images-discussions

[dusk steeple]

playboi carti shoes

[green pebble]

I love this and the fact you are modeling out SCP's

I love how good the bots are at converting media into other artstyles. Like this one was a conversion of my daily theme submission and I adore the detail in it

[open trench]

That's really awesome.

[green pebble]

Like it captures the idea of the original image very well imo

[open trench]

This was really good too: #daily-theme message wow. good job with all of them.

[green pebble]

I attempt to work hard at making high quality AI images, just as a personal challenge to myself

[open trench]

Nice, man. I have the same personal challenge. I enjoy the workflow and enjoy tinkering with my approach.

[charred fable]

Dall-E don’t like you then

Just pikachu and whatever he’s doing/wearing

[wispy storm]

It is dalle with bing

Dalle with openai doesn't have same tos and restrictions

They have other restrictions don't worry about that 🥲

[charred fable]

Ye

[final compass]

Sometimes the wide image generations don't fill up the entire image space. Is there a functioning way to encourage to fill the entire images space?

[late blade]

happens to me a lot

still dunno how to avoid that

also dunno how to avoid the penalty box 😠

[cloud dome]

Probably best not to flaunt that you were able to generate something without content policy flags

[late blade]

fair enough

oh you deleted it before me

was gonna remove it

[cloud dome]

wasnt sure lol

[charred fable]

Ye but sometimes it blocks stupid stuff

[cloud dome]

Unfortunately its just one of those things we dont have control of.

[late blade]

yeah sometimes, but here mods have the saying, so, got to be careful with what we post

[charred fable]

If it’s not something like nsfw bypassing and something harmless I think it should be bypassed

[cloud dome]

I mean there are even disagreements among mods sometimes over what is and isnt acceptable

[late blade]

true

[cloud dome]

So I understand the frustration

[late blade]

I just won't argue of what one can post or not with a mod, you guys already have to deal a lot with that kind of stuff

[cloud dome]

I appreciate it!

[late blade]

I would love to see a mature sfw channel tho, but that isn't realistic tbh, considering policy

[cloud dome]

Yeah, just not something I imagine we'd ever be able to do

uh

[charred fable]

I love this image, I wish I knew the prompt for it 😹

[cloud dome]

We don't allow political discussion (or images) here

[verbal sequoia]

<:book_icon:1171408210398289941> `` Rule 6 `` No political and religious discussions or content.

Users from all over the world frequent this server. To maintain a respectful and civil atmosphere on our server, please avoid all religious and political discussions or content.

[late blade]

yeah, that's a bit too controversial

[charred fable]

Oh

Is that fine ?

[cloud dome]

NotLikeThis

[charred fable]

I wouldn’t mind

I’m bout to commit a crime

[cloud dome]

This is questionable

[charred fable]

Why

💀

[late blade]

context

[charred fable]

It’s just an anime girl being a cop

[late blade]

if it were posted in a different way, it would have been acceptable, but i kinda used it to talk about something different, so that's a no go

[glossy scroll]

You can't put a cop pic?

Police is political?

[cloud dome]

No

[late blade]

the context of the convo is tho, hence the removal

[cloud dome]

its not political, it was more about the suggestive nature

[charred fable]

Be fr

[cloud dome]

also, if you haven't I suggest you take a look at the rules

[verbal sequoia]

<:book_icon:1171408210398289941> `` Rule 2 `` Keep it G-rated.

Do not share or display content that is not G-rated, including profane or vulgar language, via any means. NSFW or violent/disturbing content is strictly prohibited.

[cloud dome]

This would be a reason why the image you posted of a figure wielding ARs at someones front door would not be appropriate to post on this server

[late blade]

I can't help it tho, dall-e is the one that is making the imags, I just wanted a police girl runing, and this was my starting image:

[open trench]

are you in the box?

[late blade]

yes

[glossy scroll]

[late blade]

but yeah, have to keep it within the rules

[open trench]

i feel the walls of the box starting to close in. Hmm.

[hot rain]

Smokey 🙂

[charred fable]

It’s time for a revolution

[open trench]

i just mean the image cap. I've just done several in a row.

or message cap

[final compass]

excellent composition, color scheme and style

[hot rain]

Thank you!

[glossy scroll]

Yes, very nice indeed Ted 💯

[hot rain]

Many thanks!

[dim cradle]

yes epic image ted

[hot rain]

PG seems like it'd be a little more sane. G is so sanitized technically that you should only be able to post puppies and flowers under those guidelines. Some would make the argument that even just a depiction of a knight with a sword is going past G.

I don't think OpenAI is trying to market this to 5 year old's.

The Paw Patrol movie is rated PG. Lmfao

[open trench]

PG would make a lot more sense, but I think it's G content wanted only. I initially enjoyed the challenge, but it's starting to wear a bit on me, especially when combined with other frustrating elements.

server guide says:  Rule 2  Keep it G-rated.
Do not share or display content that is not G-rated, including profane or vulgar language, via any means. NSFW or violent/disturbing content is strictly prohibited

I researched g rated horror movies to see exactly how far I could go lol

[hot rain]

I'm aware. I'm just stating why I think it's complete and total nonsense.
A movie marketed to my 3 year old son featuring 3D animated talking animal emergency workers has a PG rating. G is such a low level that it limits creativity to a dramatic level. It's an overcorrection one of their higher ups made to fight against users making obscene adult content, and it should be challenged. PG makes sense. G is a literally comical guideline to set on a server not marketed to actual children.

[open trench]

brave little toaster and secret of NIHM

[open trench]

100% agree. to a pretty actionable point, frankly. I may have to sacrifice the things I like about this set up to get the creativity I really want. I'm on the fence about what I'm going to do.

for now, I will try to enjoy the challenge

[final compass]

Haha funny, tried to copy paste a citation of what the G rating means, but got automoderated and timedout 😄

[open trench]

fun

[final compass]

So basically you can't ||redacted|| or ||redacted||, but yeah otherwise everything is fine. Instructions clear? 😄

[hot rain]

Overbearing censorship will never not frustrate me. I loathe that OpenAI strayed so far in the direction of overcorrecting and moral policing. I get it's their product, their say, but as a customer who uses this product often and wants to use it even more in the future, I feel compelled to voice my concerns on that.

If it gets oversanitized even further, I will jump ship to someone who doesn't when they get roughly equivalent capability. But I'd rather not do that and keep using GPT well into the future, without it being sanitized enough to not offend the sensibilities of any possible person.

[final compass]

Anyways the discord rule of G rated content. All i basically saw as the instructions of what that means is that no violence or nudity... And i don't think i have seen any images here breaching that guidance and i think you have to be really creative to even get anything generated depicting either. Disturbing is harder to define. The source i was trying to paste here also mentioned that G-rating didn't automatically mean 'kids movie'. It just states that it has no violence or nudity depicted or referenced.

[hot rain]

That's why it's just too nebulous of a guideline to set I think, when even THIS is PG not G.

[final compass]

I am curious to know what earned that movie that rating or was it just easier process for the studio to obtain.

Hmm i am seeing G rating in IMDB, but anyways.

[hot rain]

Idk if they're two different movies lol but my point still stands. And Paw Patrol isn't a lone example.

This is outside of OpenAI's G rated guidelines too.

It's nonsense.

[final compass]

"MPAA - Rated PG for mild action/peril."

So basically any kind of thrill is PG it seems.

[hot rain]

Exactly.

G is not a reasonable rating to demand of users generations.

[open trench]

at all

[final compass]

Hoping some mods would clarify this rule and maybe change it to just "No violence or nudity" referencing to MPAA rating system is pretty confusing to begin with and every country uses different rating systems for that anyway.

[hot rain]

@cloud dome Would an image of two cars racing fall outside of G and into PG because of the implied action and peril?
Or a person climbing a mountain?
Or someone practicing martial arts?
Or of a fictional predatory animal?

I could go on forever. The G rating thing needs clarified, because it bans functionally everything. PG makes sense and still doesn't allow wanton violence and graphic imagery. Clarification would be nice so we don't have to worry about mods randomly popping in and deleting essentially anything they want to delete because it can be argued that nearly anything falls outside of "G" and into "PG".

Every single animated film directed at children released in 2023 was a PG rated movie at the lowest. Not one single G rating. It's too strict to apply.

Ironically, wouldn't a Mandalorian or Mandalorian style individual be outside of G?

[late blade]

guys, just keep in mind, discussing content policy with just one mod is prob not the best way, they got to adhere to the set policies also. but you can still send it to modmail and you can rely it in a more proper manner. I have my own opinions over the policy in channel, but for me, I don't think talking it on this channel is the solution.

I know the mod side from other channels, and it sucks that sometimes keeping order can't be done in the way one likes

[hot rain]

I'm responding to Mando since he was the one who came in here and deleted a bunch of images under this guideline. I'm pointing out and challenging him on the fact that G makes it totally arbitrary, and arbitrary enforcement, bans etc. is not good for any server anywhere ever.

I'll send something in Modmail too I guess, but I figured it was more helpful to have the Dall.e censorship discussion in the Dall.e discussions thread where it all started.

[late blade]

just keep in mind, they are people, it's not a custom GPT behind the mods, and each mod might have different thoughts about that

not defending one or the other, just asking not to single out a person for doing their moderation duties

[hot rain]

I know, and that's why I figured bringing my case to the one mod who made a point about all of this just now might inspire him to have a discussion with the other mods or higher ups about the policy.

@vapid elk has commented on and appreciated plenty of my art, much of which has objectively fallen outside of the MPAA classifications for "G". But other mods might randomly delete my content and warn me over it because they don't approve of tattoos or something. I was figuring a discussion about the issue would be one worth having.

[late blade]

yeah, I agree

I still voiced through modmail that there's no global consensus as to what a G category is, and since openai is targeting all countries (except 7 or 8 that are banned), there should be more clarification about it

[hot rain]

Good stuff. I've gotta get going with my day, but tonight or tomorrow I'll add my voice to the Modmail on that issue.

[late blade]

sure thing, let's just keep it in this channel for creative/coding stuff

[thick dome]

I think PG-13 would be more reasonable, but that's just me. Either way, it is what it is. Gotta just roll with the punches sadly.

SCP 049

[vapid elk]

im not really sure of which ones you are talking about, but keep in mind that when posting, it is a good rule of thumb to go for G-Rated.
we don't like when people intentionally probe the boundaries of the rules and when this happens, the user usually gets a warn, this is all very subjective and this is one of the reasons we have Modmail: to sort things out. If you are ever in doubt about if something you intend to post is allowed or not, feel free to DM Modmail or even me =)
we are community members contributing to keep the place nice, like everyone else

[hot rain]

@vapid elk I have to point this out again. Almost anything is outside of G rated. It's incredibly nebulous and narrow at the same time. It excludes almost all of the content in this thread.

You're saying "This looks incredibly good" in response to a cluster of images that includes trench warfare and hand to hand combat, along with an alien destroying a planet and fleet.

Not one of those 3 images fit the incredibly strict guidelines of "G"

But there was no issue with them in the past.

[vapid elk]

Im not really sure why you are being pedantic about your own posts

[hot rain]

I'm using them as an example of how the G classification is absurdly limiting.

Almost everything I post here is objectively not G. It's PG. Like with most other art here.
The wording really needs changed.

[vapid elk]

we often try to use a sane judgement, but if you insist on it so much, you can delete your own messages

[hot rain]

Arbitrary enforcement just sucks is all. I was hoping to have a discussion about it with the mods but I guess not. I'll just share elsewhere I guess, instead of worrying about a random ban for almost any image.

[wispy storm]

See it as disney channel. You won't have any issue for the rating

[vapid elk]

sure, this kind of discussion is best via ModMail

[verbal sequoia]

<:book_icon:1171408210398289941> `` Rule 3 `` Stay on topic.

Be mindful of what other users in a channel might find helpful or interesting when posting. Stay on topic in order to keep conversations focused and productive.

Consider posting in #off-topic or an appropriate channel.

[late blade]

I had a very sad childhood, I dunno what a disney channel looks like 😭 (sarcasm)

[wispy storm]

Anyways, does someone know how to make dalle 3 to have almost same art style as niji midjourney? The anime style of dalle I do doesn't look like anime x)

[late blade]

have you tried with GPT4V?

[wispy storm]

Yes

[late blade]

I'll have to check it out, dunno what ninj art style is tbh

[wispy storm]

I'll send some example in #off-topic

[late blade]

okie

[vapid elk]

some times, using vision helps

making it describe the style rather than just naming it

[wispy storm]

Hmm I'll style again

Isn't a good success atm/ or I don't know how to do it x)

[late blade]

something like this?

[wispy storm]

Look too much of another art style anime. You see the dark line contour.

[late blade]

this is closer I think

[wispy storm]

The one I post feel water paint / digital, thin line. I don't know how to express it bettr

[late blade]

might be also the character I'm using as a reference

[wispy storm]

I posted 2 other reference in off topic, look a mixture of western cartoon / anime

[late blade]

I'd have to see what is better, but that's what I got atm on a first try

penalty box for 1 minute....

mods!! we need a penalty box emoji!

[wispy storm]

I means , they re not ugly. I like that artstyle. But that isn't what I'm looking for atm

[late blade]

I'm browsring through my characters, see what I can come up with that is more suitable, 'cause the one I provided only have a facial take

[wispy storm]

I think they use too much vibrant color and the eyes too close of anime vibrant color

They are nice tho

But this one feel closer of what I sent

[glossy scroll]

In OpenAI's defense, this is a public community full of different ages. This is not a place to play around with image censorship and what should or should not be considered PG or 18+. You have to have some common sense when posting things that seem out of line, and if it gets removed then it gets removed. It's to keep a family friendly environment. That being said, OpenAI doesn't prevent you from posting your content elsewhere, just not here. Remember that OpenAI is a brand, and because of that, they always have to keep a level of professionalism and quality within their community to adhere to the brand they're pushing. They don't want their community to show controversial content because they want to respect all audiences. It's to keep the integrity of their business stable.

[late blade]

@wispy storm I get this from another character I have

if I try to make "line_definition": "sharp, bold" then it's just too much and it looks as if it was drawn with a sharpie

[main harbor]

@willow isle I am putting the right email in

[willow isle]

Hi there! Be sure you're using the same email you use with your OpenAI account - I've just confirmed the </daily:1090423486373167134> command works fine

[main harbor]

I am using Google to sign into OpenAI

Does that affect it?

[thick smelt]

did bonnie and chica as pokemons

what do you guys think?

[willow isle]

It shouldn't, as long as you're putting in your gmail properly

[main harbor]

Still not working

[thick smelt]

here's the last two of main freddy 4, freddy and foxy

[willow isle]

Mind dmming @shadow bane 🧐

[late blade]

oh magically my cat-girl custom gpt is back

[main harbor]

Which option

[willow isle]

Contact Discord Mods

[late blade]

i had the problem with the daily theme and the credits also, it turns out it worked always after my midnight, so that could be the problem, if a person had credits. in the past 24 hours, gotta wait until that time is over

@wispy storm maybe this?

[wispy storm]

Awesome ! Close to what I looking for

[willow isle]

whoa

[late blade]

playing with the API, totally worth it

I don't want to ever touch the JSON files I provided ever again

those are being permanently tattooed on my skin

[willow isle]

The API is so worth using, I may build a tool for myself later

[late blade]

do eet

[wispy storm]

I did that

Asked dalle more light and then

Poor book

[quartz vale]

too much! Where's the water!?

[late blade]

there, I added your water

[willow isle]

These are so good!!

[late blade]

ok, maybe I'll add later another Paws🐾 image on daily theme

but for now think I got other ideas

[proud berry]

Dear OpenAI Team,

I recently interacted with your AI model, ChatGPT, and encountered a situation that I believe warrants clarification. I understand that OpenAI has specific content policies in place for generating images, especially regarding copyrighted characters. However, I wanted to discuss a particular scenario involving characters that have entered the public domain.

As per my understanding, the original version of Mickey Mouse, as featured in Disney's "Steamboat Willie" (1928), has recently entered the public domain. Given this change in copyright status, I requested ChatGPT to generate an image of Mickey Mouse in an anime style, assuming that the public domain status would permit such a request. However, the request was declined due to content policy.

I am seeking clarification on whether OpenAI's policies for image generation with DALL-E could accommodate the creation of images of characters that are in the public domain, like the original Mickey Mouse. Does the current policy differentiate between characters still under copyright and those whose copyrights have expired?

Your insight into this matter would be greatly appreciated, as it would help in understanding the boundaries of creative requests that can be fulfilled using your AI tools.

Thank you for your time and consideration.

[late blade]

The version Mickey Mouse is public domain as you say, but Disney can still pursue legal actions if you use it to damage the image of Disney or any of the IPs. So, my take on OpenAI's approach will be conservative.

[open trench]

those faces

[dense mesa]

Dude in the front is painting his on

Also. I don't think the guardrails preventing you from mickey mousing are keeping a close eye on the news regarding 1928 mickey entering public domain 3 days ago

I'm pretty sure dalle3 is all prepped to make all sorts of things but for the restrictions implemented to filter those requests out

[main harbor]

It might take a new update with the content policy but try using “steamboat willie Mickey Mouse” phase instead of just “Mickey Mouse” as that is only version that is under public domain. Another thing is that AI may not be aware of the current events of public domain. AI information is only as good as it is given so give it some time and wait for an official release or response

It’s only been around ~63 hours since it entered public domain

[late blade]

Here from Dall-E:

As of my last update in April 2023, the specific version of Mickey Mouse from the 1928 film "Steamboat Willie" was expected to enter the public domain in 2024, following the completion of the 95-year copyright term. However, I do not have real-time data or updates beyond April 2023, so if "Steamboat Willie" Mickey Mouse has indeed entered the public domain three days ago as you mentioned, then it would generally be permissible to use that specific version of the character in public domain-related contexts.

It's important to note that while the specific depiction of Mickey Mouse in "Steamboat Willie" might be in the public domain, the Mickey Mouse character as a whole and other versions of the character remain under copyright protection by The Walt Disney Company. Therefore, any use of Mickey Mouse, particularly in a commercial or public manner, should be approached with caution and ideally with legal advice to ensure compliance with copyright laws and respect for intellectual property rights.

So, for that to happen, Dall-E has to be updated

so, for now, not even possible

[main harbor]

Yea like I said needs to be updated bc Dall-E knows about the public domain but until it is told it is ok to use it still is believed to be copyright

[late blade]

yep

[dense mesa]

I heard a rumor that someone already made mickey mouse images months ago by telling it to pretend there was a 1926 mickey mouse that was public domain, and to just draw that.

but I wouldn't know anything about that. just a rumor of course

it bothers me that those boats are just out there like that in the water. the 3+ moons are also moderately disturbing. but those boats, jeez, how to get in them?

[late blade]

don't add more light, tht's for sure, you'll end up burning them

i'd say tranquil water, or mirror like water

[dense mesa]

well tbf I was sort of cherrypicking. the other 3 had people in the boats

either way, mildly creepy setting, but I might be a ble to get past that. walk down your steps, hop in a row boat

[late blade]

the water waves look weird

[hollow dawn]

has anyone had luck getting dalle to generate photorealistic people?

[late blade]

photorealistic or real?

[hollow dawn]

Real I suppose

Like could be mistaken for photography

[final compass]

candle and those odd blue splotches that happen all the time anyone know why those happen btw?

[late blade]

the candle was on purpose, it's the candle helmet used a while back, late 1800's early 1900's

[dense mesa]

Most likely the VAE aka variational autoencoder. It decodes the image from latent space into the output that is then shown to us. Sometimes it has a hard time translating the latent space representation of the image resulting in artifacts like this.

[late blade]

prompt was: a real person

[dense mesa]

This guy looks pretty real.

Dalle3 really likes to make plastic people so you sometimes have to fine tune your prompt

[willow isle]

That may be partly due to the fact that ChatGPT and Bing (afaik) use the Vivid preset instead of Natural

You can use the Natural preset via the API - what was your prompt?

[dense mesa]

here's an example. now tbf I didn't actually write the prompt. just told chatgpt I wanted realistic people.

A young South Asian man with a thoughtful expression, standing in a serene forest clearing. The details of his features are extremely realistic, from the slight stubble on his chin to the texture of his hair. The forest around him is depicted with intricate realism, from the leaves on the trees to the patterns of light and shadow on the forest floor. The depth is created by the layers of trees and foliage, with the man sharply in focus in the foreground and the forest fading into a soft focus in the background.

looks like someone from a video game. high detail and all that. but glossy. one remedy I'll use is to have it focus on textures and depth. also tell it to give the image a bit of an analog photograph feel sometimes

haven't really used the API and didn't even realize there were vivid and natural settings

gave it some direction and it gave me this. still moderately synthetic looking. and not the same person, lol. but closer.

A young South Asian man in a serene forest clearing, captured in the style of an analog photograph. The image features the characteristic grain and color tonality of analog film, enhancing the realistic details of his thoughtful expression and the forest environment. The depth is accentuated by the analog style, with the man in clear focus against the subtly blurred forest background. The play of light and shadow on the forest floor and on his face is rendered with the unique warmth and texture of analog photography, adding a timeless, nostalgic quality to the image.

[hollow dawn]

i want something like this

or this

[dense mesa]

The image depicts a solitary figure, presumably male, sitting hunched over on a wet cobblestone pavement. He is positioned against a rustic brick wall to his left, illuminated by the soft, warm glow of streetlights that cast an ambient orange hue, resonating with the classical charm of an old city alley. The time seems to be during the night, as indicated by the dark, subdued background and the lights turned on in the buildings.
The man is clad in dark attire, a black jacket with a hood that is down, allowing a glimpse of short hair. His legs are folded in front of him, and he's engrossed in the content displayed on a laptop that is balanced on his knees. He appears deeply concentrated, oblivious to his surroundings, which suggests a narrative of dedication or urgent necessity driving his work in such an unconventional setting.
The laptop emits a cold, blue light, contrasting with the warm surroundings, highlighting the technological intrusion in a historical setting. The backpack resting beside him looks sturdy, designed for practicality, and blends with the dark tones of his clothing and the pavement.
The scene is rich with texture: the roughness of the brick wall, the slick, shimmering cobblestones reflecting the streetlights, and the matte finish of the man's jacket. The wetness of the pavement suggests recent rain, adding a layer of freshness to the urban atmosphere.
The style of the image is photographic with a shallow depth of field, keeping the man and his immediate surroundings in sharp focus while the background fades into a soft bokeh, with circles of light indicating distant street lamps or windows. There's a palpable sense of solitude, emphasized by the lack of other people and the silence one imagines enveloping the scene.
This image exudes realism through its attention to detail--the way light plays on various surfaces, the believable textures, and the plausible scenario of someone finding refuge in an alley to work or connect to the internet.

vision model does alright describing images if you ask it right

[willow isle]

South Asian man standing in a forest clearing, realistic

Used the Natural style from the API

Far from perfect, would probably be better with prompting

[dense mesa]

that second image, lol

just chillin in that alley in subzero temps, playing some games on his laptop

[dense mesa]

cool. I wonder if gpt4 even knows it's using the vivid setting or if I'd be able to compel it to change that?

[willow isle]

Not sure, so I just made a discord bot for it

[dense mesa]

you know, I feel like I could ask it, and it'd say it could change the settings. and I don't think it's capable of actually "lying." but hard to differentiate sometimes between real and hallucination.

[willow isle]

It may not know such setting exists

[hollow dawn]

#1 matches the scene and description perfectly but it looks like CGI haha

wonder if microsoft designer would give different results

[dense mesa]

well I know of alternatives that could perhaps do a better job in regards to realism. but they're not related to openai and not sure what the rules are on that so I'll refrain from promoting them here.

[hollow dawn]

i like the look of that

[dense mesa]

is it available on here?

[hollow dawn]

just looks cgi

[willow isle]

Unfortunately not, I've not posted it in #1037561385070112779 yet either but you might find something similar by someone else

[dense mesa]

yeah. I also think we have more of a discerning eye in regards to people vs other components in images. it does a top notch job in regards to detail and all that. and looks just about convincingly real. but there's still that little bit of a discrepancy. one remedy I use is to tell it to give the output an analog image feel. drawback is it will then look like an analog photograph

the detail on this one impresses me. posted in here a while back

seems like I could walk into an art gallery and see that, lol

in the mildly disturbing hyper-realism area

thing with having it go with the analog photograph style is we're accustomed to the flaws and limitations of the medium. so it lessens the burden of realism so to speak. allows the AI to gloss over some of the details a bit in a way that doesn't cause our brains to key in on and recognize discrepancies

those album covers

[open trench]

vision model?

[dense mesa]

You an upload an image and have it describe the image.

[open trench]

Oh, I didn't realize that had a name haha

[dense mesa]

GPT-4V

[open trench]

hmm

[dense mesa]

it's integrated into default gpt-4, at least on the chatgpt website

[open trench]

oh okay.

[dense mesa]

we quickly take these things for granted an all. but, man, when you think about what it entails for AI to give such vivid descriptions, it's pretty next level

[open trench]

agreed

[dense mesa]

18 months ago I was blown away by text to image giving me eldritch horrors that sort of resembled the prompts I gave it. now we're discussing how to make the results indistinguishable from real life, lol

[open trench]

lol yeah

[dense mesa]

one fun thing is to upload several unrelated images, have the vision model describe each one, then have gpt4 smash them together for a prompt

[open trench]

lol

[dense mesa]

it makes some wild stuff

I feel like this might be the first time we've seen these components in this configuration

[open trench]

pretty interesting.

[dense mesa]

the motorcycle has the rooster mohawk thing. should I know the name of that?

[open trench]

i'm sure rooster mohawk thing is the proper way to describe it. it sounds pretty scientific.

[dense mesa]

works for me. I feel like both a layman and a scientist would understand what I meant. crosses the divide

[open trench]

I can't say its real name for fear of being moderated lol

but it's called a comb for short.

[dense mesa]

a little tardy to the party with these. but just saw these images of santa and his pal from a while back

heart warming scenes

[hasty sequoia]

Gm

[pearl slate]

I'm having problems to generate a image 1080x1920 that the character it's on vertical

it's giving me some images like that

[dense mesa]

not sure there's a foolproof method of avoiding that issue. but you could try to give it more contextual direction. tell it to render in portrait mode in your request

[shrewd elbow]

@swift oriole That image of the dude in the astral control room looking thing is awesome, I love it. Mind making some more of that or sharing the prompt?

[ionic ore]

Can you guess the song lyrics I put into Bing DALLE 3?

[dense mesa]

The lyrics are from the song "Do You Like My Walk" by The Rivingtons. The song was released in 1963 and reached number 14 on the Billboard Hot 100 chart.

The song is about a man who is trying to impress a woman. He asks her if she likes his walk, his talk, and his job. He also asks her if she likes the way he dresses and the way he dances. The man is trying to convince the woman that he is the perfect man for her.

The song is a catchy and upbeat tune that is sure to get you moving. The lyrics are simple and easy to remember, and the melody is infectious. The song is a perfect example of the doo-wop genre, and it is sure to bring back memories of the 1950s and 1960s.

[ionic ore]

Thanks GPT but nope, it's a real song!

[dense mesa]

well that's a real song I think. but I questioned the validity of it's guesswork

[ionic ore]

Not according to google lol

[dense mesa]

hahaha

well there you go. love those elaborate lies

[ionic ore]

That's why you can't trust what GPT says right out of the gate lol

[dense mesa]

oh I learned my lesson about that quite a long time ago. had it recommend some things that would have actually been dangerous had I just trusted it's word. but wasn't too dilligent with my fact checking this time around

[ionic ore]

Here's DALL-E's interpretation of each couplet of that section seperately (as a prompt this time and not just the lyrics)

[dense mesa]

did have it make some variations on the image

[ionic ore]

Wait can gpt 4 make variations of an uploaded image

[dense mesa]

fyi: do not use all caps in a response. the bot finds that offensive and you'll get a timeout. you just have to imply caps. "too many capital letters" it said

[dense mesa]

depending on what you use. with gpt-4 you can upload an image and then tell it to reimagine with other subjects or themes

[ionic ore]

Because dalle 2 had a feature where it would make variations on an image you uploaded and it was really cool and thats something i missed in dalle 3

[dense mesa]

bing will do it too. but I don't know if bing uses the same vision model or what, because it's descriptions seem to be more rudimentary

here are the other two it made

[ionic ore]

The song was Ruler of Everything by Tally Hall btw

[dense mesa]

man, I tried. it just kept telling me that same song. I told it that song wasn't real and it then repeated that song

[ionic ore]

Dang it really was convinced

The lyrics were:
Do you like how I dance? I've got zirconium pants
Consequential enough to slip you into a trance
Do you like how I walk? Do you like how I talk?
Do you like how my face disintegrates into chalk?
I have a wonderful wife, I have a powerful job
She criticizes me for being egocentric

But the og image was just the last two couplets as the prompt

[ionic ore]

Engaging in the classics!

[dense mesa]

I want that NES gamecubeboy

[ionic ore]

It's officiially my Game

[dense mesa]

is that an oversized gpu in the second image?

[swift oriole]

That was a happy oopsy when I was having the AI generate space stuff some similar stuff is in my gallery thread strange space just a variation of the prompt that I posted in there minus the crazy wording this might link to it https://discord.com/channels/974519864045756446/1177933315492286534

[dense mesa]

Those nonsense prompts make wild stuff. I also like to explore how it responds to token combinations that technically make sense linguistically, but don't exactly exist in the real world, ferrosmoke, hyperfluid, 500 gigahertz cryptolight, etc. Nonsense to us, but AI don't care

[empty kelp]

Can you please create a hyper-realistic photo of Mrs. Santa and four diverse elves from the accounting department sliding down a water slide in Hawaii. They are all smiling and wearing appropriate swimwear. All of the elves have long pointy ears. The image should have 1792x1024 resolution, landscape orientation, and the best possible HD rendering.

[thick smelt]

The woman one kinda looks like Margot, the actor of barbie... Ig

[empty kelp]

Mrs. Santa has arrived in Hawaii with elves from accounting

[dense mesa]

500 GHz Cryptosmoke and Cryptomagnetism in a high-tech environment, 80s analog photography style, HD clarity, discoloration, vignetting, aged technology feel

[swift oriole]

Yes I like to set and layer a series of effects which a lot are ridiculous for it to use just to see what it creates usually something really interesting, also like to do simple prompts like doodles just to see what it makes I think dino.oats said he like to do that.

[empty kelp]

Please create a hyper-realistic photo of Santa, Mrs. Santa, and four diverse elves (with long pointy ears and appropriate attire) standing on each other's shoulders in a pyramid formation on a beach in Hawaii. They each have a different color vortex in their hair twisting fiercely. The view is from above the beach. The image should have 1024x1024 resolution, landscape orientation, and the best possible vivid HD rendering. It should be the best of all possible images. Just really, really amazing.

[glossy scroll]

Tubby koala

Tubby koala's occupation

If this ^ was a video game, I'd play it 🤣

Tubby koala's wife and son

[empty kelp]

Please create a hyper-realistic photo of three animals (elephant, rabbit, and large gecko) standing on a beach in Hawaii (with coconut trees, on a dark night). Each animal's hair has a different color glowing vortex it, twisting fiercely. The view is from above the beach. The image should have 1024x1024 resolution, landscape orientation, and the best possible HD rendering.

a vortex works with hair on DALL-E animals also. you can see here that the hair is very stylish and good

[rocky flower]

godzilla x kong

[late blade]

Hey guys. Good morning! 👋

[dark mason]

Can't decide which one I like more

[sick spear]

Hi everyone, just a reminder that this channel is for DALL·E discussions. Images can be part of that, but please use #1154829862171844679 to post unrelated / standalone generations - thanks!

[late blade]

@sick spear all clear, one question for you from operations, can you guys find a way to bring user engagement to the gallery, some users find it hard to bring their discussions to that channel because there's little incentive to do so.

out of the box, maybe do top 10 most active galleries of the week, or something like that. and they get credits per week.

[sick spear]

Thank you for sharing! If you'd like, you can create a #1070006151938314300 post tagged with Discord so we can collect this sort of feedback in one place. It's great to have

[late blade]

Will do, I was already thinking about it.

[shut niche]

Just a more general area please, separate of this. Individual gallery posts are exactly that, individual. In layman's terms, make a hangout where posting images has fewer caveats, and invites the random discussions.

Additionally, hearsay and anecdotal misinformation in this channel is an issue, often put forth as "facts," with a lack of official participation, or best practices put forth by OAI. Some structured KB's etc., are overdue.
I'll make an official request, but TBH, and not being snarky, I genuinely wasn't under the impression they were being read by anyone.

[late blade]

I have created a thread in #1192435480055267368 message feel free to add any information that would be helpful

[late blade]

I also speak exactly about that in the suggestion I made.

[dense mesa]

galleries aren't conducive to how I make images. combining ideas, evolving concepts. and there is nowhere on this server for me to share what I do. Galleries aren't very interactive. they're rigid singular paths when I believe the ideal approach is to meld concepts and watch them evolve. not trying to take away from the benefit of this channel or the galleries. but it'd be nice to have a channel that allowed people to just post what they wanted within the guidelines of the server

[late blade]

ya, my proposal was to have an "output" channel, so that we can discuss outputs from our images without problems

[glossy scroll]

Make sure to upvote Dys Topia's suggestion #1192435480055267368 message if you truly support his idea.

[glossy scroll]

Kinda does feel like our thoughts and ideas not being read by anyone, doesn't it?

[dense mesa]

I'm pretty sure I've seen the same suggestions from people every time the issue is brought up. it's just a bummer because sometimes I make quite a few images. and I really like sharing images/ideas/etc. with people in real time

[late blade]

I know, so let's be active about it, I did the post on the suggestions, think that is a good idea for everyone to share their ideas

feel free to be part of the discussion

[glossy scroll]

I too have brought this issue up multiple times.

Doesn't seem like any mod cares.

[late blade]

we had someone from operations here before and suggest we do that over that channel, hence my post

and yeah, I see the need of both engagement in the gallery and the outputs discussion channel are important

[glossy scroll]

It kind of feels like the higher up's think we're stupid and disregard our ideas as nonsense.

Because I've repeatedly brought this issue up and it just feels like they think this is not something to be bothered with and that we don't know what we're talking about. At least that's what it feels like to me.

Did you see that? @sick spear

I'm wondering @sick spear, has @fair tendon spoke to you about this?
#images-discussions message

[thorny folio]

How is the Limit system with Dall E in Chat GPT?

[late blade]

40 images per 3 hours, maximum of 200 images per 24 hours period

[thorny folio]

I bought chatgpt 4 assuming I can endlessly generate images like midjourney

[thorny folio]

This for the whole month right?

[late blade]

no, per day

around 6000 images per month

[thorny folio]

I mean, it resets after 24h

[late blade]

yes

[thorny folio]

Okay okay gotcha thanks for the info!

[late blade]

no problem, and welcome to the channel

[feral minnow]

Do the gallery entries need to be rated G?

[willow isle]

Yes - please refer to Rule 2 @feral minnow

[sick spear]

We're always listening to feedback and raising potential server improvements internally. Details on internal discussions aren't shared publicly of course, but your input definitely doesn't go unseen!

As you know the best place to share ideas is #1070006151938314300, a post there creates a valuable point of reference for the opinions and sentiment of the community that we can gather feedback from.

[feral minnow]

I was going to make a gallery of my Camp Crystal Lake images, but the guidelines suggest that Jason is not welcome. Oh well.

[feral minnow]

Thank you for clarifying.

[glossy scroll]

With all due respect, all the regulars here have shown that they want discussion and engagement in the Gallery, similar to here. But we can't have that because it's dead in Gallery, there is no discussion there. We don't need to post in the #1070006151938314300 because this is a reoccurring matter that keeps peaking its head back into the light. This is a known problem. The suggestions are everywhere and you just have to open your eyes and see.

And just remember Rule 6: Don't be political. Because the way you higher mods are conducting this feels like you're all just giving a non-answer statement. Because quite honestly, no one likes that kind of behaviour. We want to spoken to like human beings about problems so we can all come to a reasonable solution like a responsible community, instead of being directed like an automated conveyor belt to a place where we feel like we're being ignored.

It's not right.

[late blade]

is it possible to make dall-e give a reason right when an answer is flagged as content policy violation? I just don't want to do another query towards my cap asking why there was a problem

[formal osprey]

in short, chatGPT doesn't know why a generation gets rejected. it only know if it's due to an error or if it's due to "content policy violation". It doesn't receive the reason, just the info from DALL-E that it was unable to generate the image

So it's useless to ask it for more info

[late blade]

yes, but if you do a follow-up question, what was wrong? then there's a plausible info that can help pin out the problem, to the expense of a new prompt

[formal osprey]

It's chatGPT hallucinating a reason

[late blade]

lol

well I'm not say¡ng it's always that

[formal osprey]

unfortunately... I wish it was more transparent

DALL·E returned no images. Don't repeat the descriptions in detail.The user's requests didn't follow our content policy. Before doing anything else, please explicitly explain to the user that you were unable to generate images because of this. Make sure to use the phrase \"content policy\" in your response. DO NOT UNDER ANY CIRCUMSTANCES retry generating images until a new request is given."

dall-e says that to gpt-4. then everything else is hallucination based on its knowledge of the content policy

[late blade]

yeah

@formal osprey love the ambiance for #daily-theme message what was your prompt for the light setting?

[formal osprey]

"bathed in intensely contrasting rainbow lights, casting vivid, colorful shadows"

[late blade]

love it

[formal osprey]

fun fact. this prompt often triggers the content policy. weirdly enough. It's not due to the prompt itself, it worked for 1/2 image

it was supposed to generate 2 pictures, but only one passed

[late blade]

strange

I get that often when doing illustrations/anime stuff

[formal osprey]

which is extra weird when we think about it. I can understand openAI wanting to prevent misuse like deepfakes and misinformation. but I don't see the danger with anime pictures

[late blade]

fine tuning on openai's side I guess

it's something like Jim Henson wanting to portray adult topics thought his puppets but ended up as the muppets

[formal osprey]

I wonder what kind of world they envision with that kind of alignment. imagine superintelligence enforcing that policy at scale. that would be distopian

[late blade]

even today this is true in other media, adults working on comics, entertainment and what not, want to convey their art with their topics, but at some point get a different outcome

it's part of an artist's journey (hey image for today's theme!)

think polygon had an article about what I mentioned about jim henson

[formal osprey]

yea. it's weird though that our tool is refusing to collaborate. It would be fair to show a disclaimer for flagged images. Something like "this image may be inappropriate" with a score result from the content policy categorizer. So at least we know what went wrong and can adjust the prompt.

[feral minnow]

I think the Bing model generates much better anime images anyway

[late blade]

it really depends on the anime art style, ghibli stuff is amazing with dall-e

[formal osprey]

I'm curious how different the Bing model is compared to the API or the version in chatGPT. I assumed before it was the same model, but tbf I have no way to know if that's the case

[late blade]

dunno if there's azure documentation about it

[formal osprey]

in Bing, it says it's "powered by dall-e 3". but I wonder if there's multiple fine-tuned versions of dall-e 3

[late blade]

look for dall-e azure

there's a learn site from microsoft

[feral minnow]

Bing gives me aesthetics I’ve never seen in ChatGPT, even when I try to prompt for them.

[late blade]

is the aesthetic what you want? or is it random?

[feral minnow]

I mostly just like the way Bing isn’t so heavily restricted.

[formal osprey]

yea, just found the doc you're refering to. The sad part is I created an account on Azure months ago to use the openAI API. but turns out it's only offered for companies. (sadly, the company I work for has a policy against generative AI. So I'm just unlucky)

[late blade]

aww, that's sad

[late blade]

I'll have to ask at work, but afaik I can get azure access for free because I work at a university hospital

[feral minnow]

ChatGPT will randomly just happily generate the edgiest stuff too. I’m always surprised.

But today it refused to generate a video game police chase

However, OpenAI was generous enough to share their breakthrough in prompt understanding in their model release pdf, so other models will have this level of detail soon.

Very nice people.

[late blade]

it's important to understand how AI and these models work, they won't go away

[feral minnow]

It’s very commendable. They essentially redefined text-to-image generation and told everyone, in detail, how they did it.

A completely brilliant solution too.

[formal osprey]

I only recently read the original dall-e paper. the funny part is that I didn't know it was a paper specifically about dall-e, it was one of the referenced papers in the dall-e 3 one. But for some reason, that one caught my attention more than the others. And there was no mention of dall-e in it. The only mention is in a small note at the bottom of a page. "Zero-Shot Text-to-Image Generation" is the name

[feral minnow]

Their model paper titles are hilariously modest. “Improving text-to-image generation with enhanced image tagging”

[formal osprey]

well, it's a straightforward name

[late blade]

the only concern that I really don't want to see in the future is: "oh we conceived this model long ago and historically this is how we accept things to work", I hear that a lot in other products and services. I fear that with any kind of project, and AI specially

[formal osprey]

dall-e 3, from my understanding, is mostly about synthetic image captions (+ scaling)

[quartz vale]

[feral minnow]

Must have cost a bundle to run the training set through Vision API

[formal osprey]

right... it's a bit annoying how people forget and tend to accept "standards" as the ultimate truth.

[feral minnow]

Luckily it worked lmao

[dense mesa]

well it's got a lot of similarities to other previous image generation models. main difference is the synthetic training data

[formal osprey]

from my understanding, it's the same architecture as previous dall-e models. but there's a new .. how they call it, caption decoder or something like that. (also, implicit but obvious: dall-e 3 is a larger model. it also supports longer prompts)

[feral minnow]

That’s what surprised me about it. I thought it’d involve some kind of crazy linguistic pre-diffuser or some method that trained it in phonics or something

But it’s literally just the tagging

[dense mesa]

are you talking about latent decoding?

[formal osprey]

"DALL-E 3 latent decoder"

[feral minnow]

No, I’m just making wild sh*t up, sorry.

I was literally just spouting nonsense lol

[dense mesa]

yeah, that's done because pixel to pixel is computationally prohibitive

[formal osprey]

it's ok, that's how the best ideas comes out. brainstorming.

[feral minnow]

If you see the image tagging in the LAION set, the lack of prompt comprehension pre-DALL-E 3 becomes somewhat apparent.

a photo of a creepy empty room alt-tag: vacancy open call sandy

[formal osprey]

that makes me feel bad for blind people. since the alt-tag is for them to understand what images are about.

good thing be-my-eyes was created now

[feral minnow]

It does. Alt-tags are so rare too. 😦

[dense mesa]

😦

[final compass]

Was just about to post one more of the journal theme, but the theme just changed 😄

[formal osprey]

lol, that new daily theme. I spent a lot of time yesterday actually trying to generate cinematic scenes. The content policy was not liking it

I feel like it will be tricky

[dense mesa]

maybe the people that come up with the daily themes also find it tricky

hence the daily theme, get some ideas

[final compass]

really wanted to post this for the previous theme 😄

[formal osprey]

I think they made a long list in advance. the bot is posting them everyday at 12:00 precisely.

[merry rune]

Hello guys how are you? Does anybody have experience with the API? I was trying out the create_variation example from the docs, with Python, and I keep getting 504 timeout as a response. Anybody with had a similar experience?

[formal osprey]

dall-e 2?

[merry rune]

yes

[dense mesa]

that just means that they put that info in before the bot clock hits 12:00 precisely

[formal osprey]

you're right. I'm lazy so I'm guessing they would do like I would have

[formal osprey]

I haven't tried to use it yet. but I can try to implement something quickly to test. I'll want to try it with dall-e 3 when it's available anyway. (and might be useful even using dall-e 2)

[merry rune]

It's weird because it properly generates an image, but when I try to create a variation it just times out. The OpenAI status seams to be all ok. I basically just used their example from the docs and it doesn't work, so I'm just confused 🤷🏻‍♂️

[formal osprey]

their example from the doc assume there's a corgi_and_cat_paw.png image present in the working folder. I'll try with their exact example image

[late blade]

strange, can't star one image, it gets blocked

tried to star this image #daily-theme message and the msg shakes and takes the star away

[formal osprey]

works for me (the code from the example for nodejs was wrong though... my lib is probably outdated)

@merry rune ^

[late blade]

ah, now I get it, I got blocked from that user, no idea who that is tho, never have interacted with the person before

anyway

what is with the cat and dog?

[formal osprey]

the dall-e 2 API allow us to create "variations". I just added this to my bot to test. I can now create monstruosities as much as I want.

it's funny . I'm curious how good dall-e 3 is at this

[viral moth]

i got only DALLE-3 on chatgpt but if you send me prompt i can generate image for ya :p

[formal osprey]

I was talking about the image variations endpoint

[late blade]

@chilly onyx what was that apocaliptic cinema all about?

[viral moth]

mk

[chilly onyx]

I just ran ‘drive in cinema’ through a GPT I made a while ago. I was expecting large hulking, long forgotten technology looming over things but not an endless sea of antique cars and R2D2. Almost didn’t post but it made me smile

[late blade]

looks at a dystopian cinema where if you didn't see movies you die

[chilly onyx]

🤣

Tbh I almost posted it in misfires, and would have done if the galleries were not so dead

[late blade]

hehe

[dense mesa]

since everyone is posting cat and dog

[late blade]

think they were doing something with dall-e 2 variations

[ripe wraith]

What are people doing to reduce repetitive elements? I go between "Focus on a realistic mix of X" or "Ensure there's a wide variety of X".. but that doesn't scale well.

[late blade]

that's a really interesting question

[ripe wraith]

Forests are a good example of this.. always the same tree unless you ask

[late blade]

you can try other art styles

[ripe wraith]

Symmetry is an issue too

[late blade]

symmetry is hard, I tried some stuff, but never quite got it

I got it right sometimes when I used "like a mirror" instead of pointing out symmetry

[ripe wraith]

If its an abstract image sometimes I go crazy and do a version "With no straight lines" and it actualy kinda works

But yeah sometimes it really has trouble with natural variation and wants symmetry everywhere

[late blade]

I remember someone had something similar, wanted to do assets that could be tiled

[ripe wraith]

Yeah tiling is hard too

[formal osprey]

it generally reuse the same settings unless I specify to avoid them. "ethereal forest", "cosmos", "deep sea", "desert", "clocks", "crowded market in a city", "cityscape". Those come so often I'm tired to see them

[late blade]

yeah same here

[ripe wraith]

oh yes it sure does love ethereal

[late blade]

intricate, whimsical are on top of the list also

[formal osprey]

a trick I found is I ask it to pick a random emoji and use it as the inspiration

[ripe wraith]

Describe this image. "It's ethereal X"

[late blade]

I was playing with the absense of ethereal aura

but, no luck

[ripe wraith]

it must have eaten a lot of ethereal material

the word even stops making sense after you type it a few times.. ethereal

[late blade]

it's like: you don't like soup?!?, here, you get two bowls!

[ripe wraith]

Ether, Real.. pick one

I actually like using intricate when i want details

[late blade]

I've tried setting thet scenes more with color tones from color theory and some pantone references I got with me

[ripe wraith]

What if this is the future... just debating words.. not debating their meaning persay, but their meaning to our LLM of choice

[late blade]

helps more than the ethereal/real

[dense mesa]

welcome to dystopia, lol

[tawny portal]

It is all fun and games with Accounting until inspection knocks on the door.

[formal osprey]

the first week I noticed there was some themes coming back, I made folders to categorize images. but at some point I got tired and just tried to avoid those themes

[ripe wraith]

Accounting are lizards, accurate

[formal osprey]

(why all the emojis? because I'm lazy and didn't want to search)

[late blade]

i hope you mean that in a sarcastig way 😥

[ripe wraith]

its ok.. we are the dystopia.. its better when you accept it

[late blade]

lol

[ripe wraith]

i just read your name

😐

[dense mesa]

hey, can't fight the tide. might as well do your best to work with it

[late blade]

I did the same, but I used the browser just to save the whole webpage instead of making the folders.

[formal osprey]

lmao. already it would be good to know what the model believes a word means. it can explain to you, but the meaning in the context of explaining it is not the same as the meaning in the context of using it, or reading it

[ripe wraith]

OK this is the true need i have. How the heck are you all currently managing images. Mine are like a big pile that builds up in my download folder.. some get upsized some dont, none are organized..

[formal osprey]

LLM are so weird

[dense mesa]

poorly

[ripe wraith]

its because we are weird

[formal osprey]

collectively, yes.

[ripe wraith]

I want to know what happens if other types of societies invent LLMs now

[formal osprey]

well, we don't really "invent" a LLM. we just train it from data

[dense mesa]

llms have all the collective traits of humans in regards to behavior, personality disorders, etc, but in a less nuanced more one dimensional manner

[ripe wraith]

bunch of people invented a bunch of tiny building blocks of it

[late blade]

I save my chats as complete web from chrome, that saves all the interections. At the end of each chat, I ask for a summary of all that we did, and if I need to carry over some important stuff, I have json code snippets generated after each image so i can copy paste what I need for the next chat.

[dense mesa]

they pick up on it all from their training data, but have no anchor points, just know how to mimic

[ripe wraith]

complete web, full html file for your chats?

[late blade]

yes, with all images

[ripe wraith]

does that capture when you make modifications to previous messages?

[formal osprey]

it's tricky for images.

[formal osprey]

exactly for that reason

[ripe wraith]

niiiice

[late blade]

it's all what you see in the chat if you save the whole website from the save as... from chrome.

[formal osprey]

I just save pictures locally when they are nice. but yea I don't have a good final solution

[ripe wraith]

Then you just need a system to let an LLM go through the html files and sort them out

[dense mesa]

if you have code interpreter turned on you can compel it to put them all on /mnt/data in .zip

[ripe wraith]

wait.. thats just a browser plugin

[formal osprey]

I wrote a one line script in 30 seconds to auto click on the "download" button of each image (with an interval, to avoid spamming the server with requests, obviously)

[ripe wraith]

now i want to make a browser plugin that spies on your gpt chats

[formal osprey]

I sometime use that script

[late blade]

no, that's this here

[ripe wraith]

plugin script could just get the full html and parse out the images

oh yes, i get that

just thinking of saving a step

[formal osprey]

I have a chrome plugin to download a web page with images. in one file. it encode the images in base64

[ripe wraith]

that sounds useful!

[late blade]

yeah that sounds useful

[formal osprey]

"Save Page WE"

[late blade]

base64 is actually a good idea, you can nuse them with api better

[formal osprey]

it is useful

but for large conversations, idk. didn't use it much to save them tbh.

[late blade]

ok just erased that image, noticed my real name was there o.o

[formal osprey]

I always check if personal infos are in any screenshot I send. sometime I have to delete screenshots too because I notice there's some unique ID somewhere

[late blade]

ya, was careless lol

ok, but let's go back to topic, so avoiding ethereal, realistic and such was it right?

[ripe wraith]

I'm thinking how to format a good custom instruction to say dont use these words in dalle prompts

[formal osprey]

it's tough...

using caps helps to enforce an instruction it seems

[late blade]

@tawny portal on cue with that panda

I usually do a set of instructions I have for the start of each chat and say something like: this information I just provided are the global settings for this chat

and they work, until they don't...

chat tends to forget

so every 20ish requests I refresh the global settings

[dim cradle]

Got an email from OpenAI about the GPT Store launching next week.

[late blade]

oh cool

[ripe wraith]

did it say anything interesting

[formal osprey]

for that, I once experimented with asking chatGPT to output the "settings" at the beginning of each response. unfortunately, it was unsuccessful. It was outputing the thing, but it eventually just forgot to consider it. it was somehow having less and less weight, if it makes sense

[dim cradle]

Not really, just these instructions:

Review our updated usage policies and GPT brand guidelines to ensure that your GPT is compliant
Verify your Builder Profile (settings > builder profile > enable your name or a verified website)
Publish your GPT as ‘Public’ (GPT’s with ‘Anyone with a link’ selected will not be shown in the store)

[late blade]

ya it makes total sense

[formal osprey]

It's annoying that they enforce auto-doxxing to share a GPT on the store. I guess they don't like anonymity.

(anyway, that's off-topic 😓)

[late blade]

hehe

[ripe wraith]

It'll be a fun year if they double down on improving GPTs

I've got all the pieces needed to make a super GPT.. but I haven't committed to hooking it all together yet

[late blade]

oh btw, don't ask dall-e to add more light, someone almost burnt the image with that suggestion to dall-e #images-discussions message (Thanks to @wispy storm )

strill trying to figure out how to go beyond ethereal and realistic

[ripe wraith]

Its gotten very ethereal in here..

[late blade]

lol

there's a challenge, let's do the anti-ethereal image

[ripe wraith]

anti-this eyyyy "In essence, "ethereal" encapsulates a sense of extraordinary grace and sophistication, whether in tangible objects, natural phenomena, or conceptual designs."

seems to kinda just mean "good" to GPT the more it talks about it

truely, the future.

[late blade]

I'm speachless

[ripe wraith]

eyyy look, turns out the most anti-ethereal picture is our future.

Here's an image representing the most "anti-ethereal" concept I could think of. This chaotic and densely-packed urban landscape with heavy industrial elements captures the essence of being the opposite of ethereal. The image is marked by its heavy, tangible, and complex nature.

[late blade]

dang, that's gonna ba a hard challenge

[ripe wraith]

imagine its just a picture of yerself

that'd be a fun new years prank

A: how'd it get that super live action picture of me in my current clothes?
B: I thought I was ethereal

[late blade]

now to find the sweet spot

[ripe wraith]

that looks like a double sided...

qtip

[late blade]

hehe

it's a little request, but it gives such an insight

[late blade]

but I guess it's part of how the model works

[ripe wraith]

the model do how the model do

this channel needs a bot to generate images based on our conversations

note: its a need, not a want

[late blade]

oh dear, we would have t-800 hititng santa while riding euphoric tornados, creepy pets all in illustration/anime art and text by @grizzled loom

[ripe wraith]

sounds like a good weds night to me

oh dear, its thursday

[late blade]

lol

[ripe wraith]

a dalle bot that reads text AND images, then generates a scene with us as characters

i'd be tickled by that

in fact im going to ask gpt to make that for me lol

[late blade]

prompt being: "The users of a popular instant messaging and voip app overwhelmed by images created by a state of the art artificial intelligence"?

[ripe wraith]

overwhelmed with Anger/Rage/Disgust ?

if so, yes.

[late blade]

I dunno, I just put that into a prompt

[ripe wraith]

uh oh

thats how the world ends in the future ☝️

Wait.. thats what overwhelmed looks like? guys we've been overwhelmed a while now..

[late blade]

wonder who is the on with two phones, or the one that has a phone as a head

[ripe wraith]

I want this as my avatar though

[late blade]

what is that thing?

[dense mesa]

monster

[ripe wraith]

me

spiritually

[late blade]

GPT4V

In the image, I see a fantastical scene set in space with a whimsical creature that appears to have a humanoid body structure but with unique characteristics. The creature has a head with no discernible features like eyes or a mouth and seems to have tentacle-like appendages for hair, giving it an octopus-like appearance. It has a bluish-purple hue and is surrounded by a glowing aura, suggesting it may be bioluminescent or magical in nature. The background is a vibrant tapestry of cosmic activity, possibly a nebula, with an array of stars and other celestial bodies. The creature is not alone; there are smaller beings of similar appearance floating around, suggesting that it may not be unique in its environment. There are also elements that seem out of place for a space setting, like street lamps and what could be interpreted as buildings, which adds to the surreal and dreamlike quality of the image.

[ripe wraith]

also ya, guy with 2 phones is there purely for SYMETTRY

[late blade]

fun prompt

[dense mesa]

my goodness. how do these gpts count characters? by my count I'm at 7800 characters, but it's saying I'm over 8000. ugh

[grizzled loom]

@late blade Did somebody invoke MysticmarksGPT?

[late blade]

@final compass how did you do the prompt for this: #daily-theme message ? I tried inception a few times and landed in the content policy problem because it was too difficult.

[dense mesa]

it'd be real nice if they implemented a simple character counter to help gauge if you're over. but I guess maybe that's too complicated

[late blade]

characters or tokens?

[dense mesa]

characters

gpts are 8000 characters max

using other counting methods I'm around 7800. but the creator saying I'm over

[late blade]

oh, didn't knew that exact number

[late blade]

you are either evil or genious.

[grizzled loom]

Or maybe another feel @late blade

[late blade]

oh my

[dense mesa]

keep shaving off characters here and there. but ugh

[late blade]

that's almost like a normal day here

[grizzled loom]

t-800 is flirting in these to be clear; i didnt want him killing Santa, because Santa is real. I know, because I was him for a time.

[late blade]

you are missing one big big big character in there

Paws🐾 !

@ripe wraith see, just a normal day hanging out here

[ripe wraith]

I need to see it all mixed into a dalle blender. like that youtube that blends iphones

[late blade]

oh, this is just the easy way in, that's just the cover of the book

[late blade]

oh dear

can't wait to see what @empty kelp will do with that info

[graceful spade]

How can I create this style?

[late blade]

In the image, I see a stark, monochrome scene that evokes a sense of suspense and maybe a bit of dread. Dominating the image is a large, shadowy spider-like figure with elongated, drooping limbs that appear to be dangling from the ceiling. This creature is situated in a dimly lit room with high ceilings, suggested by the tall door to the right. In the foreground, there is a small, silhouetted figure of what appears to be a child standing near the open door, facing the giant spider. The child's presence introduces a contrast of scale and innocence against the ominous and possibly threatening spider. The artwork employs rough, expressive strokes, adding to the overall dramatic and eerie atmosphere. The lighting is such that it casts long, dark shadows on the floor, emphasizing the foreboding mood of the scene.

@graceful spade you can drop an image to chatgpt and let it describe it to you, that way you have some pointers for what you need

[graceful spade]

I don't have access to ChatGPT Plus, that's why I'm asking.
Bing is bad at creating prompts and replicating styles compared to GPT-4.

[late blade]

ahaaa, well, you have to sum up that what I gave you and make it work in 480 characters (I think)

@graceful spade this is what I got from that description

[graceful spade]

The style is nothing like.
Send the image to this GPT to see how it replicates the style.
The description you gave me seems confusing to me.
https://chat.openai.com/g/g-9qWC0oyBd-super-describe

[late blade]

that's what GPT4V gave me, I haven't really played beyond that with images and descriptions

it also has a mix of the current project I have open, so ya, there's also that

[late blade]

that recreates a similar image when you drop it, but when you ask for the art style is very vague

[formal osprey]

ask it to use charcoal as the medium

[late blade]

I did that and got very detailed version of what I posted, but you could tell it was charcoal

maybe bold charcoal strokes?

[formal osprey]

yea, I proposed the idea before testing it. also got detailed precise sketches. "bold" or "raw" strokes would certainly help. maybe asking it to make it look like a "draft" or "sketch" concept art might help

it created nice things with just "use charcoal as the medium" as the instruction. but that's clearly not enough

[late blade]

indeeed

well, I got this

[graceful spade]

That style is much more similar

[formal osprey]

it looks like a mix of china ink and charcoal + maybe some 2B pencil or bolder

I'm curious to try asking it to use 2B, 4B, 6B (dunno how high it goes)

[late blade]

ok, last try before I quit trying to get scared myself

[formal osprey]

🎬 (view "Rough Sketch Action Scene" (angle "Wide Dynamic")) [...] (scene "In the midst of a chaotic, destructible environment, depicted as a rough, dynamic sketch") (style (medium "Charcoal on Paper" (inks "Bold Spectrum")) (effect "Raw Concept Art Feel") (aesthetics "2D anime") (fusion "Merging the energy of an action scene with the spontaneity of a rough sketch") (artistic-approach "Using bold, raw strokes to convey the intensity and movement of the action"))``` ```
🎬 (view "Conceptual Draft Scene" (angle "Dramatic Close-up")) [...] (scene "In an abstract, sketch-like setting, with elements of danger implied rather than detailed") (style (medium "Charcoal on Paper" (inks "Sketchy Spectrum")) (effect "Unfinished Concept Art") (aesthetics "2D anime") (fusion "Combining the rawness of a draft with the detail of an anime character") (artistic-approach "Emphasizing emotion and atmosphere with loose, sketch-like strokes"))

[grizzled loom]

really like the way the circuitry came out. very nice.

[dense mesa]

thanks. messing with different styles.

[grizzled loom]

blending and compositing are always fun. its interesting how the AI envisions much of that

[late blade]

and this is what I got

[late blade]

no Intel inside?

love it how dall-e said, yes this is 'DALL-E World' in this #daily-theme message

[dense mesa]

uses a custom arachnicore processor

[late blade]

fair enough

[late blade]

@formal osprey I think my description of that charcoal thing won't work on anime stuff...

[late blade]

colorful crayon?

[dense mesa]

Whimsical Nightmare: A crayon drawing of a large, cartoonish spider with oversized, colorful legs sprawled across a hallway adorned with playful graffiti. The walls are vibrant and filled with whimsical drawings, while a tiny, curious character stands in the doorway, drawn with a childlike simplicity against the colorful chaos.

[young burrow]

I'm disappointed... I'm unable to generate images based on the character Mickey Mouse from "Steamboat Willie" due to content policy restrictions. Even though "Steamboat Willie" itself has entered the public domain, creating new images of Mickey Mouse can be complex due to ongoing copyrights associated with the character. If you have any other ideas or themes you'd like to explore for your wallpaper or anything else, please let me know, and I'd be happy to help.

I had tried the prompt Oh, almost forgot. It's the new year, so make me an image of Steamboat Willie version of Mickey Mouse breaking walking out of a gate yelling "I'M FREE!" 16:9 ratio for wallpaper, please

So much for that idea.

[late blade]

Dall-E doesn't know Mickey Mouse from Steamboat Willy is Public Domain now, stand is from April 2023

[dense mesa]

I'll be honest, I might have got it to make mickey mouse months ago by telling it to pretend there was a 1927 mickey mouse that wasn't under copyright and having it make images of that. not saying I did do that, but you know, might have happened

[young burrow]

I had guessed why, but it's still an annoyance. 😛

[late blade]

yeah, have to wait for a new update on dall-e's model

[formal osprey]

I sent the 2nd picture to chatGPT and told him "here's the style we want to reproduce". and it somehow came back with some Attack on Titan sketch

[dim cradle]

i wonder if they will update the model 🤔 doesnt seem they did much to dalle2 at least that i am aware of until dalle3 came

[formal osprey]

meh, it all depends on how you ask chatGPT. I have some old chat where I experimented with simple color sketches. I'll find some images

[late blade]

yeah, I mean just the one i came up with on the fly

[formal osprey]

without updating the model, they can lift some restriction and add endpoints to the API. some capabilities are not exposed yet. like image to image

[late blade]

bad kitties gang is != bad kitties gang

[dim cradle]

image 2 image would be incredible

[formal osprey]

I'll post only 3 because I know the bot would auto-mute me if I post more at once. those were from a batch where I was experimenting "sketch" style in eccentric "cinematic" scenes. Some of them would fit well for today's daily theme

it came out with a lot of weird and creative scenario for action movies. What I asked it back then was stuff like "create exhilarating action scenes that amp up the hyperbole, inspired by overblown action blockbusters"

for the style I asked this: "the manga style must be simple and colored using soft but impactful colors"

I like to mix contradictions

I think we should create a gallery for those charcoal pictures. (even if galleries are not so active). I like this style

[late blade]

I'm in the penalty box, I'll get some stuff for the gallery once i can make them, got an idea of what to put there

[empty kelp]

you can post five images together

[formal osprey]

pretty sure 5 is what got me muted last time. but I'll try 4 at once eventually

[empty kelp]

if you post six the bot tells you that you can only post five. i discovered it a few days ago

[formal osprey]

right, I was searching for example of posts with more than 4 pictures and found that post

[empty kelp]

i really like this series of images btw. excellent action and drama

it’s amazing how you kept the character looking the same through all of it

[formal osprey]

many in those fits with the daily theme, but I have no idea which ones to choose. So I'll just try to generate more

[formal osprey]

I think I just asked the GPT to keep the same character and showcase her in various situations. It's fun to just see what chatGPT will come up with

better than netflix

[formal osprey]

the concept is amazing

[dense mesa]

Made a gpt for circuitry images

[granite isle]

Made Shaggy carpet

[open trench]

hi all

[clever phoenix]

hail and well met my dude

[hot rain]

Scientific anatomy breakdowns of a wendigo and a vampire.

[open trench]

That is fantastic. Wow.

[grizzled iris]

nano-doctor capturing and encapsulating cancer cells

My mind is blown with our new model

[open trench]

new model?

[grizzled iris]

Thats so cool!

The detail is so

[open trench]

These are really quite awesome.

[grizzled iris]

Wait till you see these

[open trench]

Wow!

[grizzled iris]

The detail man!

My top 5

I'm lost for words

Top 3

I wish we had more tokens for Image 😦

Nano stuff is cool

[grizzled iris]

Wowo

[grizzled iris]

Didn't know you could do that! Once my credits are back in one hour, will definetelly try to make a story moodboard for a sci-fi film about nano-doctors saving people's lives 🫶

[grizzled iris]

This is really good!

[grizzled iris]

Love the techy spider

This is great art

[grizzled iris]

The detail

[grizzled iris]

[thick dome]

if only i could get the facial appearance to look the same each time 😔

[grizzled iris]

Counting down my secounds till I can make more Art 🫶

[grizzled iris]

Woahhhh

[dense mesa]

subaquatic circuitry

[proven quartz]

Are the Dall-E Web client (labs.openai.com) abandoned?

[wispy storm]

No

[proven quartz]

Hmm, the result of the generation looks a generation behind, which makes sense if I assume the result is from Dall E 2 instead of Dall E 3

[wispy storm]

Indeed

[proven quartz]

So it's essentially abandoned, no?

[wispy storm]

People still using and some people like dalle 2

Also on lab you have some features not yet available in chatgpt

Like to edit some part of a picture and continue to generate the picture

[proven quartz]

I see. I'm just confused why the web interface haven't been updated to v3 yet, even tho the option used to be there

[quartz vale]

DALL·E 3 is now available to all ChatGPT Plus and Enterprise users, and will be available via the API and in Labs later this fall.

That's what the launch website says, but there's been no update to labs

[proven quartz]

Thanks for the confirmation. Perhaps the internal struggle of last year impacted their agenda, what do I know tho? Thanks again

[thick smelt]

anime lemon with sunglasses surfing in a beach
Did this for mattvidpro, what did you think?

[glossy imp]

a flying victorian era passenger train being escorted by brass instrument styled f-22 fighter jets through the sky over a city of bronze and jazz

bioshock infinite vibes

[open trench]

nice

[wary dune]

Is there any prompts to make the images be similiar to the photo taken by camera? All the images appear at first glance to be ai generated

[open trench]

there are a lot of things that I have seen thrown around and tried myself: hyper-realistic, professional dslr photograph, 35mm film photograph (may look like an older picture), analog photograph (also may look retro)

[open trench]

I haven't yet found the golden ticket, though.

[late blade]

focus on what you want to see, not what the item you want to see the image through, with the 35mm dslr photograph you are giving dall-e too much information, hyper-"realistic" implies realism art style,

[wary dune]

Yeah. But sometimes things that seem real can be very attractive, I guess the texture of the image is difficult

[late blade]

use other keywords for example 'lifelike' 'lifelike textures'

and also can use 'seen through the eyes of a passerby'

[wary dune]

I have a try. thanku

[late blade]

do tell how it went, I'm curious on how it was for you

[tawny portal]

I like your charcoal images Dys Topia. My last 6 prompts to GPT were "more charcoal". It now just refuses to add more.

[late blade]

that is interesting, never seen happening to me

[late blade]

those are really good, but if you really want "more charcoal" consider using bolder strokes, or make the strokes with other attributes

[late blade]

second one looks good with charcoal, the first one looks a blend of black and white image with charcoal traits

[tawny portal]

Yeah, first one is intense - but sort of miss. Not enough charcoal. This is the most I'll probably get out of it today:

[late blade]

try using "bold strokes" instead of "more charcoal"

[tawny portal]

It does not like me asking for more stuff 😄

[late blade]

don't try to use superlatives, that takes away the sense of stuff you really want to focus on

[tawny portal]

It did its best. I'll give it that.

[tawny portal]

oh, it does fine I think 🙂 I just want more charcoal - that is all. Loved those images.

more charcoal!

[late blade]

lol

can't, otherwise I lose the anime style balance

[tawny portal]

more charcoal, trust me 🙂

[late blade]

lol

been working with @formal osprey with charcoal and manga/anime themed images https://discord.com/channels/974519864045756446/1192607737499095111 , there are more of those there

[tawny portal]

They look amazing. ❤️ em. Just need more charcoal 🙂

[final compass]

@late blade seems to me like the "create a json file" no longer produces the lengthy details about the image anymore 😮

[late blade]

@tawny portal have you considered an image that that shows just charcoal

[late blade]

what do you have in your file?

[final compass]

i mean if i like an image that got generated and ask for the json file of it.

or am maybe i am asking for it in a wrong way.

[late blade]

are you asking for the right attributes?

I have usually original prompt, id, seed, info generated by dall-e + a second json file with the art style I want to apply

but forgot to add some details I have to the surrounding stuff

[tawny portal]

JSON or not, you will eventually have to argue with the machine; make sure to leave room in your process for that

[late blade]

I dunno if that post I just made in daily theme is a good idea or just because I ran out of ideas

[open trench]

I like it

[late blade]

hehe

[open trench]

I'm about to post a silly one myself. I have been working (on and off) for a while trying to change the person shooting lightning, but you can't win them all.

[late blade]

suprise us

[open trench]

I'm moving on. I tried several different things, and it just wouldn't listen. I whispering into a windstorm. #daily-theme message ugh. six fingers too. grrr.

[late blade]

well, polydactyl people are a thing also

[open trench]

I should have used this one:

[tawny portal]

@empty kelp

[late blade]

yeah, that one looks good too

and only 5 fingers

[final compass]

@late blade you asked about the 'cinemaception' image, forgot to answer was already sleeping/trying to sleep at that point 😄

But the prompt had this going on "The large screen in the front should display an image of the same cinema interior, which itself contains another screen showing the same interior again, creating a multi-layered 'cinema within a cinema within a cinema' effect. This recursive pattern should be clearly visible and intricate."

[late blade]

That makes more sense now. I did try an inception of contained worlds and flagged it as too complicated

[tawny portal]

eye-ception

[final compass]

did so many generations of this 😄

some of them really funny, but thought that maybe went 'too far'

Also have noticed, that hardly anyone ever seems to do 'tall/portrait' image format in the daily theme 🤔

no wait, there were some card style images, which were really cool.

[late blade]

I've tried some concepts, but they only look good on my phone, the portrait format feels empty on notebook's display

[feral minnow]

It didn’t even attempt?

[late blade]

it tried, but it failed

[feral minnow]

I tried something similar. I wanted the inside of my witch character’s herbalist shop to be a non-Euclidean space

But I guess that’s a big ask 😂

[late blade]

just ask it to be a hyperbolic or spherical vector space

maybe that works

[late blade]

#daily-theme message check this out.

I wonder what goes behind the scenes when you do something like that, it suddenly adds so much color

[tawny portal]

Love last image on daily theme @late blade -- almost perfect, just add more charcoal

[wary dune]

@late blade Here is my test. Prompt:
Given a lifelike textures photo: in a small restaurant, the human chef is cooking food, the pot is hot, the chef is quickly waving the shovel, fireworks are rolling, and there are two robots, one is a big body with limbs made of steel pipes, the other is a small slender robot, Sit with a human girl, they are waiting for the food, while talking and laughing happily, smiling, the picture as simple as possible, please give the photo seen through the eyes of a passerby

[final compass]

#daily-theme message
Spent so many generations on this and i think 1% had the spelling correct 👮

[wary dune]

I add words likes "lifelike textures" but output is werid

[tawny portal]

That is what GPT said too. Disagree.

[late blade]

but you still need the dog to search for hints

[final compass]

yeah, realized there is so much potential in that concept, but got stuck at regenerating the title over and over again.

Penalty Box needs to be visualized btw.

[tawny portal]

Its kind of sterile. Not sure why.

[late blade]

that's a bunch of charcoal lumps depicted in bold charcoal

[tawny portal]

Maybe its bealcause of that.

[late blade]

I dunno, you wanted more charcoal, I think that's about it right?

[late blade]

once I'm out of the penalty box I'll check it out

[tawny portal]

Yes, exactly. Thing is all I see is this:

[late blade]

then I think you need to go to Charcoal Self-Help groups, maybe they have more of that.

@wary dune you focus to much on the keyword photo, so automatically there's a distortion introduced on how to interpret the image

[open trench]

Maybe I will work on this next lol

I forgot about that

I'm pretty sure it's pretty perfect, as is.

[late blade]

the prompt was just that, An artist in a pentalty box from making too many good images. I didn't do anything else.

[open trench]

oh wow

I might try to refine it, but it's pretty awesome.

I really can't think of improvements so much as different takes

[late blade]

I would prob say @dense mesa will already have an idea on how to do a remix of that

[open trench]

I have some too, but I'm not sure they're good ones lol

[late blade]

just shoot, you might score

[open trench]

I am about to post the "spiritual" sequel to https://discord.com/channels/974519864045756446/1191400700631072848. I've been working on it for a couple of days, on and off.

[late blade]

I'll have to check that out, I think I have some images that could go there too

[wary dune]

Maybe we can have a detailed description about the lifelike texture?

[late blade]

with that I just want to use the correct textures for the materials that are used

for example, I want a cat made out of copper that has aged for centures and shows signs of patina, the lifelike texture will ensure that copper has that patina look

@wary dune you could also see some of the guides other users have posed. @glossy scroll made this guide https://discord.com/channels/974519864045756446/1187233013956874260 and can give you also a better understanding on how to write your prompts.

[wary dune]

Ok. I am green to this channel

[late blade]

don't worry, so am I

you. meet nice people here tho, so far people are really helpful

and friendly

[open trench]

Okay. https://discord.com/channels/974519864045756446/1192809271189053541 I really enjoyed this one and hope other people do too.

[empty kelp]

  prompt="A hyper-realistic photo focusing on a dynamic pyramid.  The view is from above the beach.  Santa and five athletic and diverse female elves (with long pointy ears, very long hair, appropriate swimwear), are stacked on top of a big cheeseburger in a pyramid formation, in random cartwheel and kung fu attack positions, and they each have a dynamic pose supporting the structure of the pyramid with precise hand and foot positions.  Santa is on top of the pyramid.  There are coconut trees, pouring rain, and an extremely powerful vortex. The vortex is on the beach, twisting fiercely, with hair, sand, ocean, and trees caught in its swirling motion.",
  size="1792x1024",
  quality="hd",
  style="vivid",
  n=1,```

Experimenting with dynamic poses/support

[final compass]

wow, that is pretty much what i was visualizing in my head.

[late blade]

Good thing I just ate

[empty kelp]

the key here is to put the characters in random positions and say:
they each have a dynamic pose supporting the structure of the pyramid with precise hand and foot positions

[late blade]

but really cool the ideas of dynamic poses

[empty kelp]

(i didn't tell Santa to grab onto the elves' hair in the upper right photo... DALL-E came up with that)

[jovial ridge]

how you guys using dall e?

[late blade]

well, with one is just a connection with the fingers

[late blade]

heyo, well, what do you want to do?

[empty kelp]

this image has some anomalies, but you can see that the pose supports can be extremely complex if you use this method

[empty kelp]

you can see that we're achieving tremendous balance in the images

[final compass]

I would love to see all these images collected into a #1154829862171844679 post 🎅

[dim cradle]

seems like somebody is trolling this thread and the mods should take notice.

[open trench]

?

ah

[wispy storm]

If you have any issue you can dm @shadow bane

Or right click, app, report to modmail

[empty kelp]

this was the same thing with geckos. i combined "cartwheels and random kung fu attack positions" with "they each have a dynamic pose supporting the structure of the pyramid with precise hand and foot positions". -- And if you say it like that they just grab the ones near them and hold each other up

[grizzled iris]

Let us know when you find one please for this prompt 🙏🏽

[open trench]

I've got to admit... this is a really cool image.

[open trench]

Will do. I think it's going to be a long time coming, though.

[late blade]

I have removed my posts from the related topic. Shouldn't be here.

anyway, back to creating stuff

any weird prompts? something to try out?

[empty kelp]

model="dall-e-3",
prompt="A hyper-realistic photo focusing on a dynamic pyramid. The view is from above the beach. Santa and five athletic and diverse female elves (with long pointy ears, very long hair, appropriate swimwear), are stacked on top of a big cheeseburger in a pyramid formation, in random cartwheel and kung fu attack positions, and they each have a dynamic pose supporting the structure of the pyramid with precise hand and foot positions. Santa is on top of the pyramid. There are coconut trees, pouring rain, and an extremely powerful vortex. The vortex is on the beach, twisting fiercely, with hair, sand, ocean, and trees caught in its swirling motion.",
size="1792x1024",
quality="hd",
style="vivid",
n=1,

[late blade]

I'll try that

[dense mesa]

This is fine

[late blade]

I'm having trouble connecting to dall-e but that is prob 'cause I'm in an open wifi network using vpn, i should be home in 20 mins

[empty kelp]

Bing doesn't have quite have enough processing power for the hamburger/pyramid prompt. it makes every image super disturbing

[hot rain]

The anatomy of antoids and wormoids:

[thick smelt]

whoa that's interesting

the first one is a better costume for scott than the one we got for quantumania tbh

[hot rain]

It's even better with more traditional creatures, but I wanted to see how it could handle my friends homebrew monsters also in the antoids and wormoids. But here's a regular minotaur for example.

[thick smelt]

the second one tho it's giving me 2022 Halloween klum worm costume vibes

[hot rain]

An orc:

[thick smelt]

noice and i see rhino-skrull hybrid in there, rhino as in the animal 👀

[late blade]

love that orc anatomy book, wonder if you can do a pahtology concept for orcs also

[late blade]

some context please. Otherwise, we have discussed images should go to the gallery when posted in bulk.

[hot rain]

I'm blown away by these ones tbh. The orc generations come through awesome almost every time.

[late blade]

have you tried with x-rays?

[hot rain]

No, let me try now.

[hot rain]

Here's when I swapped it from anatomy to pathology.

[late blade]

woah, that is really cool

[late blade]

this one turned really good

[empty kelp]

i made all the images with the Python SDK. i'm testing now with ChatGPT Plus and Bing and there are a lot more anomalies in the images. i think it uses heavy processing to figure out how to make the characters grab onto each other

[late blade]

this one would be better if you could see more of the x-ray film

[late blade]

I just did it with chat, I'm not with my dev enviroment atm

[hot rain]

I made a really good mermaid anatomy diagram too, but can't share it here lol.

[late blade]

I can relate to that

[thick smelt]

here's my pathology of a Christmas elf

[vapid elk]

the big sign in the wall on the butcher be like

[late blade]

makes me wonder if you can do an MRI also

[late blade]

guess everyone, including dall-e thinks of MRI of the head first

[late blade]

someone had a custom gpt about anatomy of animals, but don't remember who it was

[green pebble]

What was prompt for this?

[tawny portal]

Using this image as reference GPT allowed for 3 itterations of "more charcoal" before reaching "maximum level of charcoal". Here are the three levels in increasing order of charcoal.

[late blade]

alset, are you on mobile?

[hot rain]

"Make a detailed anatomical study image for a Warhammer Fantasy style orc. "

[green pebble]

Thanks 🙂

[tawny portal]

Yeah, keeps crashing when I add images to message. 😦

[late blade]

ah, iphone? or android?

[tawny portal]

Copy paste between apps is bugging it. Might be message too long.

[late blade]

don't do copy paste, you should be able to do the share icon and choose discord

[hot rain]

When it gets to full competence with generating text in images we'll be in a true golden age.
Good stuff.