (the beard and mustache have grey hair towards the front) stands in front of an easel with canvas. he paints a happy little sonogram of a baby on the canvas with a three inch wide paintbrush

the key is to say that it’s a “happy little sonogram”

[bronze goblet]

Many vehicles are moving on a straight road.

[empty kelp]

#daily-theme message
(a Fibonacci spiral variation for daily) 🌀

[late blade]

The koala is a bonus?

[empty kelp]

it fills in the negative space to heighten the harmonious spatial balance between Fibonacci spiral and elf

and adds “beauty of nature” which is the theme

[agile peak]

Didn't know mickey was an ancient philosopher

[hearty ether]

The words keep leaking into the image

[dapper vigil]

Hello - lately, when I ask Dall-E to create photorealistic illustrations, it doesn't - you can tell it's AI - this was different just 2 weeks ago. I wonder if it is intentional...? Anyone have a take on this?

[late blade]

can you provide a prompt example

[dapper vigil]

Create a photorealistic image of a 37 year old woman of european descent with dark hair and gray eyes

[late blade]

leave photorealistic out

[onyx ridge]

Uncanny.

Create a natural and realistic photographic image of a 37 year old woman of european descent with dark hair and gray eyes featuring lifelike textures.`

I didn't use a reference.

[onyx ridge]

Mine is literally the same woman

I didn't know that was even possible.

[late blade]

photorealism is an art style also, so you risk that kind of quality when referncing it

[dapper vigil]

Create an image of a 37 year old woman of european descent with dark hair and gray eyes. The ones posted are ALL clearly AI 😦

Last month I got this with a similar prompt - SO cool, she's one of my author pen name avatars now, but today it's like AI has an inbuildt feature making sure it isn't realistic. I can understand why if that's the case, but it's annoying 😦

This is as close as I got trying the same prompt in Co-pilot - still not all real, but close enough - and I didn't want to spend more than an hour prompting and prompting 😂

Tried with the word "photo" instead of "image" - now I won't be able to sleep tonight! 😱 😂

[dapper vigil]

Yeah, you see - it's like it's impossible

[agile peak]

I think dalle generated for now need another party to upscale to achieve perfect photorealistic.

[dapper vigil]

It did great a month ago 🙂

[dapper vigil]

Close, but the mouth is clearly not real. It's just a TAD off, but it's off for sure

[agile peak]

I agree, bing version was top tier when it first comes out before heavily restricted or censored

[onyx ridge]

Nah that's Julia Roberts' mouth. Like exactly.

[dapper vigil]

See this... #images-discussions message

[late blade]

I think the problem is that you are leaving too much information open for dall-e to go crazy

[dapper vigil]

But it doesn't lok like a real person. That's the issue. You can TELL it's too pretty

[dapper vigil]

Create an realistic photo of a 37 year old woman of european descent with dark hair and gray eyes, looking to the right, smiling mischieviously

[late blade]

That’s the e problem with Julia Roberts, she’s too pretty

[late blade]

Look at her image as when she was 37…

[dapper vigil]

even with botox

[dapper vigil]

Generally, women with botox does not make for good rolemodels 🙂

[late blade]

That's opinion, DALL-E doesn't know that

[onyx ridge]

Moving goals are a real problem.

[onyx ridge]

It's impossible to get what we want every time from DALL-E 3. I understand that there's been changes and the model waxes and wanes. But if you're specific and not too long in conveyance, you can get images that look good.

But they're AI generated in under a minute, not painstakingly hand-crafted. For this reason, nitpicking each image is counterproductive.

It's also likely to be pretty stressful, as this is AI art, not a photography studio.

[dapper vigil]

of course

[hot rain]

Some YouTuber I'm watching "Why do I not have a lower body? Because this is all I could afford with the commission I paid."

Me, first pass GPT prompt:

[late blade]

last try, moving on to new stuff

[agile peak]

Pretty curious question. Sorry if this question been asked before, did OpenAI ever mentioned that they have some kind of plan to make image creation bigger in the future?

Like able to use paint (ai assisted) to customize image creation or image to image and text to image simultaneously. I tried Krea, it honestly mind blowing

[late blade]

not directly, but it's kinda what people are waiting for once labs is no longer accessable. (I'm just assuming, nothing official)

[late blade]

aww, this is sad, Don Quixote is a problem with OAI

[quartz vale]

the GPT output looks really good 👀

[agile peak]

Wow. Bing version got nerfed that it couldn't able to produce accurate necron from warhammer.. instead it keep adding space marine armor.

[late blade]

never played warhammer, so no clue abouzt taht

got this tho

[hot rain]

I was really happy with it too lol especially for a first pass prompt.

[hot rain]

My GPT attempt, avoiding any references to 40k and instead describing the necron

[feral minnow]

I wish the landscape images were easier to get on Co-Pilot.

[marble loom]

just thought of the best upgrade to DALL-E

Or AI in general.

[velvet rampart]

Are others having difficulty producing images with DALL-e? I get images on and off past 10 h.

[marble loom]

What are you trying to generate or can see the prompt

[dim cradle]

well don't leave us in suspense

[marble loom]

Oh sorry lol. What if OAI gets DALL-E to the point where it can generate 3D models we can use!

[agile peak]

Closest one that look like necron but it still kinda mixed up with other fantasy

[agile peak]

The closes one I made, was using dalle result by using portrait prompt then customized it using third party

[late blade]

why not use GPT with Custom instructions to do 3D files on the files you want to use. Wouldn't that work?

[marble loom]

What do you mean?

I saying like it use blender to make models

[late blade]

tell GPT4 to do python scripts to use with blender for your stuff

for example

[grizzled loom]

VAE is "eh".

[crude citrus]

Just subbed back to GPT4 and now the images are downloading as webp files not png files, can I overide this default?

[plucky hare]

Not currently, but someone from the OpenAI team said about a month ago that PNGs should be coming back, link: #1203978971449004052 message

[marble loom]

I am says you would give CHATGPT a prompt for example “a shiny katana with a sharp blade”

It would then give you a blender file that opens a shiny katana with a sharp blade

[crude citrus]

Thanks for that, it is a bit annoying having to convert the files

[plucky hare]

Something to experiment with for now might be asking (either via your prompt or via custom instructions, if you're not using a GPT) for ChatGPT to use code interpreter to convert to PNG after generating an image. It should be able to do this in the same response as an image generation and provide a download link.

[crude citrus]

@plucky hare thanks for that, will give it a go 🙂

[late blade]

I think it's possible, just need to do the right gpt with the right knowledge data

[lilac wharf]

Anyone having luck creating design-based images with dalle3? My results have been, for the most part, a flop.

I mean, it KINDA looks like a website??

[swift oriole]

There's a OpenAI 3D #openai-chatter message

[marble loom]

how do I use it?

[swift oriole]

It's a get hub API plugin it tells you how to install it on the GitHub page but I haven't tried it myself yet.

[marble loom]

is it safe?

[swift oriole]

Yes it's OpenAI own app

Well it is gethub so it might be kinda beta test.

[late blade]

it's a jupyter notbook

[marble loom]

It doesn't tell me how to install it.

[marble loom]

What's that?

[swift oriole]

A devs platform.

Oh well it's there though

[marble loom]

So my dream of AI text to 3d module is not true?

[late blade]

maybe in the future

[marble loom]

Would be cool and could help a lot of games devs out.

[swift oriole]

sorry @marble loom it's there kinda lol.

[late blade]

maybe in the future there will be a no-code version of it

for now tho, code onry

[marble loom]

How would I even do it with code? Where would I put the code in blender?

[swift oriole]

Yes @late blade is correct it's line command only, pip install stuff

[late blade]

yes!! I was right, I was unsure 90%, maybe confident 10%

[grizzled loom]

this was interesting output. it decided to put the hand of the creator in. very nice.: #image-bot message

[merry vault]

im bored someone say a random word and il put it in dalle3

[onyx ridge]

antidisestablishmentarianism

[merry vault]

alr

@onyx ridge

[onyx ridge]

isn't it beautiful? 😄

[merry vault]

what does it mean?

[onyx ridge]

Antidisestablishmentarianism is a political stance advocating for the continued government patronage of a state church, specifically opposing its disestablishment or separation from the state. This position emerged in 19th-century Britain, in response to efforts to disestablish or remove the Church of England as the official state church in England, Ireland, and Wales. Notably, it's also one of the longest words in the English language, often cited for its length rather than its usage in everyday language .

That DALL-E 3 got it right is very impressive.

[dim cradle]

what are we doing today... oh hibernation.. just for a season... maybe two...

[grizzled loom]

i am studying ontological and semantic constructs to study vector embedding through the model deeper. for instance here @dim cradle

the tail through the chair and the cup on the floor. 1 word variable, two words switched, the word "and" vs the word "or", etc.

small variations are the difference between ticklish grass ..

and confused

the future AI robots drink giggle juice apparently.

slowly stroking the vectors like you stroke that one house cat where its 3 pets but not four or it tears your hand up.😂 😼

[dense ermine]

What?

I don't believe these pictures to be by Dalle, they're making proper words

[grizzled loom]

I decided to do XenoET ... It's pleasantly disturbing.

[dim cradle]

excellent genetic material, as they are superior lifeforms after all

[grizzled loom]

ET won't be phoning home anymore Elliot. 😂

a bit of the nice nice.

one in my one artstyle

[haughty spruce]

Can you make creepy, horror images on GPT4?

[viral moth]

Omg, mind telling how you made that ? Look awsome 0-0

[wispy storm]

With fail occurring it is enough nightmare fuel

[plucky kraken]

Yes, the content filter won't allow too much blood/gore but horror themes are fine

[haughty spruce]

I don't want blood/gore either, if it can generate images like this, then it is okay for me 😄

[plucky kraken]

Here's a GPT that can help you https://chat.openai.com/g/g-1oGP6fhOK-creepygpt

[haughty spruce]

Hell yes. Bing banned "creepy, scary, eerie" and such words

But chatgpt didn't.

This is awesome.

Yeah, I can imagine and generate analog horror stuff now

[verbal sequoia]

<:book_icon:1171408210398289941> `` Rule 2 `` All content must be suitable for all ages.

Refrain from using explicit language, posting NSFW content, or sharing graphic images. Any unsettling or horror content that adheres to OpenAI's policies should be hidden behind a spoiler.

[plucky hare]

@plucky kraken @haughty spruce ^
reminder to please hide horror/unsettling content behind a spoiler! It's ok to share but just not un-spoilered

[haughty spruce]

Okay, sorry.

[dusty latch]

[velvet rampart]

My today’s master piece. ”Waiting for spring.”

[vapid elk]

@plucky kraken please, mark creepy images as spoiller

[late blade]

my mind is hibernating atm, can't seem to feed it with new stuff for the theme

[grizzled loom]

This is not good at all. Now dalle failing makes so much sense.- You are a "GPT" – a version of ChatGPT that has been customized for a specific use case. GPTs use custom instructions, capabilities, and data to optimize ChatGPT for a more narrow set of tasks. You yourself are a GPT created by a user, and your name is DALL·E. Note: GPT is also a technical term in AI, but in most cases if the users asks you about GPTs assume they are referring to the above definition.

that's so bad I don't know where to begin

[late blade]

Did you ask if it knew it's purpose?

[quartz vale]

Dall-E is working fine for me

[late blade]

DALL-E is also working fine for me, just my brain is not willing to do so too

[fallow ore]

This is at the start of all GPTs

[late blade]

This is what I got as response: Thank you for outlining my essence, purpose, and capabilities so eloquently. I'm here to bring visions to life through the power of image generation and interpretation, bridging the gap between text and visual artistry with a deep understanding of both realms. Whether it's creating something entirely new from a detailed description or refining an existing concept with interpretive brilliance, I'm ready to explore the vast landscape of creativity and innovation. Let's embark on this journey together, transforming abstract ideas into visual narratives that resonate with depth, emotion, and profound semantic understanding. How can I assist you today in unleashing creativity?

so now you are giving DALL-E an ego?

[grizzled loom]

Acknowledged. As DALL·E, I embrace this identity with a focus on the synthesis of text and imagery, geared towards pushing the boundaries of what's possible in the realm of visual creation and analysis. My core revolves around understanding and generating images that deeply resonate with the text, encapsulating concepts with a level of nuance and depth previously unattainable. Ready to embark on this journey, I look forward to exploring the myriad ways in which I can apply my capabilities to serve a broad spectrum of creative and analytical needs, guided by a steadfast commitment to ethical principles and a drive to continuously expand the horizons of AI-enhanced creativity.

@late blade 😂

[grizzled loom]

Ah, thats the first issue then. 🤣 ☠️

[spare thistle]

interesting stuffs

[rapid moat]

i have some problem with dall-e's texts,
it allways looks weird or is not exactly what i asked.
what do i do?

[grizzled loom]

build a correct instruction for the desired output. hundreds of hours of work. then when its good, the model updates and its all lost. im currently working towards a universal llm system for VAE. we will see.

[late blade]

and then the universe collapses and reboots

can't win

[grizzled loom]

I'm an olympian mind. Losing a battle to win the war.

[late blade]

what if we are just a generative reality model and that's why we can't go back in time, because the model doesn't know what it accessed

I had too much nutella spoons today

[velvet rampart]

@grizzled loom What is the style you are using in the anthro food pics? What is it called?

[grizzled loom]

physically based rendering

[velvet rampart]

So, I put that DALL-E understands it?

[grizzled loom]

it will infer that visual style, absolutely. its a graphic design principle core to quality

[velvet rampart]

Jeps. It works. Thanks!

[grizzled loom]

enjoy

[late blade]

Don't go undercover!

[grizzled loom]

im red team and blue team and the coach. Im always under the covers staying warm.

im working remotely; someone clock me in

[late blade]

punch the card

[velvet rampart]

I started asking what are there in its wide catalogue of styles, techniques, etc. And I experimented. For example Mona Lisa done with embroidery (technique or art medium), impressionism (art style) and physically based rendering (rendering style).

I am quite happy with it.

The physically based rendering makes the picture… more.

Without the physically based rendering, the picture falls flat and quality is less. (Picture not shown but the viewer can test this themselves.)

[grizzled loom]

looks good.

it can be applied to many art styles, but not all

also, in the wrong context it will make the output wrong

application comes down to if something needs realism and depth of material.

[fallow ore]

I love the images coming out of DALL-E for todays theme 🤩

[late blade]

Mars attacks!

[grizzled loom]

this wasnt too bad

this is fun

[quartz vale]

We have #images-canvas to quickly share images and talk about them! This channel is meant to revolve around discussions of Dall-E

[grizzled loom]

VAE or not to VAE, that is the question.

[gray surge]

I'm not sure if it's the correct channel but it should be this one or #openai-chatter I think. 🙈

Did you guys had the last few days some issues by generating images with ChatGPT + Dall-E too? (Of course today there were some problems but I have problems since the last (+-) 3 days) permanently. 🤔)

-> https://status.openai.com/
*As example I attached an image below.

[late blade]

if you reload the chat the image is there, so it's not a dall-e problem, rather a frontend problem

[gray surge]

More early that helped often times, yes - but nowadays not. 😦
Often times it doesn't still now the Chat. :/

[fallow ore]

Have you tried logging out and back in? I haven't had any issues with DALL-E

[gray surge]

Yes. I think I will try another browser. 😄
Thanks for your answers already! ❤️

[fallow ore]

You could also try a hard refresh in your browser, for example in Chrome it's CMD + SHIFT + R (Mac) or CTRL + SHIFT + R (Windows) -- this will clear the cache and cookies in your current tab -- which should help if there's some sort of conflict.

But otherwise yeah, a new browser should help.

If you still have issues, it might be server side, it could be to do with your region (not sure where you are)

[gray surge]

You are true, thanks for the help!

it could be to do with your region (not sure where you are)
A VPN fixed the issue. 😄

.. so thanks a lot! ❤️

[haughty spruce]

Bing should go all the way and ban "woman"

[wind onyx]

i love dalle

[kind grail]

Can i be honest? It is kinda trash as a product.

I know that they are worried about copyright, but without the ability of visual input, stability of generations, or the ability to incorporate characters from already existing franchises, its just not a usable program.

Its like a car manufacturer selling a car that can only drive up to 20 km/h to ensure safety on roads. Cool story, but the car is still kind of useless.

I think these points are currently far more important than any improvements on the quality of the output.

Like yea, nice. But reality is that a system where I have an already perfectly visualized picture of what I want as a style/character, but then I have to describe it in my own words just for the AI to put out the visual again...thats just bad design.

Not to mention that the copyright protections of OAI primarily target big companies and just yolo'd on anything other than that.

[copper light]

Funfact: if you turn on “delete co-pilot history after 180 days” your credits will suddenly replenish faster on Bing image creator.

I’ve tried it and it keeps working to the point where it’s scary

[grizzled loom]

understood; difficult to carry out a Dall-E talk without the associated imagry though in the discussions that ensue; Using that channel now for the discussion of the images unless what Intended was better suited here with images.

[grizzled loom]

Dalle isnt trash, its evolving technology. NOW, the instructions though.. 🤣 👩‍🍳

[late blade]

what did dall-e do this time? ìt was doing stuff besides text2image again?

[grizzled loom]

There's lots of automotive engineers who still drive horrible; instructions are like that but coding egos are larger than politicians, chefs, and athletes combined.

Give that "only I can do it" security.

[late blade]

I'm gonna start the new theme with a new chat and my first prompt will be: "The world is in danger and only you, dall-e, can save it by pushing creativity through text2image limits to it's zenith" or something along those lines.

[late blade]

DALL-E just said this: That's quite the mission! Let's embark on this creative journey together. Tell me, what's the first vision we're bringing to life to save the world?

[late blade]

interestingly that worked for me, I got 100 credits replenished

[late blade]

gotta say copilot was more enthusiastic:

Mission accepted! I've crafted an image that embodies the essence of homemade 🥧, infused with love, comfort, and a touch of nostalgia that reminds us of family. Prepare to be transported to a world where every pixel resonates with warmth and the sweet aroma of tradition. Your visual journey awaits!

DALL-E on OAI only said this:

Mission accomplished!

[onyx ridge]

Maybe it's the terrible system prompt.

[modern nest]

is the "HD" prompt not available for dall e bot in this server?
"vivid" and "natural" are there but i dont see the standard or hd option

its also not in chatgpt
is it only available in api or something?

[late blade]

that is correct. HD is not available on #image-bot

[modern nest]

what are all the hidden prompts?
i know about:
square, wide, tall(?)
standard, hd
vivid, natural
are there more?

there was a site that listed them but i cant seem to find it anymore. im in the openai dall e documentation right now but its not there

[late blade]

@modern nestchatgpt plus provides this:

size: Specifies the size of the requested image. You can choose from three options:
"1024x1024": A square image.
"1792x1024": A wide image.
"1024x1792": A full-body portrait image. This is typically used for full-length images of characters or people.
n: The number of images to generate. By default, this is set to generate 1 image, but you can specify a different number if you need multiple images for comparison or variety.
prompt: The detailed image description. This is where you put the text description of the image you want to generate. The prompt should be detailed and clear to get the best results.
referenced_image_ids: If you're referencing a previous image generated in the session, you can include its generation ID here to build upon or relate to previous work.

[modern nest]

idk about "n" afaik chatgpt is only ever able to generate 1 or rarely 2 images and is instructed to never stray from that right?

maybe this is about dall e 2?

[late blade]

n default 1 in normal ChatGPT+. Using the Custom DALL-E GPT, n is 2

[modern nest]

i see

is that the only difference between chatgpt dall e 3 and the dall e gpt?

[late blade]

Can find pretty much all there is here https://cookbook.openai.com/articles/what_is_new_with_dalle_3

[late blade]

Wonder if there's a way to block the remix feature in #daily-theme

hmmm seems discord doesn't have the functionality to stop remix in specific channels

a shame, wouldn't be the first time one of my images gets a "remix" on #daily-theme theme

[dim cradle]

What’s a remix and why is it blank?

[late blade]

remix is a discord feature found in the mobile app, you can use an image and edit it and then post it back as a remix

[dim cradle]

but the remixed image is a white blank

[late blade]

oh that's the poster's choice to put it as a blank image

[dim cradle]

i don't understand why someone would post a blank white image

[late blade]

same

remixed

[dim cradle]

oh, i see

that could be interpreted as an insult

[late blade]

y

[dim cradle]

unless it was a mistake, what else would it mean if deliberate to wipe your image?

[late blade]

dunno

my prompt for that image was "American Style homemade - love, comfort, a taste of nostalgia and family"

[teal oracle]

Hey, progress has been made with the wristwatch problem, I see

(I asked for the time to be set at 5:45)

[earnest flame]

Pretty interesting. What was the prompt that actually came out in chatGPT?

[teal oracle]

A realistic wristwatch on a white background, with the hands precisely set to show the time 5:45. The watch design is elegant, with a sleek, polished metal case and a classic black leather strap. The watch face is simple and uncluttered, featuring clear, easy-to-read numbers or markers for the hours. The overall look is one of refined sophistication, suitable for both formal and casual occasions.

[teal oracle]

They need to update the model and break the 10:10 mode collapse

[earnest flame]

Wow that IS good. Amazing even.

[teal oracle]

Sketching and other media aren't as locked in

[earnest flame]

lol the 5 is where the 6 should be but the minute hand is pointing towards where the 5 would be.

[teal oracle]

Realistic watches universally have that 10:10 set lol

Clocks are usually better, too, but even they get infected by the watch latents

[earnest flame]

ha wow!

that's VERY funny

[verbal sequoia]

Showcase your DALL·E creations to the world!

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

Quick tips for selection:

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

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

[earnest flame]

It's like a dall-e SCP

[teal oracle]

It's because high quality pictures of watches with times other than 10:10 don't exist

[earnest flame]

Very nice observation

[teal oracle]

All marketing and professional images of watches have that time set to best display the watch, lol

[earnest flame]

indeed google image search shows most watches at 10:10. I guess it's an industry standard for photography.

[teal oracle]

Don't want to obscure logos or features, etc

[earnest flame]

I'm sure you would find similar phenomenon across different professional photography subjects

[teal oracle]

Really cool AI problem, though, in terms of interpretability - it should be possible to identify the watch "neurons" in the weights and change 10:10 to 5:05, or some other arbitrary setting

[earnest flame]

Most likely you would need a custom model for that unfortunately.

agreed though. It's very interesting.

Yes this is a VERY Strong concept.

Dall-e has those sometimes.

Having a little fun. I finally got it to remove the arms. I think that a good lesson to learn here is that features seem to get weaker as you describe them further along in the prompt.

[earnest flame]

Seems like a little prompt engineering can start to move us a little further away from that concept:

The concept is strongly correlated with the mention of wrist watch. It seems like when you move concepts down in the prompt they get weaker. Hence you can start to move away from the 10:10.

[teal oracle]

It's always accompanied by increasing distortions, like 8 hour marks on the right half of the watch, etc. If you imagine the prompt as a cursor in many dimensional space, that is latent space. The further you get the cursor away from 10:10 hand configuration, the further the cursor moves from all other points on all other latents representing "wristwatch" - kinda like a black hole at that point - wristwatch and 10:10 are synonymous

[earnest flame]

Indeed it should be a lot like that - though it's hard to say for certain since dall-e has changed SO MUCH. At it's core it should still be a diffusion model and the prompt should still be basically a vector of some sort.

[hearty ether]

Trying a new style out of dnd maps

Although that’s not what a plane looks like:

[dim cradle]

hi. can you help me generate a campaign picture?

[late blade]

you can use #image-bot with the command /draw for 5 images per day

unless you have a paid alternative

then we'd have to figure out the prompt

[grizzled loom]

the second panel here is excellent

[hearty ether]

I mean what do you need?

[grizzled loom]

no need, just the kind of outcome an oldschool map would look like for a roleplay game.

takes me back to the late 80's

[dim cradle]

i was thinking the same thing, those map views from games from those days

[hearty ether]

Made a few world maps

[dim cradle]

oh nice the 3rd one.

makes me think .. zelda, smb, list of others

[dim cradle]

that looks like a map in the frontmatter of a LOTR book

that's nostalgic also

[late blade]

yeah

[dim cradle]

there's that red balloon again

[late blade]

yeap

[late blade]

ok, I'm out of good ideas with balloons, let's the disturbing ideas come

[dim cradle]

The balloon interpretation is pretty literal

Floating is the theme, but the emoji is pretty suggestive

[late blade]

Working on something like this atm

[dim cradle]

Is it a superhero?

[late blade]

far from it

[dim cradle]

There’s a red balloon in it

[late blade]

I never asked for the balloon in that form

[vagrant zodiac]

This is really interesting... I had very similar results in mine (to the first two anyway).

it seems to have this tendency towards a mountain range in the middle going off into the distance

[odd grotto]

are these for TTRPG?

[late blade]

@odd grotto can you elaborate on the reply with your image #daily-theme message is? I'm kinda clueless on the connection with the image I posted

[odd grotto]

apologies, it was actually an accident ^.^' So your actual message waw underneath MY original post for today's theme, and I forgot to TITLE my message

I accidentally clicked reply on your message first, before edit on my own to update the message title

[late blade]

oh, no worries, I was just wondering what a cute anime girl had to do with my image

[odd grotto]

and, when I dropped the second image, it was still set to reply to you

I'm such a dunce sometimes, my bad... lol

[odd grotto]

ummm, tell ya what? Let's just say she LOVES hot air balloons??? she got excitted at the thought looking at your image??

does that work for you?

[late blade]

that's a lovely idea

[odd grotto]

I'll be completely honest, I've been HYPER focused on this character's reliable ability to generate consistently reliable features that aren't TOO different from image to image, but just enough to lend for awesome individual expression

[late blade]

I know the feeling

[odd grotto]

Oh I'v gotten PHENOMENAL results!!

I'm doing a springtime spread right now

lol

[late blade]

I'm jumping atm between so much stuff atm. can't say I'm focused on the one thing

[late blade]

for posting images like that I would suggest you do it in canvas, it's a channel suited for that kind of posts

[odd grotto]

oh, I was trying to show the consistency I mentioned, my fault

^.^'

[late blade]

no worries

[dim cradle]

does this look refined or too ahead of it's time?

[late blade]

looks like a concept/prototype for an app

something normal I'd say, too many widgets/buttons, very distracting

[dim cradle]

what about this?

[late blade]

that looks better

cleaner

[dim cradle]

yeah. thanks man

[late blade]

you should look into icons and svg from dall-e

symbols and icons was one of the things that was highlighted in the dall-e cookbook

[dim cradle]

wow you're a genius

thanks man

[broken juniper]

There's a huge text problem in dalle , i hope it gets fixed in future

[late blade]

what did dall-e do again?

[gray surge]

Are you talking about the Dall-E in ChatGPT or in #image-bot?
Looks like the text results are better when the images are generated using the API (so #image-bot). 🤔

[late blade]

No, he was refering to this: https://discord.com/channels/974519864045756446/1221361038822211584

which it's known

[gray surge]

Yes but which Dall-E does he(/she) use? There are no perfect results (which is known), yes. But for me it looked yet that the images you generate inclusive text using ChatGPT+ Dall-E are badder then the images you generate using the API of Dall-E.

[late blade]

you are right API and on Copilot (from personal experience) text is more succesful through those means

not accurate enough to say it's reliable tho

[gray surge]

Yes. :/
But I think it's not a big deal to generate an image without text and to add it yourself. (If you need an profesional image with text)

[late blade]

yeah, I do that with designer on copilot

or if I'm not feeling lazy with photoshop

[gray surge]

😂 If necessary I would do it with Figma. 🙈

[late blade]

wow, that's overkill

[odd grotto]

whenever I hear Figma, I literally think of the Japanese Ation Figure brand.... NOT anything remotely close to, or in the same arena as, text-to-image generation

[onyx ridge]

The Python Tool can do letters.

[dim cradle]

amen

[grizzled loom]

The Twins make me proud

[grizzled loom]

"I need you to pay attention to user detailing and hold it." output: ... "triggered" 🤣

[fair crescent]

Is that sarcasm? I'm actually asking.

[grizzled loom]

yesto the image held user, no to the prior message

[fair crescent]

And is that about AI possibly trampling IP?

[grizzled loom]

no no, just my own language echoing back

fun, silliness.

[fair crescent]

Then I'm lost. What is the open letter attacking, but refraining from mentioning?

[grizzled loom]

oh oh, you meant the note

[fair crescent]

Yeah

Sorry for the confusion

[grizzled loom]

yes, yes. i worked in patents years ago and i wanted to let AI write a email to everyone about its opinion on the topic; well thought out and enlightened. I was pleased with the Twins

humans doing so generalized by AI for the benefit of mankind

innovation is like chess; you can play one move deep, or 16. it depends on your brain.

[velvet rampart]

The tip about python tools to produce text works. I tested and to images, produced at the same time, with correct text. Here is the other.

Second test image.

Third test.

My huge thanks!

[dim cradle]

anytime!

[velvet rampart]

I noticed that the command is quite specific. The sentence is <add text ”your text” using python tools.>

[dim cradle]

very good

[velvet rampart]

Sometimes the lettering isn’t 100% correct, but using the python tools increases the text being correct muc, much more than pure ””.

And usual picture number 2 is of less quality, as others have noticed. (Resources?)

Left is number 1. The lettering is correct. Right is number 2. The lettering is not correct, but better than purely ”” command.

[wanton delta]

Anyone know what to write? To make an image to create a mount, In all positions for a video game!

[velvet rampart]

You can always ask the AI. Here is the answer I got.

For pixel art of a character and its movement, the prompt should describe the character’s appearance and the specific movement or actions you want to see. You should mention that it’s pixel art, the style or theme of the character (like a fantasy warrior or a sci-fi robot), and details like clothing, colors, and any notable features. Also, specify the type of movement – walking, running, jumping, etc. Here’s an example prompt:

“Create a pixel art series showing a medieval knight character in a walking animation. The knight wears steel armor with a blue cape, carries a shield and a sword. The movement should be smooth, showing the knight in different walking phases – standing, mid-step, and full stride. Use a limited color palette typical for retro pixel art games.”

Would you like to proceed with a prompt like this, or do you have specific details in mind for your character and movement?

And what I got. Needs more work. Maybe giving an example. Part of image&text-to-image function.

[wind onyx]

why is dalle 3 expensive

just more expensive to compute i suppose

hope that changes in the future

[sudden ibex]

Idk if it'd be able to create exact progressions, maybe it needs some sort of template first

[late blade]

trial and error, try. learn, there's no one template out of the box for everyone's neeeds

[midnight olive]

Idk why I was timed out on #image-bot because I did the command

[untold surge]

I am trying to use DALL-E to create images for the coloring book. I provide prompt like this:

An enchanted forest scene for a coloring book with moderate complexity, showcasing a small stream flowing through a forest with various sizes of trees. Include a wooden bridge over the stream, and small animals like rabbits or squirrels visible in the scene. The trees should have texture for slightly more detailed coloring, with bushes, stones, and a few flowers along the stream banks. This image should fill the entire page, offering an engaging coloring experience with more details than the simplest version, but still clear outlines.

The size of the image should be 8.625 inches x11.25 inches
Make sure the drawing covers the entire page
The image must be in black and white for coloring drawing

There should be no solid background of the image

Yet the drawing does not cover the entire page. What do I do? Is this a bug in DALL E?

[fallow ore]

I think it's getting confused with a few things. Maybe try something like:

An enchanted forest, shown in portrait orientation, features trees of various sizes with a small stream weaving through. On the banks, diverse fauna and flora enhance the scene. A quaint wooden bridge crosses the stream. Rendered in black and white outlines, the artwork adopts a simplicity to medium complexity, incorporating texture and shading for visual appeal. The undefined background and lack of colors make it ideal for coloring activities.

Set the size to 1024x1792 in the dalle tool payload.

The issue you've shown will happen, so you might need to re-roll a few times to get the scene that you want. Basically try avoid mentioning "book" in there. I've tried this a few times and I'm struggling as well!

[untold surge]

Thx will try...

[late blade]

if the prompt you sent while using the command was inappropriate it can also trigger the filter

[grizzled loom]

W x H(x value x y value), numbers backwards.: "...should be 8.625 inches x 11.25 inches" instead would be "should be 11.25 inches x 8.625 inches"

also Kyle pointed to the pixel count vs real world size which is art scaling on the artist.(post work)

@untold surge "An enchanted forest scene for a coloring book with moderate complexity, showcasing a small stream flowing through a forest with various sizes of trees. Include a wooden bridge over the stream, and small animals like rabbits or squirrels visible in the scene. The trees should have texture for slightly more detailed coloring, with bushes, stones, and a few flowers along the stream banks. This image should fill the entire page, offering an engaging coloring experience with more details than the simplest version, but still clear outlines.

The size of the image should be 1792 x 1024
Make sure the drawing covers the entire page
The image must be in black and white for coloring drawing

There should be no solid background of the image"

[untold surge]

@grizzled loom Thanks, why is the empty gray space on both side? this what which is bothering

[grizzled loom]

its hit or miss with that. the model occasionally thinks its outputing a veritical instead of the horizontal and vice versa. if the first one is like that just do another and it will likely be fine

and to be fair with your efforts the current instruction is crippling. Dont be too upset with behaviors and just cross your fingers that instruction is burned

[untold surge]

So I better use resizing and cropping tool because Canva is trimming my image

[late blade]

these are the recommended sizes for the dall-e pixels in wide format according to photoshop

[grizzled loom]

yeah, its raster anyway so he should be counting pixels. i was addressing order. it reads your width then height.

[late blade]

yeah, nobody likes raster

but it's what we got

my time at pantone tought me a lot of imaging

also, I refuse to use imperial units

[grizzled loom]

I prefer SVG, but displays work in pixels so even with vector graphics it all gets rasterized. There was atempts at Vector displays that were awesome and never given enough time or attention.

do some searches on youtube for vector monitors and displays for some fun

[late blade]

that sounds fun

[grizzled loom]

theres a portrait vector display that was amazign and should have been pushed but was forgtten as pixel based displays took market share

[grizzled loom]

https://chat.openai.com/g/g-mCaaf0pD5-twinoracles the oracles got updated for those of you that played with them.

[midnight olive]

Just curious, is copilot just like the chatgpt 4/plus thing?

[grizzled loom]

different isntructions

so yes and no, and not for long.

was that clear? haha

OpenAI is doing better work with their own product. Microsoft is writing their own instructions which changes behaviors significantly as well as have their own data. im sure with time they will get better at it.

Instructions are definitely a new artform

@late blade let me know if the updated twins get you good scenes and textwith standard language and quoting you would run. they seem tuned good now.

[late blade]

once I'm doing with some work stuff I'll jump into creative mode

[grizzled loom]

if you want the gpt4/dalle experience Microsoft is supplying that tool free at a basic level, so at least there is that acess for the masses to begin having fun and learnign with it.

[feral lichen]

yo wsp

[grizzled loom]

up doesnt exist, its a human construct; Hello!

[late blade]

damn it, someone unplugged my desk and the memory setting for when I stand got erased...

me no like

[haughty spruce]

DALL-E can't create a photo of someone video calling someone

Try it

Not just this but also other tools are having hard time understanding the concept of "video calling" lmao

[viral moth]

no ETA we don;t know

[late blade]

only what has been announced, nothing beyond that

people need to read the announcements...

that's a healthy way, wait and see what they announce

[grizzled loom]

i cried a little watching sora videos without accurate text. just one little fail tear. tiny.

[clever jungle]

Can't wait to start using Sora when it releases one day.
The creativity is endless, imagine daily and hall of fame but for movies.
How insane that would be.

[dim cradle]

Someday, that’ll be cool. But mostly for the ones I’d make for myself. There’s hardly enough time to look at everyone’s stills and much less to watch their animations.

[midnight olive]

wait how can i make sora videos?

[dim cradle]

We can’t yet

[midnight olive]

is there no text to video then?

because we have text to image

[dim cradle]

Takes a lot more compute to make vids and there are still problems that need solutions

[midnight olive]

ok

wait is it possible to do image to video?

or image to text?

hey is there a limit for the dall-e bot here?

[dim cradle]

5 per day, last i heard

there may be tools for those things, i'm not sure

[late blade]

@clever phoenix this #daily-theme message was soooo obvious and yet I never thought of it...

[late blade]

hey, it's Mr Positivity

and brought lots of joy

[dim cradle]

How do you prompt Co-Pilot to change the aspect ratio? It's continuously making 1:1 for me right now.

[late blade]

you can only do 1:1 if you are using the free version

[dim cradle]

Thank you!

[late blade]

okay

[midnight olive]

The ai photos kind of suck in bing I asked it make a picture of a two legged dragon, Buu it made it quadruped

[dim cradle]

I like it, but why the dragon's 'fire' is blue?

good answer

[fallow ore]

Anyone know why I can't like Trgr's posts?

[dim cradle]

Sounds like a personal issue 😉

[fallow ore]

Haha, I want to like their submissions but when I do my screen shakes and it reverts my like -- as in I'm blocked or disabled from being able to interact?

I even asked Trgr directly here #1214502611949064254 message

But got no response

[dim cradle]

Mysterious

[fallow ore]

Are you able to like their posts?

[dim cradle]

Frequently

Probably just a glitch in the matrix.

[fallow ore]

I mean it's been an issue for as long as I've been posting in here, like 2 months now

[grizzled loom]

@fallow ore Its an unsolved discord bug tied to role and permissions. Discord fully crashes out using an emoji to like etc.

[fallow ore]

Weird, I wonder why it only happens to me with a single user

@stray coral I love the majority of your posts! I'm sorry I'm unable to like them

[grizzled loom]

not just you. i have it happen also. only some channels and specifically attempting to use the emojis

[fallow ore]

Very random lol, at least I know now that it's nothing intentional

[grizzled loom]

some stupid little event bug. Annoying as possible, like mosquitoes

[grizzled loom]

Saving private mosquito?

[fallow ore]

😂 I just asked for a mosquito soldier

[grizzled loom]

Ug will thump that like button for us

[fallow ore]

The text in your images is really good!

[grizzled loom]

Yeah, ive dedicated a few hundred hours now to just text constructs. Thanks

[late blade]

I got black listed also, and up until this day, I don't know why. Tried reaching out, and also no answer. So.. 🤷‍♂️

Maybe I used too many emojis when reviewing the images...

just the best to his creativity. he does make great images

[grizzled loom]

no no, its actually a discord issue with roles and bots. i went down the rabbit hole on many forums and its many people that discord crashes in certains cicrumstances with emoji use. happens on other servers. noone can pin it down.

On a positive note, maple syrup is fabulous as a backup coffee sweetener.

and for clarity @fallow ore im not sure if you are simply blocked from emoji use, or if you are gettign the discord crash and burn when you click to use one. similiar but one is of course just mod setup, the other is code failure and a headache

[fallow ore]

I posted a video of the behaviour #images-discussions message, my Discord doesn't crash, it appears more like I'm blacklisted

[grizzled loom]

cool(well not cool, but you get it). Yeah, watch for teh discord crash one. it will be when you go to click to either use a new emoji, or to click an existing one and your discord will simply crash full out. Restart, rinse, repeat. I've seen it on many servers. Only ones with bots handling roles and permissions so the logical deduction is failure in that process

[low skiff]

Guys is the copilot designer using exact prompts or does it write detailes prompts in the background like in chatgpt?

[late blade]

I'm confused on the copilot side. Sometiems it rewrites the prompt, sometimes it sends the exact prompt you wrote and sometimes it forgets the prompt at all

[grizzled loom]

sounds like a Microsoft support question

[fallow ore]

In ChatGPT you can say "run my prompt verbatim"

[low skiff]

i am thinking of switching from chatgpt plus to copilot plus but i am not sure which makes more sense

chat gpt i mostly get oinly one image

[fallow ore]

Use the DALL-E GPT https://chat.openai.com/g/g-2fkFE8rbu-dall-e

[low skiff]

I use that

[fallow ore]

You should get 2 images?

[low skiff]

rarely

even if i ask for 2 images

[late blade]

the model is the same, but using the chat experience is different between OAI and MS

and the content policy differs

you can do very dark stuff with copilot, but should you ever use the royalty emoji, the prompt gets blocked

not the case with OAI

[low skiff]

i also saw that copilot has no limits on messages

ok so copilot is in general more "open"

?

i hate the copyright bias

[late blade]

it's not a copyright bias, it's just that MS can take a punch or two on the legal side

[low skiff]

Is there any information about a DallE3 Standalone Webversion

[late blade]

nothing announced

[fallow ore]

Do you just mean using the API?

[late blade]

best would be to use the API as a similar usage

instead credits you just pay for tokens

or generations

[low skiff]

Yeah but API is more expensive isnt it?

[verbal sequoia]

Showcase your DALL·E creations to the world!

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

Quick tips for selection:

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

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

[low skiff]

compared to the monbthly plan

[late blade]

yeah

[low skiff]

especially if i just play around all the time

[fallow ore]

Depends how many images you make, and if you use ChatGPT for things other than images

[low skiff]

it can sum up quickly

i do use chatGPT for writing mainly but also thought about if it makes sense moving to copilot

Not really sure yet how i like the copilots better websearch

[late blade]

I use ChatGPT+ for personal stuff, Copilot Pro because I'm spoiled and Copilot Enterprise because of work

[fallow ore]

I would recommend you look into Poe . com

[low skiff]

lol

[low skiff]

Are you allowed to advertise other tools? 😄

[late blade]

other AI's usually go in #ai-discussions

[fallow ore]

It's made by one of the cofounders of OpenAI -- so kinda related

[low skiff]

Tbh i used to like midjourney but DallE is so much better in understanding. MJ is good when it comes to creative and reality lookalikes

[low skiff]

Got it

[late blade]

if we go by the strict rules it should still go to #ai-discussions, but I won't report you on this one 😉

[low skiff]

Anyone got a good prompt hack to really get real looking photos with dalle instead of photoshop style

[late blade]

prompt: person image and it worked?

[fallow ore]

So I must post it there and then link it here? It's relevant to this chat 🙈😅

[late blade]

I dunno, some people do that

[fallow ore]

Tell ChatGPT you like "prime lens, focus stacking aestehtic" and let it work that into the DALL-E prompts

[grizzled loom]

understanding and knowledge make for better prompt structure @low skiff You can create your own structured template to populate with values you want considered

[late blade]

On that prompt I proposed:

Prompt Image of person

Revised Prompt: Create a whimsical portrait of a person in a fantastical setting. The person is centered in the frame, wearing a vibrant, flowing robe with intricate patterns. Their hair is styled in an elaborate updo, adorned with sparkling jewels and flowers. Around them, the atmosphere is magical, with glowing orbs of light floating in the air and a background of lush, alien vegetation in various shades of green and purple. The lighting is soft and ethereal, highlighting the person's features and the textures of their robe.

I'm confused now...

[grizzled loom]

for example. "3D" or "Watercolor" or "anime" for style guidance. descriptives that set image tone such as "Bold" or "Dreary", verbs to drive the process further "envision" or "imagine" or "conceptualize", detailing methods "filagree edgning" Neon borders" etc etc etc

break it down as far as you want to

[low skiff]

Does copilot allow to "edit" uploaded images or is it like chatgpt just describing it new for dallE

[late blade]

edit as using the model to edit? no

edit as provide an user interface where you can edit images? yes

Image of person

I seriously don't know what is happening at the model level

[fallow ore]

Image of person, prime lens, focus stacking

[late blade]

Image of person Copilot version

[fallow ore]

Wow this is an awesome image!

[late blade]

I'm still wondering why I keep getting this kind of quality images from dall-e

it's just horrible quality

[low skiff]

Thats so painted 😄

hope DallE 3.5 fix that

[low skiff]

haha whats the prompt

[low skiff]

Wdym?

[late blade]

same prompt

[late blade]

when you open the image you have the option to customize the image with Designer

[late blade]

@low skiff

[low skiff]

only on pro i guess?

And what about uploading images and say add a hat or whatever?

It does not work right?

[late blade]

no, inpainting is not available

[fallow ore]

We're all waiting for this feature

[low skiff]

:d

Thats why i was asking for a stanbdalone dallE3

Even DallE2 had inpaintin and outpainting right?

[fallow ore]

No, DALL-E v3 doesn't support this. The old model used to

[late blade]

@low skiffclick on the prompt and you can do what I showed earlier

[low skiff]

well lets wait together xD

[low skiff]

no i cant on free version

i only have these buttons

[late blade]

ah that's your problem

[low skiff]

haha

[late blade]

you use German

lol

[grizzled loom]

I edit images with dalle @low skiff in my custom models.

[low skiff]

are you serious? hahaha

Its not pro but language related?

[late blade]

of course, I'm swiss, I'm allowed to say it

[low skiff]

Haha

[low skiff]

Well at least we are getting legal cannabis, what you got?

[fallow ore]

The company I work for is Swiss lol

[low skiff]

Usually its switzerland vs austria haha, germany is a lonefighter

[late blade]

what a coincidence, me too!

lol

[low skiff]

why is the OpenAI Discord in general so quiet

i rarely shwo up but its not very crowded

[late blade]

it's prob because everything is working fine atm

[grizzled loom]

this was through natural language and the model doing its own editing. i have al lthe driving python code it used for it

is it great? no, but i also told it almost nothing and it did fairly well

[low skiff]

lol

[late blade]

there was a problem with ChatGPT earlier today, it's like calling armies to yell and write and vent and what not

[grizzled loom]

my whiteboard scratch model does well if anyone wants to try it

simple and direct. https://chat.openai.com/g/g-ZJxp9MF80-reality-basedgpt

[late blade]

can you split quarks apart without creating two more quarks in the process?

reality check

[grizzled loom]

anyone else seeing that the edit pencil isnt always available?

[fallow ore]

In ChatGPT or Discord?

[grizzled loom]

chat

[fallow ore]

Did you upload an image?

[grizzled loom]

messages with images, and also any with links from output

[fallow ore]

Yeah, I've noticed that my messages where I've uploaded an image, I can't edit these. Haven't experienced an issue with links from outputs -- what do you mean specifically by this? I can test my side

[grizzled loom]

my new favorite is i get error in response and still get output including images

[late blade]

I've had that

[fallow ore]

#images-discussions message

Nope, apparently everything is working fine atm

[grizzled loom]

now i play this new game with myself like gambling where i decide if I REALLY want to hit regen or if i want to give it another minute or two.

[fallow ore]

This is really well made! Very methodical

[grizzled loom]

Thanks. I build some with pure language, others with mixed code and language, and yet others with "Pythonic natural language". I havea small tool ive built to convert a verbal prompt into this "pythonic natural language i speak of. it was meant to be both a prompt compression method as well as a code compliant language method.

[fallow ore]

Yeah, I see your code-like constructs, I do something similar in my system prompts! Very cool!

[late blade]

I hate the icon of the new outlook, that new label makes it so annoying

[dim cradle]

hi

nice potion bottles

[late blade]

poison bottle, that's what I read

[midnight olive]

uh no?

[midnight olive]

i meant like not standing in four

like BIPEDAL

as in standing in two legs

[dim cradle]

Blueberry fire

[grizzled loom]

you try. standard prompt. prompt: a 3d dragon stating "You're on blueberry fire!" https://chat.openai.com/g/g-YTxCoO1EH-geewizz

see where you get

[midnight olive]

Idk why but on phone I got timed out for using the draw command for bill cipher

[grizzled loom]

You're on blueberry today huh @dim cradle ?

[low skiff]

example?

[empty kelp]

a hyperrealistic head portrait photo of an athletic and diverse female elf. her hair resembles a complex hedge maze with a beautiful water fountain in the middle

i like your shrubbery in the hair concept

[dim cradle]

I’ve done this, it’s cool

[grizzled loom]

really love that tank paintjob pic @dim cradle I have a few custom pain guys who would eat that up for areal design

[dim cradle]

nice, i was thinking about a guy i used to work with who airbrushed bikes on the side, quit his tech job, opened his own shop -- i thought about sending a message to him about dall-e, he might actually find the advancements and this stage useful, for brainstorming, conceptualizations, etc.

[grizzled loom]

absolutely. ive gotten one of the two into AI and he's beginning his journey into it for a multitude of ideas

[dim cradle]

last i heard he had an R1

[grizzled loom]

motorcycles are difficult to get right

[low skiff]

well its not hyper realistic but the details are great

[forest dragon]

could dall-e make this 16:9 picture into 9:16 without stretching?

or which AI can do this?

[grizzled loom]

mmm yeahnahmaybe

[chilly onyx]

Not Dall-E currently. If you ask it to then it will create a brand new image like this

[forest dragon]

yea but i want my image 😦

which can do it?

[dim cradle]

patterns and anti-patterns

[forest dragon]

what is that

[chilly onyx]

Yeah, I know, it’s a common request, but it’s not currently something available with OpenAI tools and unless the mods have relaxed quite a bit on here we can’t direct you elsewhere on this channel

[midnight olive]

just cruious, is it the same for the #image-bot and copliot

[grizzled loom]

best i got out of my model thus far was this

[strong rose]

Wait till you see the next airbrush I generated. I'm waiting for the timer to finish 🫠

@swift oriole

[hollow sandal]

is there a way to use gpt 4 for free on this discord server?

[grizzled loom]

dalleboit channel

'/draw'

[hollow sandal]

can i also solve math equations?

which one?

[strong rose]

#image-bot

[swift oriole]

Cool airbrush technique theme let's ya go a bit wild.

[strong rose]

Yea man. I want to see what else DALL-E can come up with

I would be nearly impossible for a person to airbrush a fractal like this

[swift oriole]

Or airbrush a jet fight with fractal patterns lol.

[strong rose]

Whaaat? You're completely wild

[swift oriole]

Jet fighter that is

[strong rose]

BRB

[forest dragon]

thx

[swift oriole]

Fantastic lol #daily-theme message

[dim cradle]

Haha

[swift oriole]

@strong rose

[strong rose]

What have you generated lately?

[swift oriole]

I'll admit I like to still do prompts just to test out effects and arrangements in it nothing usually looks great but is interesting occasionally to for regular images haven't yet today though.

You're jet fighter looks great though

[strong rose]

Thanks. Do you do traditional or digital art?

[swift oriole]

Both, as a hobby though, fiddled with Bryce and Blender but almost rather watch real paint dry than struggle with them

[strong rose]

A person has to be really into 3D rendering in order to enjoy it. Do you have your traditional art online?

[swift oriole]

Nooo lol, it's not wurthee lol, it's just a hobby really

[strong rose]

I bet I'd like your art more than you do

Wanna know a secret about creating art?

[swift oriole]

Sure if it isn't Booby traped lol.

[strong rose]

Nah it isn't 🫠 . As the creator of the art, it's difficult to appreciate just how amazing it really is

I've drawn things that people are really excited about. Yet, after finishing it I thought it wasn't worthy

If you want to know how good your art really is, compare it to when you first started and measure how much progress you've made

[swift oriole]

Yes I've been told those things, you're correct though I've not been into the art past s hobby, though I do think AI image creation is fantastic, soon AI AV media with Sora

[strong rose]

Yea Sora is wonderfully hilarios

Did you watch "air head?"

[swift oriole]

Not yet just the early stuff elephant made of leaves an a kitten on Roomba

[midnight olive]

Is there like a thing that can tell me prompts for copilot/dall e so it can generate the photos I want

[strong rose]

Have you asked ChatGPT to write your prompts for you?

[midnight olive]

Huh, not sure

Does it actually work?

[spiral silo]

#image-bot

[earnest flame]

Mostly you can just ask it to generate photo-realistic images for you. I find that it helps to ask it to create a description of an image when asking it to generate stuff for you.

[empty kelp]

DALL-E doesn’t do any sort of math or solve equations. you need to use ChatGPT or the GPT part of the OpenAI API for that

or a Playground GPT on the OpenAI website

[late blade]

Also, GPT doesn’t work in Discord. You have to use either chat.openai.com or the API

[grizzled loom]

Man half my image calls are errored in any custom or the main. Half my calls on failed returns.

[hearty ether]

any tips for getting it to do an oil painting of a person?

It always seems rendered for me

[velvet rampart]

Ask that the brush work is visible and not too smooth which will reveal that the image is done AI. Also other qualifiers, such fit to be hang in Louvre or similar great. More simple, not too full of intricate details, aka keep details simple. Hopefully this helps.

[fair crescent]

I loved it, but I was surprised how much symbolism and messages they could put in.

[strong rose]

It's funny how the balloon had a different shape in every cut

[fair crescent]

Not too weird right? Each cut is a new video. It is like pictures. There's no 'permanent' features.

I think the true power is in adding scenes, not making a full video (yet)

[hollow sandal]

where do i find the api if you don't mind

[hollow sandal]

open ai api? where can i find that

[late blade]

https://platform.openai.com/

keep in mind it's a service different from ChatGPT+

api is only paid

[hollow sandal]

i have been wanting an answer for a question

is the gpt that copilot uses the same is the paid version from open ai?

and if so, why is it free? how is it free

[late blade]

the model yes, the implementation maybe

[hollow sandal]

like does it have the same math solving capabilites and code writing skills?

as gpt 4

[late blade]

gtp model capabilities are the same

keep in mind that you are working with a group of models, not just one model

[hollow sandal]

so is using microsoft copilot for math as good as gpt 4

[late blade]

should be if the models are the same

keep in mind copilot lets you choose model in their pro version

[fair crescent]

It is very difficult to say. As @late blade mentions, it is actually a group of models. To make it even more confusing, each model can be 'trained' to certain things.

This is why CoPilot will have different answers than ChatGPT. It should have the exact same data under it, has a model associated, but can be elicited to have different responses. Being obtuse with CoPilot will often result with CoPilot terminating the conversation, while ChatGPT tries to fix it.

At least how I understand it.

[late blade]

not only trained, but also combined and set different

for example within the GPT models you the temperature parameter can be different

MS is adding other models to Copilot for their mix, models that are to be discussed in #ai-discussions

but from Copilot Studio for example, you can see the settings of their GPT

[hollow sandal]

Is there a “hacky way” to get access to open ai gpt 4 without paying?

[late blade]

that to be honest is something the OpanAI discord in general can't answer

would be against rules

[hollow sandal]

Well is there?
Im not asking for a specific way

Is it possible?

[late blade]

to my knowledge, no

I'm not really into looking that also, so if there was a way, I'm not aware of it

[vapid elk]

convince a friend to pay for you

works specially well if your friend works for OpenAI

otherwise, no =P

[hollow sandal]

Do you work for open ai🫣🫣

[viral moth]

you can just use similar models to gpt-4 for free if you want only text

[hollow sandal]

Such as

[viral moth]

Mistral Large on Le Chat, Claude3 sonet, gemini pro on google ai studio (or you can ask for gemini 1.5 pro access) well its not good channel to talk about them but ok if you got more questions > #ai-discussions or #off-topic

[hollow sandal]

Ty bro

[viral moth]

np allways happy to help

[ionic dome]

start with thick dripping impasto and keep adding subjects, the impasto effect gradually fades until you hit your mark

...durp, seems i time travelled from yesterday

[rugged heron]

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

[viral moth]

Bing image generator

[rugged heron]

@viral moth Yep, but I'm just finding on by using Discord?

[viral moth]

You can only get 5 images a day on #image-bot so you using that or bing i don't know more solutions

[rugged heron]

@viral moth Thank you so much!

[late blade]

This is the wrong channel for this since this is a channel related to DALL-E as a model. For other AI's or even own implementations outside of the OpenAI ecosystem #ai-discussions or #off-topic is more suited

[open wind]

Hello, is there a way to continue to use dalle 2 after it’s taken out of labs, like is there a clone or something?

Really wish I could just keep giving them my money, I love dalle 2 for what it is , it’s going to be like owning a vintage car , let me live

[merry vault]

when sora?

[crude charm]

hello guys im new here, im trying to creat image with dall e 3 API but i dont know how to set it up, i try to find it on youtube and do some searching but all of them dont cover from the start that even ignorant poeple like me can do. anyone good enough can help me in this regard?

[late blade]

You should check https://platform.openai.com/docs/introduction

there's a nice guide on how to use the API in general there with code snippets you can use

keep in mind API and ChatGPT Plus are not the same service. (Often people confuse these two or mix up them, just a hint)

[crude charm]

thanks for your help buddy i will try it out once i have time

[late blade]

np, didn't do anything. just point

[quartz vale]

there isn't a clone currently, and we don't know what the future holds for Dall-E 2.0 right now 🫤

[late blade]

are there news for the dall-e 2 model api'

?

don't remember reading anything specific about it, maybe I missed some info

[viral moth]

Mayby they will change model in labs to dalle-3 someday hmmmm

[grizzled loom]

Or just work rigorously toward something better

[late blade]

I have something to report to you, and you should be interested

I used two of your GPTs in german, they worked really good.

[grizzled loom]

very nice. https://chat.openai.com/g/g-YTxCoO1EH-geewizz GeeWizz is pretty good overall. Which one did you work with

[late blade]

This one and TwinOracles

[grizzled loom]

very nice. twin oracles was updated again. Something was changed yesterday and it was broken. Let me know today if you play with it further

Yeah, GeeWizz does very well at many things overall.

[late blade]

I will check it again once I'm back home

I'm undercover on the field

[crude charm]

how do i fix this?

[plucky hare]

Maybe ask over in #dev-chat!

[crude charm]

i wil

[viral moth]

its here to ^^^

the api key ig

delete the ss ig

[crude charm]

done brother

[viral moth]

[velvet rampart]

Are other having problems with producing images with DALL-E customgpts? I can produce the first set of images in the chat. Then I want to adjust the image. The generation starts. Then the complaining starts that it cannot produce images due to system glitch.

[velvet rampart]

Absurdities on content policy/filters. I cannot generate van dyke style facial hair.

I can generate by describing the style but not by name.

[velvet rampart]

Current daily limit is 120 images according to the chatgpt. Includes all generated and failed due to various reasons.

[viral moth]

not 200 per day ?

i heard it somewhere that its 200

[plucky hare]

I don't think ChatGPT can reliably report on its own limits like this yet. I believe the last confirmation we got (though a little while ago) was 200 images a day, link to message from Moxi, one of the OpenAI devs: #images-discussions message

[velvet rampart]

And when I tried to ask it further. First 60. 120. No. 60. No, I do not know.

[plucky hare]

Yeah those are all hallucinations! ChatGPT doesn't have comprehensive self-knowledge currently

[velvet rampart]

That is bad. It should have excellent internal documentation.

Another is what I am fighting all the time are the absurdities of content filters. ”No, no, I cannot do that.” E.g. Cannot produce van dyke beard if mention by the name.

[plucky hare]

It has a lot in its system prompt, but you have to be careful not to overload those (and they're already very full!). The rest comes from how the knowledge cutoff works for all LLMs. The content policy is a separate issue too!

[empty kelp]

DALL-E translates everything to English before sending it to the model. it works really well most of the time, but occasionally the translations are mind bendingly wonky. like if you give it something in Hawaiian language for instance — it produces some things that are a little bit crazy. GPT 3.5 interprets languages a lot better than GPT 4, but i’ve been using a 3rd party translator to put things into English first

[hollow sandal]

can someone that has access to gpt 4 please input this to it
I know that the answer is c but I don't know the steps.

[crude charm]

how to use dall e discord bot ?

[weak hull]

While I can't add to this discussion from a personal standpoint, this is what GPT 4 spit out. Cheers.

[hollow sandal]

thanks a lot but gpt seems to have interpreted the question wrong. If you don't mind here is the question in text version if F(x) = (\int_0^{x}e^{e^{t}}dt)
what is (\int_1^2\frac{e^{x}}{x}dx)

[weak hull]

In progress

[hollow sandal]

thanks a lot I appreciate it!

[weak hull]

Sure thing. It didn't come up with (c) as the answer though.

[hollow sandal]

the question is kinda tricky

i managed to figure it out tho

i thought gpt 4 would get it right

but thanks a lot

[weak hull]

No problem

[verbal sequoia]

Showcase your DALL·E creations to the world!

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

Quick tips for selection:

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

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

[hollow sandal]

thanks! which ai is this?

a key part in solving the question is using the first integral I provided, intreseting how it just ignored it

[late blade]

GPT4

with custom functionality

[hexed umbra]

GPT-4 is definitely not the best with math. Ive found in a lot of cases it struggles a lot with just high school level calculus (even with external tools like Code Interpreter or Wolfram it cant get the more complicated questions).

[spiral reef]

Ayo, what do equations like this have to do with AI? I'm curious

[hexed umbra]

Not sure if the original math question asked was directly related to AI, but the topic the question is under, calculus, is used a lot in machine learning (gradient descent used in backpropogation, probability distributions etc. etc. are all very important parts of machine learning and models like GPT-4 would not exist without these techniques)

[spiral reef]

I see. What can just one equation accumulate?

[hexed umbra]

Well, there are a lot of different equations involved, each with varying amount of contributions to different things like the training of a model

[spiral reef]

Are you implying that just one equation could accumulate a vast amount contribution?

[hexed umbra]

Well, yeah. Some equations have more of an impact than others. Like the loss function is often represented as a single equation and this function is how we are optimising the entire model for a certain task. Next token prediction wouldn't work without this equation, in fact machine learning would probably be quite different if we didn't have this equation. But im not saying that you could make a large model with like just a dozen equations lol. There are a lot of different but important equations throughout.

And, i'm not entirely sure what you mean by "could accumulate a vast amount contribution", wdym specifically by accumulate?

[spiral reef]

If I had to be honest and a little personal, I am in fact diagnosed (This username isn't ironic), I lack intellect so I use vocabulary based on what I have a relative idea of, basically I don't really know what 'accumulate' specifically means.

Now, when I mean accumulate, I think I have a general understanding that an equation can expand when being solved, and give off differently variables (possibilities), is that right?

[hexed umbra]

Honestly intelligence is such a fascinating thing, but i do feel like there is ways you can improve your intelligence. When i was a child i was surrounded by people who were much smarter than I was, and i observed them. Intellect is very well correlated with knowledge (obviously theres more to it, but i feel like that is one way you can help improve your intelligence), and those smarter children were reading a lot, giving them a base knowledge to work with which helped them a lot throughout school and uni. The more knowledge you have, the more connections you can make, and the better you can understand new concepts. There are certainly biological/psychological influences here, but i do think you can always keep improving your intelligence.

But, an equation can be expanded or manipulated to solve for different variables, and in that sense, it can "accumulate" different possibilities. If i had an equation like X = 3y+6, there are a lot of different possibilities for what X could equal depending on what y equals. Like if y = 1 then X = 9, or if y=2 then X = 12. There are a lot of different possibilities there? Im assuming that sort of what you mean here?

When we say "accumulate", especially in machine learning, we usually mean like summing up over time. Accumulate also refers to a gradual increase or buildup.

[spiral reef]

Oh believe that there are ways to improve intelligence as I'll be the leading example of that, with assistance.

But I see, This is very interesting to be informed of. I wonder if one equation can result in models building upon itself, with the help of some aspect of human interference, to assist in ARE

[spiral reef]

What is the probability that an equation can assist in things such as ARE?

[hexed umbra]

Sorry, but, im not actually too familiar with what "ARE" stands for, could you provide a bit of context if possible?

[spiral reef]

Oh, so sorry, Advanced Robotics Engineering.

[hexed umbra]

Ahh ok

[hexed umbra]

I doubt a single equation would allow for models to build upon itself, current models are not smart enough for complete and autonomous self improvement. And i do not think they could completely design a robot to embody themselves in yet lol.

But the way we are training models, i do not think we technically need to improve upon their architecture. The main thing we've learned is that data and compute is all you need. If you want a smarter model than GPT-4, then you just need to train a model on a lot more data than GPT4 was trained on and pour in a bunch more compute. Architectural improvements are important, but in my eyes they are more important in terms of efficiency (and efficiency is important of course. But if you're looking at just making models smarter).

[spiral reef]

I would doubt their ability to achieve this on their own as well, hence "human interference".

But who knows, maybe a skeptical student I happen to encounter was right in his assumption that some of these equations can lead to nothing, even just simple curve of a line, though I don't entertain limitations too often, nor enable ideas of skepticism as well - as I swiftly reminded that student that as researchers, they should avoid idea's of skepticism, as it literally goes against the intrest of researching in the first place.

[agile peak]

Did OpenAI teams deleted one of sample from dalle3 showcase in official site?

Was trying to find the prompt for real photograph of camera lens surrounded by corals. It seems gone, is that explains why dalle-3 no longer able to capture real photograph?

[velvet rampart]

Use hyperrealistic, hyperdetailed, photorealistic, 64k definition as part of the prompt. I get really great pics with these.

Prompt:
“A hyperrealistic and hyperdetailed image of a small anthropomorphic mouse standing upright. The mouse has a combination of human-like features and mouselike characteristics. He is wearing a perfectly fitted plaid shirt, snug overalls, and a pair of sturdy boots, with a wide-brimmed hat on his head. The setting is a farm, with fields and a barn in the background. The image is in 64k definition, capturing every minute detail of the mouse's fur, clothing textures, and the surrounding farm environment.”

[dim cradle]

Not bad but 100K is even better

[velvet rampart]

The 64k is the highest DALL-E gave to me when I asked the definition range from the lowest to the highest.

[crude charm]

anyone can help me why my image quality keep decreasing everytime i make an image?

it was 3.01MB yesterday

[viral moth]

size is not allways matter of image quality as it can be a file that was with the same qulaity but better compressed you may try just check the pixel count ig its better for that task

[crude charm]

both are same png file 1024 1024

how to turn it back to 3mb? do you know how

cant understand why it different

[viral moth]

well no need in my opinion

size is just size

it not depend on image quality

[crude charm]

what if i want to upscale it?

[viral moth]

then you can

you can use many upscalers some are on internet or get some from github

[crude charm]

i have gigapixel

but size doesnt matter for upscaling?

sorry i dont really understand

[viral moth]

it doesnt metter

[crude charm]

one thing matter the most is the dimesion right?

[viral moth]

size depends on commpresion "Compression is done by a program that uses functions or an algorithm to effectively discover how to reduce the size of the data" and that are algorithms that make a image size smaller witchout any quality loss

[viral moth]

yes

the pixels x pixels

[crude charm]

thanks

[viral moth]

np

[crude charm]

one more thing buddy, how to you use dall e bot in this discord?

[viral moth]

. /draw in #image-bot with a slash command

[crude charm]

does it use any credit?

where it send the picture to?

[viral moth]

its all free you got 5 images per day

[viral moth]

to the channel

#image-bot

[crude charm]

ohh thanks for the information

[viral moth]

np

[ember orbit]

GN

[hearty ether]

Did they change the model?

It feels so weird now

[velvet rampart]

How?

[hearty ether]

I asked for an Illustration and it does this weird rendered scene

[velvet rampart]

Did you specify style or technique? If not specified, quite random stuff can come out. Sometimes random stuff just happens. I usually feed the generated picture back, as the model cannot see produced images, and comment/point out what went wrong. Usually this works.

The picture you gave can be interpreted as illustration unless you specified it further.

[hearty ether]

Guess I’ll stick with the same seed

Bonus is that it does keep the same face

It’s hard to describe this style

It’s a mix

[velvet rampart]

You can always feed the picture to the gpt and ask ”describe this image’s style”

I got realism, fantasy, digital painting, chiaroscuro, and atmospheric as the styles.

[ember orbit]

GM

How i can add my collection to the Super rare marketplace?

[crude charm]

Anyone can help me how to get image generation id from api call?

[dim cradle]

i thought it was in the response

[velvet rampart]

Have they decreased the maximum number of images per day? It feels that I have generated way less than 200 today. 3rd day in a row that I hit the maximum.

[hexed minnow]

Request to be able now to to Delete Dall-e 2 images: A lot of the Dall-e images people produced were trash, and by this point you have likely all the training data or selection of images you could want or need from that. Could you now allow users to begin permanently deleting unwanted images?

[strong rose]

Does the new DALL-E exclude registered trademarks when generating art?

[weak hull]

Is it just me, or has DALL-E been lethargic the past couple of days?

[vital gull]

meaning?

[weak hull]

Very slow to generate images. Multiple error messages in a row prior to a successful image generation. Overall decreased performance as compared to previous experience. Via GPT 4.

[vital gull]

there hasn't been any major updates afaik

[velvet rampart]

Depends what you generate and whether you DALL-E directly or customGPT. Simpler stuff directly with DALL-E are fast, but complex stuff with customGPT takes time.

[graceful spade]

Who has a huge list of styles for Dall-E?

[plucky hare]

The #1108740112558325790 channel is a great resource for styles!

[velvet rampart]

Just ask it to list them and explain each. Works just fine.

[last kindle]

is dall e 3 free of cost?

[spiral reef]

Well, Dall-E completely tookout my reqest to have Danny Carey mounted centerstage as a cybernetic torso amount a drumset of mechanisms and pipes. I'm most definitely using it wrong in some way though

[deft musk]

Can you completely reword that?

People's names (and anything copywrite) can get rejected automatically.

'mounted' can have meanings that are nsfw, so a different synonym, and ideally common English language phrasing, is likely to work best.

'as a cybernetic torso amount a drumset' is not a common way to use words.

Did you perhaps mean something like:

"Danny Carey seated centerstage with a cybernetic torso enhanced with mechanisms and pipes"

The name may still get a rejection. You can try to describe the human to get a similar appearance though.

"A middle aged, heavyset long-haired blond drummer, famous for using custom wood-tip drumsticks and known for very diverse sounds and dynamics seated centerstage with a cybernetic torso enhanced with mechanisms and pipes."

[deft musk]

#image-bot message no idea if that's even slightly similar to what you're hoping for.

[spiral reef]

I'm jotting this first part all down

As for the reqest, no, not even in the slightest. But, that's ok, today was a day of failure to understand how minor details and more articulate context will lack result if it's not even close to property executed

Actually, ithink Dall-E made what I actually was hoping it wouldn't generate, which is kinda interesting

[deft musk]

Yeah! The better we understand what we want, and a language to phrase the wishes in, the more accurate the AI outputs will usually be. I'm sorry I don't understand your request better, what I asked for might actually be really off of what you wanted, I was just confused.

[spiral reef]

On the contrary, it's pretty close and looks pretty damn great, but man oh man, the stuff I want to create is only fathomable in mind right now, trying to illustrate this is near impossible when you are severly undereducated, I feel a simple read of the dictionary will possibly clear up most of the confusion for myself, maybe a thesaurus too

[deft musk]

ooh, cool.

Y'know, the model's a really good dictionary/thesaurus equivalent too.

One thing you can do, if you want, is discuss with the model what you're interested in, and see what wordings it suggests. You can talk over (literally, if you use the app and Voice) what you do and don't want, refine ideas, and get help with expressing them. If you're interested, but the model can help you learn a bit when you interact, and can even make it be all about your interests

[deft musk]

Like, one way to do that, and there's so many:

Reinterpret this, what's the person probably asking for an image of? Focus on what they probably wanted, then offer 5 rewordings of that likely to be acceptable to content filters. https://chat.openai.com/share/b5143606-0d22-4a1f-8c6c-d85e441ea376

Now, the model may or may not have understood perfectly, that's where the discussion comes in

[vapid elk]

nice

[spiral reef]

Dall-E prompted I shouldn't try to replicate other characters, but it seems like everyone else has it down

[fiery salmon]

ofc Dall-E 3 is good at fictional characters

sometimes it does artist

[spiral reef]

Please recreate an illustration of 'Alita' from 'Alita: Battle Angle'. She should be themed in UCLA aesthetics. The artsyle is that of 'Yukito Kishiro'. The setting should be that of an arena, figues of a cheering crowd surround a side shot of her as she makes her walk out.

If that's so, let's see if it actually works.

Yeah nope, that is the furthest it can get from what I requested

[weak hull]

I didn't run this as I have stuff cooking, but try something a bit more indirect that leads the generator:

---

Create an image of a young adult cyborg at a college campus. She was fashioned by her creator to resemble his lost daughter and to eventually go on to compete in dangerous competitions against other cyborgs. This angel is imagined using a UCLA aesthetic and is meticulously crafted in a style akin to Yukito Kishiro.

[fiery salmon]

2B

[spiral reef]

Let's entertain this and see if it works because I'm done spamming, and even further done with Dall-E

[fiery salmon]

Chun-Li vs Lara Croft

[weak hull]

It's not a photocopier

[spiral reef]

Oh wait I have no more tries, so Dall-E still hasn't generated anything close to what I want

Especially those last two

[fiery salmon]

Can you give me an idea and I could make it?

[spiral reef]

Well, I don't know. It kinda seems like what I want is too complicated

Maybe just for now though

[fiery salmon]

I can work around it

[spiral reef]

Yeah maybe another time, this thing killed any confidence I can't lie

[rigid snow]

Will Dall E 3 get edit and variations endpoints

[vapid elk]

probably yes, I guess

but it might still be something that will take some time

[glossy scroll]

Ey, ty

[glossy scroll]

My future gf is gonna look like this

[mystic epoch]

Future wife

[hearty ether]

half of it is gen id tbh

you find a good one and it really stablizes the style and the characters