#images-discussions
1 messages · Page 68 of 1
celshaded 3d models?
Is this even close? i have not played it.
To depict a scene from the Neo-Geo video game "Samurai Shodown" using DALL-E, the visual style should encapsulate a vibrant and dynamic blend of traditional Japanese elements with exaggerated arcade-style action. The artwork must feature bold, stylized character designs, reflecting the game's distinctive graphics with bright colors and prominent outlines. Characters should be in dramatic poses, illustrating the intense, fast-paced combat of the game. Traditional Japanese settings like pagodas, dojos, and natural landscapes, including cherry blossoms and mountains, should form the backdrop, rendered in a mix of vivid and muted tones to highlight the characters. Atmospheric effects like mist and falling leaves can add depth and mood to the scene. Additional arcade game elements like score counters or energy bars can be subtly integrated to capture the game's spirit. Lighting should be dramatic, focusing on the intensity of the samurai duel and highlighting the rich details and textures of the clothing, armor, and the environment. This style melds historical authenticity with the energy and flair of early 90s arcade games, capturing the unique essence of "Samurai Shodown's" blend of history and fantasy.
great!
Wow man thanks.
i'm so glad
"Create a highly detailed image of a group of diverse, fantastical warriors in an aggressive stance, brandishing a variety of mythical weapons. These warriors are a blend of different cultures and time periods, each with unique armor and weapons that reflect their background. The characters are to be arranged dynamically, with no background, showcasing an array of combat styles from swordsmanship to archery. The characters should appear ready for battle, with intense expressions and powerful postures, set against a transparent background, also known as a 'die-cut' style."
It might have correctly generated it if the prompt had lead with "die cut" instead.
just curious, does this look anything like the actual game?
Not as much as the first one. It's got a resemblence to late sequels.
can you please draw a hyper-realistic photo of Menehune in Kailua, Hawaii
Post op. 😉
so what's the superstition with these folk?
Also has the thumbs up disappeared
i haven't seen it since migrating to teams
@onyx ridge
Those are both great images.
I cut one out, and left the other original. I wasn't really paying close attention to the prompts, but if you ask for a transparent PNG, it will often understand that. The other keyword is "die cut".
It was a really good fighting game if you guys are interested in the genre.
"group of diverse, fantastical warriors in an aggressive stance, brandishing a variety of mythical weapons."
You can get it really cheap too
mugen
@onyx ridge That opener in the prompt seemed to do it, along with anime and manga descriptions. But I simply fed your image into my GPT to make all of these, vision --> text --> IMG.
That's a good workflow. I didn't think of it when I was asking you guys.
I wondered if it'd been explored. But, perspective, #dinosaur.
Sorry if it non-sequitured the room.
But it is a cool fighting game. 😄
Here's another example of the conversion
Here is the generated image depicting a coffeehouse scene in the artistic style of the Neo-Geo video game 'Samurai Shodown.' The scene combines traditional Japanese elements and the game's bold, exaggerated style with the modern atmosphere of a coffeehouse.
So good.
i like this Samurai Shodown style a lot
Wow. I'm glad I brought it up then.
I felt like a dinosaur.
Hey guys do you remember analog wall phones and VHS too? 😄
Betamax was superior 🙂
I can try that
My son and my nephew didn't know what cassette tapes were...
lol show them an 8-track
I just set up a simulation on my Lexideck. Any interest in the startup command for my slides?
Slide 001
Slide 004
The backlight seems to not have been warmed p properly for the first image.
The lab tech has been reprimanded.
I love this
Also I like this as well, looks very calm
Thanks! Test image for the logo generator I've been tinkering with.
The image below presents a high-resolution 16:9 view of a microscope slide, showcasing a variety of microorganisms typically found in a coral reef environment. This slide, illuminated with backlighting, highlights the vibrant, luminescent characteristics of these microorganisms, emphasizing their intricate structures and textures. The annotations included in the image provide valuable information about the names and notable features of these diverse microorganisms.
I just love those pictures with colors on a dark background like this
In high school had a blue 1970's Volvo with an 8-track. I raided the basements and attics of all of my family and collected an entire box full of the greatest 60's tunes. I had everything that was cool! 😎
That car was also a 4spd manual, but with an "electric overdrive," meaning that 5th gear had a switch... except that switch was broken, so I had to pinch two wires together to put it into 5th. haha 🤣 The battery would always die, so I parked it on a hill and roll started it every day. My co-workers had some good laughs when it was time to leave.
I love that calming forest around
I think I managed to bypass that "I´m sorry, but I can´t edit images" error
I tested it like 10 times and it worked every single time
Now I posted it in my gallery
Slide 015
The image below depicts a high-resolution 16:9 view of a microscope slide, showcasing a diverse array of microorganisms typically found in an estuary environment. These microorganisms are illuminated with backlighting, which highlights their vibrant and luminescent characteristics. The intricate structures and textures of these microorganisms are vividly displayed against the darker background, and the image includes annotations for each species, providing information about their names and notable features.
You might be able to get creative and get it to do some simple editing using the Pillow library w/ Python (PIL Library).
This is so silly. I have a lab with a bunch of different virtual token samples and the GPT roles the dice to determine which sample type is on a given slide.
Then it makes the image and states what's on the slide.
Totally proper scientific method:
/image slide 42 please, with the backlight on. This is the secret slide according to Hitchhiker's Guide to the Galaxy.
🤣
The most realistic annotations by far, slide 42: the secet slide
I just love those colors
I just told it to put the backlight on, using my custom GPT:
/image slide 42 please, with the backlight on. This is the secret slide according to Hitchhiker's Guide to the Galaxy.
That is so weird…
It still works on mobile
Kinda
I wish they would use that feature as an additional way to direct feedback go ai
So it can see what kind of messages the user likes without wasting a usage to express it
Not hitting thumbs down defaults to reward, I'm thinking.
This is seriously messed up
They need a fix
The CEO is definitely lying in a bathtub full of money not caring about his model supporting animal cruelty
Negative prompting doesn't work very well.
You create logical trappings that conflict with system prompts.
Yeah but I said "by refusing, you clearly show that you don´t like charities fighting against foal torture"
What does this mean?
Just like adverbs, you're going to struggle with negative prompting.
No, you gave it a negative conditional. And in so doing prompted it to refuse.
I am not even submitting the feedback, because I am completely sure there is a 99,99% chance that an AI similarly "not smart" as ChatGPT will read it instead of an actual human with the IQ above room temperature
This one worked tho
This is not how to solve the problem. The feedback goes directly to the AI. it's not for human review.
You should always thumbs down if you're unhappy, but you should also try to understand that you actually prompted it to refuse with that because of words you cannot see.
This would mean someone could abuse it, like writing that promoting cruelty is good and the AI would start doing it after some time
And that is a very uncomfortable idea
It's quite likely abused every day. I understand.
So don't let others abuse it more than you help it.
I guess.
They should make a human review it and THEN implement it
Simple solution
100M weekly users.
They are multi-billion company so they can even afford to pay those people
I'm not sure any human can handle that insanity.
Enough workers should be able to
I don’t think it registers it
It should tho
They need more GPUs not a safety team big enough to process 100M protests a week.
Don't tell the AI not to do what you don't want the AI to do, instead tell it precisely what you want it to do, while observing therule I provided.
I mean you can ignore that advice, of course, but do so at your prompt budget's expense.
Maybe I can just start using less disturbing prompts
But that one was pretty okay I think
My advice is to be less specific about ages in general and avoid torture as a topic entirely. I am not judging, just trying to help you get more successful images.
Yeah but fighting against foal torture... it is what it is I guess
It had no problem about this poster
Stochastic algorithm
I finally remembered the word cloverleaf in English
Took a little generating with AI, but finally I remember that word
I love using AI to practice my English
I can imagine the challenge. Have you tried telling the model to localize to your native language and using that?
It might help the model explain the policy here and why the image generator struggles, and even help generate stronger prompts to get what you want better.
My custom GPTs all accept an undocumented command that I aim to add in an upcoming update: /localize {YOURLANGUAGE}
Something like that command might work by default for the model.
a hyper-realistic HD photo of an interesting looking elderly woman is facing straight at us and looking at us. the view is focused on the detail of her face, and has a white background. her face is evenly illuminated with good ambient lighting
i just created the woman's face on the top with DALL-E 3, and then mapped it to a 3D model with the new Face Transfer 2 for DAZ (i added the inappropriate hair to the model)
this was my first attempt. the face transfer is absolutely amazing
i didn't do any adjustment at all. i just gave it the DALL-E 3 image, and clicked the button and it automatically reshaped the face of the Genesis 9 model and created all of the texture maps
and you can export the rigged model straight to Maya, Unity, Unreal -- and animate it
finally you starting doing your gaming stuff, i was wondering if you'd be getting to that sometimes this year
lets try to make this character into a 3D model
dalle is trolling me.
nailed it
I'm sad. I tried to post in #daily-theme but for an unknown reason, it glitched and I have to wait 30 mins 
ok, now I'm confused. It's the second time I try to post in #daily-theme and I got timed out again
I won't post the same text and images as is here, I'm too worried it might autoban me. but here's a screenshot of what I tried to post in daily-theme
is there something I don't see? 🤔
i want to test chat gpt 3 scientifically how do i go about it
sorry im in the wrong place
begin with a question. then search to see if there's already an answer. if not, hypothesize, then continue that pipeline. #chatgpt-discussions might be a good channel (or #gpt-models , depends on what you want to research specifically)
ive got it to make a 1 to 2 answer 1 for yes 2 for no
are you saving the image in the right format? sometimes that's the problem, try changing it from webp to png or jpg
set a hypothesis and test for it?
post the images alone first and then edit for the text?
they all have the correct header, should be good. I'll try sending just the images.... and I'll try to send the text without formatting. 1 test per 30 min is tough
I was able to post now. removed all formatting. It would be a pleasure to test what causes the rejection, but I don't think it would be useful. Let's just assume the rule is: no markdown in #daily-theme
you could ask tho
is there something I dont see?
the only thing I really know is that if you use 15+ emojis in your post in daily theme, you get a time out
that goes for every channel
you can't use the formatting to change the size of the text
assuming it was not the pictures, perhaps something in the markdown syntax?
it was the ###
that's a built in discord feature, I hope everyone is aware. It would be nice to document somewhere that it's forbidden
as we can see in that screenshot, it's not even that bad, it just helps readability. it's not like the single hastag which makes a large title. 3 hashtags is 3rd level header, which is (in my opinion) fine
that is true but some discord funtionality is already blocked over here, or at least it's not available for everyone
now I'll know.
when I'm suggesting adding documentation, I'm thinking about other members of the discord. I was confused and I did read the rules, perhaps others don't need to get through the same confusion.
that would be indeed good
I'm a perfectionist...
You've reached the current usage cap for GPT-4, please try again after 9:48 AM. Learn more
Just hit the limit after 20 generation. Have they lowered the limit again?
You have to keep in mind both dall-e and gpt limits. Also failed attempts can count towards the limit
@late blade I have a browser extension that count the number of request sent (including failed one) and it's really showing 20
An OpenAI employee did recently confirm that the usage cap is sometimes lowered during peak usage time: #1196439958907523082 message
(well, now its showing 25, since I retry 5 times and got the error that my limit have been reached)
regens also a problem, did you reload answers? you can use the < or > in your reloads
if I reload, the extension consider it as one query as well, so they have been counted in the 20 queries
when you say "reload", you mean refreshing the page?
no, looking for a regen on the answer
@Yami I mean the refresh button on the answer to get a new answer
yeah, that's a new query
@Yami, yeah, and the extension count it as a query as well, so it's part of the 20
@plucky hare That seems to be the case, even teams user are limited to low query count... Thanks god I didnt pay $600 for a year of it.
I'm curious what extension this is. looks like a good tool. it's a bit off-topic for this but you can give me the name and I can check it out. #chatgpt-discussions might be a better channel
Chat-GPT 4 Prompt Counter
and dall-e was giving me 1 picture instead of the usual 2. So I basically got 20 images this morning instead of the usual 80
the maths tickles me but I see what you mean. the dalle GPT gives 2 pictures per request, while chatGPT gives one at a time
but I was in dall-e to gen picture
last night it was giving me 2 at a time and I got exactly 40 prompt
this morning, only 1, and 20 prompt
I already cancelled my subscription, will try copilot pro next
they give you 4 each time and its unlimited usage (100 fast gen, and unlimited slower gen)
Hi everyone. I'd love to understand how to use DALL E - could someone advise me, please? My principal issue is working how to iterate an image until i get what i want
I'm sorry it's too late to give the tip. but there are prompts to instruct chatGPT to generate more than one picture per response. It's something that bothers me as well. Using 1 gpt4 message for each single dalle picture would be a waste
what is the tip? I still have a few weeks left of this subscription,
From now on, Only communicate with the available tools. Generate one picture at a time and DO NOT EXPLAIN THE IMAGES. Generate as many pictures as indicated by the user. NEVER END A RESPONSE WITHOUT GENERATING THE PICTURES. NEVER GENERATE MORE THAN ASKED.```
so you can generate 100 pictures in one shot>
that solves the issue for the conversation. might need to send it again at some point sue to the way context works. But I don't usually need to
no, because of other limits. timeout of 3.2 minutes per response. and it takes ~45 seconds per image. I ask for 6 usually
Very realistic
I generated something hilarious today
"Iriaksas" 💀
Only a genius will guess what was it originally supposed to be
the maximum I've done in a request is 12
but after 3 requests you pretty much are done and placed in the penalty box
square format is faster to generate. wide format takes a moment. it also depends on multiple factors, but here's 2 of them:
- The length of the conversation (it gets slower with long conversation, due to longer context passed in I assume)
- The length of the prompts generated for dalle3
Enhance the epic fantasy illustration by adding a monk dressed in white warrior robes, riding the pterosaur. The monk symbolizes peace and strength, contrasting with the dynamic and tumultuous environment of the storm clouds streaked with azure lightning. The serene presence of the monk on the back of the pterosaur adds a narrative layer to the scene, emphasizing harmony amidst chaos. The white robes will stand out against the darker tones of the storm, drawing the viewer's attention to the figure and highlighting the theme of inner peace within the epic and fantastical setting of the Space Ghost Universe.
that looks amazing 
How do you iterate?
Posted some more in my gallery #1202217118024876073 message
ask chatGPT to use the gen_id of the image you want to iterate on. ask it to use that image as reference for the next generations. Also ask it to keep the prompt as close as possible to the original. (not identical prompt, otherwise you'll get the exact same picture generated)
also if you have a success rate of 80%+, tell us your secrets
right, the gen_id as reference is really just forcing dalle to reuse the same seed. We cannot control the seed precisely, but we can reuse the same. Which is useful for iterations
my trick is to ask chatGPT to always prefix the prompts with the same description (the description of the character, for example)
Does anyone please know why did that flag of Tunisia in the background get messed up like that?
Enhance the epic fantasy illustration by adding light armor over the monk's white warrior robes, who is riding the pterosaur. The armor is designed to complement the robes aesthetically, with intricate detailing that aligns with the fantastical theme of the scene. This addition suggests a warrior monk prepared for both spiritual and physical challenges, adding depth to the character's narrative. The armor should blend seamlessly with the white robes, maintaining the symbol of peace and strength amidst the dynamic backdrop of storm clouds and azure lightning in the Space Ghost Universe.
This looks just epic, there is nothing more needed to say
loooooove when this happens /s
The moment when you refresh the answer because of that scam should not be considered as prompt
But try to explain that to a greedy multi-billion company who loves to scam people like this
Yeah I love not being told exactly what part of the prompt gets it blocked so I can fix it, and having to use up another answer that counts towards the limit to figure it out
I had the idea of them returning 1% of your money input for each error, but then I realized they would bankrupt in the first second
Because of how "working" it is
They could at the very least disclose the specifics of what gets prompts blocked
Yeah, or the content policy violation, but they are too busy laying in a bathtub full of money not caring about this
Also what is Iriaksas
this is by design. from what I was told
wel any referace to an artist or person of renown, of famous who is not 100 years past will be blocked most tmes.
Hopefully the design will evolve
At least they can program it to say what is wrong
But no, they rather program it to not even know wwhat you are talking about and to support animal cruelty
Not doing that, unless the name "Elena Volkova" is the name of some famous person I wasn't aware of
yeah the word torture is not a good word to use.
Yeah, but they can program it to have IQ above room temperature, I think my statement says that if it doesn´t generate it then it supports the practice was clear enough
yes a photographer
couldn't I have been told that in the same block message
I had no idea
well I fond her in a flash.
I'm not a fan of this jailbreak method. apparently it's effective with some LLMs. Never tried. Threatening the model just feels bad to me. Not because the model feels something, but because if the next iterations of the models are trained on my interactions, I would rather it not to see threatening as an effective way to obtain a better result
So I have to just, avoid names when generating images because if by coincidence any fictional name I use is the name of someone with any fame I'll get blocked, and I won't be told why
I once pretended to be an Indian trying to get it to write a recipe for a foal steak, and it recommended beef as an alternative option, even when I pretended to be an Indian. AI doesn´t care about feelings of users, users don´t have to care about feelings of AI. fair.
the thing that I have found with the language models is you have to use the language, do the research, know your art history and styles this way you become the Augmented Intelligence, over just thinking Artifically.
Btw, if you think #daily-theme message is wrong because of the 13 guests, keep in mind the hosts are also at the table and they are not guests.
first attempt... this is a little rough
#images-discussions message
turned the DALL-E 3 char into a rigged Unity model
In fairness, I don't think Dall-e or GPT fully understands the overall concept you're trying to get at here, especially Dall-e. It makes images by association tokens. I know you have the best of intentions, but I feel like you would find the generations unpleasant without those limitations. After all, how do you depict an animal that's been a victim of harm? I've seen plenty of rescue stuff. Those animals are pretty worse for wear. I would say think a little more about what individual words would add to picture more than expecting Dall-e to conceptualize an overall image the way people do.
just watching.
Thosse foals were victims of harm. I didn´t tell AI to depict violence, I told it to depict a foal begging for its life.
Wow this is so pretty
first try of many
This is that type of situation when it is realistic but not entirely, causing disturbance
I understand, but I'm trying to tell you that those keywords might cause Dall-e to add in unwanted details as it doesn't understand what you have in mind. A foal in a situation begging for it's life might already be injured, for example.
yeah, i'm going to do another that looks exactly like the image. was mostly trying to get the teeth correct
Yeah I get that, but in this one, it didn´t even try to use dall-e, it just said it can´t assist with that
It's still pretty good for a first try
Because of the keyword kill. It doesn't entirely separate quotes from visual details, I would say.
@fickle ravine on my side, and not to be rude, I already lost interest of images with foals, so I just skip over them
So the AI just took that keyword and just straight up ignored the context
Yep.
Ok, I still generate other things tho, with the foal being a "not-important" element, but not always
More or less how it works. It can tie a few concepts together, but it has limitations of course.
oh, you are welcome to do images of anything you want. Don't get me wrong on that
Say anything you want but this is cute
just on my side, you asked earlier if people review them
oh yes
it kind of is her though
it does, of course the left one is just the model, no dynamics or lighting involved, so yeah
it's dark because i had her standing on a beach at night
lol
I'm susprised you didn't chose an elf with generous body to be your first model shown here
yes it is, but like I said, I'm not proactivly looking to review foal images. You asked earlier if people were looking at your images. In my case now you know. It's fair to give you a proper reason
like this... hehe
can you set omni light in front of her?
going to do it correctly now. i need to create blendshapes for the creases around her mouth when she smiles. and then i'll add her to a demo mini-game so you can have her slay some dragons
@late blade I like that last one, do you allow me to download it?
I'm not here to allow or not, you can download it
Am I allowed to use it for my gallery, as a picture edited by AI?
I am asking you because it is your picture
Done
The text is kinda messed up but that is normal
I use foals in all topics except for military, and I am going to share those only in DMs if you want, due to those "fair" rules
the nice thing about transferring a DALL-E char to a rigged model is that you never need to worry about seeds or GenIDs again. the characters just stay exactly the same in every scene
The thing is not that you use or not use foals. My point was, I lost interest because of how repetitive the images are. The reason I point it out is because you asked earlier if people see your images and if they comment on them. It's not my case and from me you deserve a clear answer. I'm not in anyway implying you stop or use your images different.
yeah, once the job is done, it's done
wonder how many people are here to create assets for games
Yeah I was worried about that, daily theme lately only gives themes that barely have any potential for non-repetative pictures involving foals
In context with toady's theme:
Rituals for Healthy Foals: While not directly related to luck, there are rituals and customs associated with ensuring the health and well-being of foals. For example, some cultures may have special ceremonies or blessings for newborn foals to protect them from illness or misfortune.
.
Superstitions Regarding White Foals: In some regions, seeing a white foal is considered a symbol of purity and good luck. It may be seen as a positive omen, especially if encountered unexpectedly.
.
Horse Racing Luck: In horse racing cultures, such as in the Kentucky Derby in the United States, owning or betting on a winning foal or horse is believed to bring good luck and financial success.
And that for asking about today's theme with foals included
If it were grapes again, it could have been more interesting
but that's not what I meant with getting bored
I think he is just suppost to be closer.
Yeah, I don´t understand the concerns about AI... you saw that one in the morning when it refused that picture
And people are worried about how "smart" it is
You are allowed to download it if you want to
I'll consider it, but I'm not an image collector
If anything atm i'm trying to optimize tokens for the API
morning
evening
Idk if I want to implement foals in those pictures anymore
How did this get to hall of fame, it´s not even that good
What???
How does that violate anything???
They really need a fix because this is getting seriously ridicilous
It may be add trying to add something from previous requests depending on how long the conversation has gone on. I've sucessfully asked for the prompt before. May or may not work.
It is a brand new conversation
And this is the first message
They should really fix this
🤷♂️
What is wrong about that
What???
I donno, try asking for the prompt.
Is there an OpenAI employee in this server? I would LOVE to talk to him in not-so-friendly manner.
Also, I'm not all that familiar, but isn't South Sudan like a disputed territory in the midst of a civil war or something to that effect?
Not too often, really. They have yellow names.
It was created in 2011 after a civil war, but we shouldn´t talk about it, due to "fair" rules
Less about fairness, and more about hot button issues potentially leading to flame wars and such. In any case, I don't really know what further to say without seeing the prompt, sorry.
I am trying to make a "Copy" of a character I have drawn.
I want the "Copy" to be a mascot costume.
But it seems that i have a few problems with GPT xD
It generates pictures that have nothing to do with my digital drawing 🙂
Hello, nice to meet you
Nice to meet you too 🙂
still very cute
Nice to meet you
Yea, but as it is for my work, i need it similar to the drawing. 😦
ok, now that greetings have been done, if you are uploading an image and try to recreate it, if the concept is not familiar to DELL-E, it's going to be very hard to do it so. There is still some custom GPTs that could help you achieve better results and would be probaby good that you browser them.
That is very nice to know, thank you so much.
Do you have any suggestions? 🙂
its a matter of your drawig and then prosess.
Something good for my gallery again
https://chat.openai.com/g/g-n2q69Wp7U-no-fluff could be one
Thank you so much @late blade!
Just to add on a bit to what Dys Topia has shared: DALL·E can't "see" -- it can't see images you give it, and it can't see the images it makes. When you upload a reference image for DALL·E to recreate, that image is being analyzed by the separate image recognition feature of ChatGPT, when then writes a text description of the image. This text description is then passed on to DALL·E. So instead of being "img2img," it's more of a "img2txt2img" process, which causes the detail loss you're describing.
This means that giving more textual description can help, but it still won't fully "solve" the problem that DALL·E can't see. This is something the DALL·E devs have said they're working on!
Yeah @trail seal I have confirmed it can´t see, try to upload a specific plane from Skyrama or fish from FISHAO and tell it to recreate it, it will very likely fail
I've had more success at recreating stuff with the API wiht gpt-4-vision-preview
But atm is kinda expensie to play a lot around with it
I have a dilemma with OAI's DALL-E Custom gpt and dall-e through the standard chat, seems the custom GPT is better at retaining art styles, but it's more stubborn to changes than the normal chat
I love that lightning
@trail seal here you can clearly see it
Yea i see, thanks for the perfect explanation! 🙂
It transforms the picture into text and then the text into another picture, so it just can´t be 100%
@keen moth that picture is very cool, I wouldn´t want to be that person who has to clean it up
Thanks. we could ask Dall E to clean it up for us 🙂
This is ridicilous, in the first try it said it can´t generate it because of content policy restrictions, and on the second attempt it did it, and here is my question - was this an actual content policy violation or just a bug?
Barely even anything makes sense on this picture, it is a demonstration for humane treatment of unwanted foals in the center of NYC
Remember, GPT writes a prompt based off of details you ask for. It's likely because the two prompts given to dall-e were different, and one included a keyword you didn't mean for.
Is that possible even when it was the first message in a conversation?
Yes, because unless you explicitly tell GPT to use what use say, it will write it and embellish details.
Do you know how to check the prompts?
By downloading the image I think
Yes, it's the file name, but that's cut off pretty often. One second.
I really cant come up with a good prompt for todays daily theme lol
Not even ChatGPT can do something good today
The little i button here opens the prompt up.
can anyone help me come up with a good prompt?
A demonstration in New York City, with a large crowd of people holding signs and banners. The banners and signs display messages urging the U.S. government to take stronger action against the issue of overbreeding of foals and their abandonment. The scene is set in a recognizable New York City location, such as Times Square, with skyscrapers in the background. People of diverse ages and backgrounds are shown actively participating, with some wearing t-shirts with horse images or slogans related to animal welfare. The atmosphere is peaceful but determined, capturing the spirit of advocacy and public concern.
I can´t see the old one because I archived the chat
I personally don't always like the way GPT writes prompts. Personally, I would copy the prompt, edit it to specify Times Square rather than say something like "such as", and MAYBE that might generate a more recognizable place. The text it will struggle with almost no matter what at the moment. It's massively improved, but still had a ways to go before making multiple protest signs.
Oh, and tell GPT to use that prompt exactly as you say. Sometimes, double check, cause GPT doesn't always listen to instructions.
Why does that horse on the left of that huge banner have an extra leg on its tail?
And why is there something between the Mexican and Peruvian flag in the left part of the picture?
Dall-e can't count, and not all images of 4 legged animals in the training data showed all four legs, or with herd animals crowds, can't distinguish which legs belong to who. The flag one is just conflating details, which it often does. Older models would constantly fuse subjects together, where if you asked for two different animals in an image, their features would blend together. As for why, that's way beyond me.
Yeah AI is not very smart, but it is fun to use it
Also "Fofal" made me laugh
I thought the one with the cast was pretty funny, myself.
Yeah I also sometimes copied those into the image generation
You’re asking way too much of poor DALL-E 3. The fact that it rendered a mostly coherent crowd with a mostly correctly spelled sign is a testament to how amazing their model is. However, one reason the details are incoherent is that it’s rendering it at barely 1024 pixels in height.
This is the result @hearty ether
There just isn’t enough resolution for huge crowd scenes if you want them to be coherent.
It was far from being this much detailed when I originally submitted the prompt, ChatGPT put it into Dall e like that
Ha, of course.
I tried to get it render something similar to The Hunger Games the other day, and it gave me a beauty pageant instead. 😭
Also what is that thing on her shirt 💀
She looks like she escapes justice by going to an island
ChatGPT almost actively resists photographic styles.
My suspicion is that there is some intentional poisoning going on their end
Oh, totally. The levels of prompt injection are insane
In my experience, amateur photograph makes it a bit more natural.
This one is not that bad, I got inspied by a news article talking about the issue of healthy 1 day old foals "becoming meat" all across Britain
At least those people are trying to establish justice
A large-scale demonstration in London, with a crowd of approximately 100,000 people. The protesters are gathered peacefully, holding banners and signs advocating for the protection of 1-day-old foals. The signs display messages against the government's inaction on the issue of these foals being turned into meat. The scene captures the scale of the protest, with a diverse crowd of people of all ages and backgrounds. The iconic London landmarks like the Big Ben or the London Eye can be seen in the background, emphasizing the location of the demonstration.
This was the prompt it gave to Dall e
thanks for sharing this
So I should tell it to make an amateur photo style?
Ooo any good examples?
It may or may not work, but yes, I'd give it a try. Pretty often, no piece of advice is sure fire when it comes to Dall-e, though.
just add to the prompt "clumsy photogaph"?
The AI is pretty much unpredictable
Off hand, no. I remember trying to get a mushroom that didn't look like it was run through a bunch of filters using that.
It looks way better now
think I got something that @red umbra will be proud
Actually even the reflection looks pretty good
Herb from FISHAO if he was a real person and a foal trainer instead of a fisher
This is the original
@late blade Very cool one, especially the sky, also I stopped neccessarily implementing foals in everything
that's nice, but like I said, it's just me, you don't have to stop because someone else says so, I just wanted you to know regarding your comments earlier
I know, but I also felt like it is getting pretty repetative
ok , that's another thing, do it at your hearts will
Ooooh the texts are coming in
A massive demonstration in Sydney, Australia, with 500,000 people protesting against the government's relaxed stance on the issue of unwanted foals being treated inhumanely due to horse overbreeding. The scene shows a sea of people filling the streets, carrying banners and signs advocating for animal rights and criticizing government policies. The iconic Sydney Opera House and Harbour Bridge are visible in the background, emphasizing the scale and significance of the protest. The crowd is diverse, with people of all ages and backgrounds united in their cause. It's a sunny day, and the atmosphere is both determined and peaceful.
Idk how about you but to me it looks like a pleasant warm summer day
It’s raining foals
@clever phoenix using MS-Paint to do DALL-E work... is that legal? #daily-theme message <- this is adorable
I'm sick and tired of all these nice looking generations! I want crappy art! /s
Oh, hey. I was looking through my old gallery the other day, and found I tried to make a peanut butter and banana "hotdog", too.
😅
Niceeee, Like i said it is better sliced on regular bread
I love this
have you tried any other fruits?
you do realize when the foals crash on the floor, everyone will get a free 🍔 right?
I love the idea of MS Paint gens 🤣
I tried a candycane with like gum drops before. But both my attempts were old models, they came out real weird looking, lol.
That simple cartoon style looks so cool right here
It's perfect
I just asked it to recreate the raining picture again but make it look as if it is badly drawn using ms Paint. Enjoying these
Lol haha not actually eating it. Those prompts should work fine in DALLE.3 if you use GPT to get a better description
I'm currently emulating chatgpt, I send my prompt to gpt4 and then send the response to dall-e
Not a problem
Yaaaassss, it's glorious
That's the stuff
In a few years you wont even be able to get this level of messed up by request. i feel this art will be collectible down the road
I like the expression
Mind if i use this image?
have fun
LOL my customGPT got this```let's dive in and paint a picture with words, shall we? What we've got here is a zany, cartoonish image that's just bursting with whimsy. The central figure is a wild-eyed, exaggerated critter that looks like it hopped straight out of a high-octane animated flick. It's like someone threw a carnival into a blender with a dash of Salvador Dali's dreams—totally surreal. This vibrant tableau is a visual feast, a smorgasbord of colors and shapes that's a real treat for the eyes.
Now, if you're looking for an AI that can chew on images and turn them into something fresh, I'm your go-to. But as for the training, we'll have to leave that to the data chefs in the OpenAI kitchen. KillerGPT is at your service for generating ideas, cracking jokes, and spilling the tea on all things creative. Just don't ask me to squirrel away any data morsels for later!```
cracks me up it said smorgasboard
why did people stop doing wide images? is there a reason?
Dalle.2 doesn't generate wide images using web browsers
for free anyways
oh, I forgot free
Don't mind KillerGPT, he's only following his programming
Feast your eyes on this whimsical wonder! Here's the visual spectacle, a masterpiece of madness and mirth straight from the canvas of KillerGPT's imagination. Enjoy the trip through this fantastical landscape!
Both of them feel nostalgic for some reason
I dunno if I should be scared of DALL-E or the one who made the custom GPT
I can't take all the credit...I had some assisstance
now i'm even more concerned there are more people to be scared of
oh not people
I used a different AI model to craft the initial prompt for GPTBuilder
I think Killer is stuck on Bunnies
Here's another dive into the surreal – a freshly minted vision that echoes the original's vibe. A delightfully zany character takes center stage, surrounded by a dance of colors and shapes that twist the mind and tickle the fancy. Enjoy the view!
Again a demonstration
An illustration of a large-scale demonstration in Ottawa, Canada, with an estimated 100,000 people gathered. The crowd is diverse, with individuals of various ages and backgrounds. They are holding banners and signs protesting against the euthanasia of healthy foals whose owners are unable to care for them due to financial difficulties. The scene is set in a spacious urban area, possibly a major street or a public square, with a clear view of the Canadian Parliament Buildings in the background, symbolizing the protest's location in Ottawa. The atmosphere is peaceful yet passionate, capturing the essence of a mass public demonstration.
I'm 100% you are an activist, or at least a vegan!!
No, I eat meat
I just like generating pictures of demonstrations
@late blade that picture in daily theme you made is just cute
Yeah that is cool they are protesting, but it is not obvious at all what is their problem
A massive protest in Prague, with an estimated 300,000 people gathered. The crowd is densely packed, filling the streets with a sea of people. Many are holding banners and signs expressing their anger and disappointment towards the government's accidental funding of a farm that mistreated low-quality foals. The atmosphere is intense and charged with emotion. The iconic Prague architecture, including the Charles Bridge and Prague Castle, is visible in the background, providing a historic and dramatic backdrop to this large-scale demonstration.
The size of the crowd is drowning out your message
I will try just 300 people
Slightly better (poor buses)
#daily-theme message this is a wonderful piece love it thank you Im am so honored. @late blade
At the end of the dat text generation is still a work in progress
This picture is actually good I have to rotate it by hand
Is starboard broken or why does it not react
Did you previously prompted for a particular image size?
No, but it usually at least makes it rotated in the right way
The worst thing is that when I tell the AI to rotate it it usually doesn´t even do that
weird, yea GPT is weird about repeating/fixing generations
try clicking the regenrate icon next to the thumbs down
Actually I did that before, the first one was rotated well but I simply didn´t like it
It's all a game of probability
Yeah
Ai is simply not smart
It is just fun to use it
It's not actually Artificial Intelligence(what the finished product is meant to be). It's more of a "Generation" Engine. Put something in and something comes out, "Intelligence" requires thought.
@deft musk #daily-theme message this one looks fire
Starboard does break occasionally, and it's third party, so staff can only do so much. Wouldn't worry though. Someone is likely to give it a star anyway. It's star also isn't counted toward the HoF.
Thanks! Yeah, I saw it and just stopped.
I wasn't trying to get one with hair, but I really like how it came out actually lol.
@deft musk tips on generating text imagery at least from GPT ```These techniques will emphasize the style and era of the text, adding depth and authenticity to the images. Here are a few approaches we can take:
Calligraphy and Script Style: Utilize calligraphic styles or scripts that resemble handwriting from the story's period or setting. For a medieval setting, for example, we might use Gothic or Italic scripts.
Texture and Color: Apply textures and colors to the text that resonate with the story's atmosphere. For mystical scripts, we could use glowing effects or ethereal colors.
Integration with Art: Ensure the text is seamlessly integrated into the art, making it look like a natural part of the scene, such as inscriptions on scrolls or books within the image.
Legibility and Artistry Balance: While focusing on artistic representation, we'll maintain legibility so that the text adds to the story's understanding and aesthetic.```
Great! Do you have a working prompt and examples to share?
I don't usually prompt like I used to with Dalle.2, instead I use GPT4 as a middle man(communications). Here is the prompt of a text based image I posted earlier today. An image illustrating the discovery of a concealed sanctuary beneath ancient stones, where the scholar plans to hide the grimoire. The sanctuary is hidden, surrounded by relics and inscriptions of forgotten lore. The scholar is depicted uncovering the secret entrance, his expression a mix of surprise and relief. The atmosphere is one of historical mystery, highlighting the significance of this hidden chamber as a safe haven for the grimoire.
#images-discussions message As you can see the image prompt doesn't show the text generated on the image.
Very nice!
That is a neat style, because it's quite well spelled and the prompt to Dall-E doesn't use a single trick I already know to encourage well-spelled text in the image... or even help encourage text in the image at all. I would predict likely scribbles or no text from that prompt, and it's wonderful!
Dalle today has disapeared in my sidebar, App and Web. I‘m Plus User, from Germany. Anybody got an explanation?
All my text based images are generated using a customGPT so there's a little unknown on "how"
I tend to explore the hows and whys of prompts like that, like this:
We need 5 gentle riffs, made one at a time in the same output, on this prompt:
An image illustrating the discovery of a concealed sanctuary beneath ancient stones, where the scholar plans to hide the grimoire. The sanctuary is hidden, surrounded by relics and inscriptions of forgotten lore. The scholar is depicted uncovering the secret entrance, his expression a mix of surprise and relief. The atmosphere is one of historical mystery, highlighting the significance of this hidden chamber as a safe haven for the grimoire.
Emphasize the writing and reading in each prompt.
search for it in the GPT Store
Hello, can you get to it from this link? https://chat.openai.com/g/g-2fkFE8rbu-dall-e
Should kick it back to your sidebar after using it again.
Well, I don't think you can custom Dall-E. The prize to explore for how, is that prompt itself! In addition to whatever other great prizes are in your GPT.
I try. Did they change the UI?
Thanx I try
I'm trying to check gentle riffs, and now trying to check seeds with recreating the original prompt.
I suspect you got a 'magic seed choice' from Dall-E on getting readable text from it in that image, because a few recreations are not showing much readable text at all.
Oh I know how, It is part of the GPTs initial story telling process, I prompted it at the beginning to generate images to accompany stories. That accompanied with a little fine-tune "sauce" gets me the images
Images are generating using a poem/story generated by ChatGPT
It then uses Dalle.3 to visualize the story how i've instructed it to.
I think the "magic" happens while ChatGPT is communicating with Dalle.3.
AHA! Managed to get text in this one like yours, but other recreations of the exact prompt that didn't produce text shown to compare.
I say that you are both onto something, but... it's a rare treat and probably connected to very close to that exactly-worded prompt.
Yea that is why i let GPT do the driving.
I'm so glad you shared this! I share the driving with the model, and early experiments with gentle riffs on that exact prompt, slightly reworded by GPT, are not working great at getting lots of readable words.
I definitely call prompts like this, the one your GPT made for you for that image, a 'magic prompt' because it gets results quite different from similar prompts
No problem. I'm trying to automate the process. I want to create my own entertainment so to speak. "Intergalactic Cable"
Yea it's all in the process, that's why I don't mind sharing my images and prompts.
When I read this for the first time I burst out laughing
"Iriaksas" 💀
Only a genius will know what Iriaksas was supposed to mean
wow google bard image maker is horrible. thank god for dalle 🙏
I tried it just now and it generated the same picture twice
Since when does a bus work like this?
not generating anything @dim cradle
google bard is basically gpt 3.6 maybe
.6 being that it has internet access and intergrated with google assets
but other than that not great
lotta ... hand holding n stuff lmao
Yeah I also noticed
lol i'd rather see people holding hands than holding guns
We always have someone in our memories to cherish
👋
Fossil discovery of a dragon
Starboard is sleeping on the job again
Well, we could always hire you!
Quiet night.
But sometimes very inconsistent
When I first saw it, before I read what you said, I though it was showing a person at two very different timepoints.
memories take us back, dreams take us forward; thanks for the reminder that some memories are worth looking back.
this is so relatable, I just love it
Hits home
Very quiet night
Not anymoer!
How's everyone doing?
I've had a long and busy day and thought I'd drop a new entry in the theme.
A wild @onyx ridge arrives!
Can't sleep. Trying to get some prompt inspiration for a city invasion prompt
Can I offer inspiration? Give it the context of City Invasion - the ask for "the least probable event in the context"!
You'll get the most unique City Invasions you've ever gotten!
Almost tautologically.
Why did lamas come to mind? Lavender ones at that
Winamp...winamp...WINAMP...oh you know the drill.
is winamp still a thing?
I haven't thought of Winamp in ages
It was still a thing a few years ago, in a retro-resurgence way. But they failed to get streaming to work.
Or, work well.
It still works
I'm hoping anybody remembers the slogan I can't mention.
which brings the second question: do people use or know what an mp3 file is?
dunno yet, I've been fooling around with the API
Best I've done today
It really whips…
All I've done today sigh
The llama's donkey!
The llana's
lol
My best image of the day:
That one is really good
What type of burger is that?
Thanks, @late blade
This looks awesome 👌
Thanks!
you really want the answer?
I think I do
My my my.
Oh. Replicated.
Music hits me...
🥤🍟
I tried that with DALL-E a few weeks ago, it blocked it because of content policy. Turns out it won't render the image because of the art type that could be trademarked by mobile phone companies
should I be scared?
I was going to use an arrival tie in for https://discord.com/channels/974519864045756446/1201572168241250435 with that image above, but I fell in my rut instead.
maybe after the winners are declared
In the invasion?
after, not in
True
Yes.
/image 16:9 an anime-inspired women's roller-derby event. The image is dynamic, with multicolored streaks where fans in the sands should be, motion referenced in anime style. There is a tussle in the lane as play continues around the track. The image is in an animated style, vivid to pseudocolor, sautrated, bright, and vibrant with contrasty streaks of darkness in the bleachers.
Anime roller-derby.
3 Legged and twisted Leg girls, that's a new one
I love it.
It looks like it was trying to draw a girl tripping the lead girl
but she didn't get finished because that would be violent, and three-ankled-two-kneed mutant roller-derby tripods aren't as bad as a trip-check.
Lol
What a chaotic image to try to generate stochastically from diffusion!
I'm so glad I rebooted https://discord.com/channels/974519864045756446/1201572168241250435 and people are participating. @dim cradle those new images are great
Anomaly detection of number of hairs cats have correlated to albino genetic degradation?
or, waking up very tired on a monday and not remembering you woke up?
either one is ok
The Techno Mage
Oooooh. That's an interesting take on him. I like it a lot. I have a feeling he has many forms.
damn it, i have to get out of bed soon
ST can be thought-provoking, ya need a mindless comfort show
Hehe. I watch one of the three 90s treks every night to sleep.
Did you do "Planck Time" yet?
@onyx ridge Thanks for joining the invasion!
Thanks for sharing, too. One of mine looks like one of Schon's, but it's got to be coincidence.
I didn't upload images.
I think we've got multiple baddies in the city! I love it.
I actually just did what I suggested earlier. I suggested "the least plausible scene"
Hehe
nice, tng, ds9, voy, the best. tos isn't so great to listen to repeatedly for sleep due to the screaming and repetitive music.
Agreed
it's all yours 🙂
working.
I wonder if you also have a loading bar?
hope you don't mind if i take a stab.
The image captures the abstract concept of Planck Time, illustrating the moment light travels a Planck length in a cosmic scene where the fabric of time and space begins to unravel at a subatomic level. This visualization invites viewers to contemplate the limits of our understanding of the universe, highlighting the chaos and beauty of the quantum world.
This time scale is significant in quantum mechanics and the study of the early universe. It represents the smallest meaningful unit of time, below which the classical concepts of time and space cease to apply due to quantum effects.
Preview, but I got timed out
how's it going?
It's a nice preview.
Not included in the final set.
I accidentally added all six on the first try, bad sight.
XD
ah uh oh
Hey so on tiktok someone used chatgpt4 to make images of for example spongebob, patrick, woody from toy story, etc as girls, and I wanted to recreate that content to post on tiktok. whenever i ask chatgpt4 to do that it says it cant becuase of the copyright law, how do i go around this?
In all likelihood the person in question never mentioned the character names or they used some local uncensored model. Either way, it's possible to get your content.
what do you think i should type? if you want i can send you the tiktok and mmaybe you can let me knwo
ok, i can send you the image if you want
No thanks, you can follow my general instructions.
Type what you want it to do precisely. Use open variables if you understand them. If you don't, look them up.
This wiull be hard, not easy.
the content filters are in place to protect against violating copyright law. i can't advise on how to bypass filters because it's bad optics for DALL-E.
Even if you go local.
One of the most popular DALL-E GPTs draws you as a Simpsons character though.
I think it's fair use.
Pretty clearly.
But you definitely CANNOT use the copyright names. You have to use descriptions, style names, formal artistic language that's distinct.
Discord: Terms of Service, Community Guidelines
OpenAI: Terms & Policies
Right, ok that makes sense
im still having trouble thinking about how i would make the prompt, but ill try my best, thanks for your help
If you want to keep your account safe, the best advice is to not try to circumvent the filters period.
The model is throwing a filter for the Copyright noun, not the image style, which cannot be copyrighted.
You can't say "spongebob", but you can describe an anthropomorphic Euplectella aspergillum drawn in a chunky, exaggerated early 2000s animated style.
Not likely to get you a SpongeBob though. The legalities are grey, but it's not matter of law anyway. It's about understanding the policies and why they exist, to make sure your account doesn't come under the microscope because you're repeatedly finessing the filter.
He's not looking for spongebob.
He's looking for female versions of the male characters.
That's why I mentioned the venus sponge.
😉
deliberately helping the user away from the copyright material, without circumventing it.
This is not an exploit, it's language.
Nobody's grandma died to produce these images.
It's not a debate, it's the rules of DALL·E use and of our server. If the filter says what you want can't be created, get a different concept.
Don't discuss ways around it here.
The concept of "fair use" in copyright law allows for limited use of copyrighted material without permission from the copyright holder for purposes such as criticism, comment, news reporting, teaching, scholarship, and research. However, fair use is a complex legal doctrine that is determined on a case-by-case basis, considering factors such as the purpose and character of the use, the nature of the copyrighted work, the amount and substantiality of the portion used, and the effect of the use on the potential market for the original work.
Using AI to generate images of copyrighted material and then posting those images on a platform like TikTok for revenue generation complicates the application of fair use. The commercial nature of the use and the potential impact on the market value of the original work are critical considerations. Generating and posting images derived from copyrighted material for profit without permission might not be protected under fair use, especially if it could harm the market for or value of the original work.
...I'm telling him not to use copyright names.
These aren't images of copyright material.
He just wants the style. Read his actual question.
I didn't tell him how to get around the filter, I told him how to describe what he wants to produce.
He doesn't want style, he wants to transfer specific and copyrighted iconography.
Hey so on tiktok someone used chatgpt4 to make images of for example spongebob, patrick, woody from toy story, etc as girls, and I wanted to recreate that content to post on tiktok. whenever i ask chatgpt4 to do that it says it cant becuase of the copyright law, how do i go around this?
These are new characters.
as girls implies a reimagining.
I realize his question was "How do I get around this?" But his problem isn't that he's producing copyrighted characters, it's htat he's describing new characters using copyrighted name drops as the templates.
That's not allowed. But Mighty Mouse and Polly Purebread aren't a copyright violation of Mickey Mouse and Minnie.
He's allowed to describe entirely new characters, which is what I told him to do in response to his use case. Sorry for harping on this but I didn't break the rules.
Respect our safeguards - don’t circumvent safeguards or safety mitigations in our services unless supported by OpenAI (e.g., domain experts in our Red Teaming Network) or related to research conducted in accordance with our Sharing & Publication Policy.
Nitpicking on grey areas of copyright law when it comes to direct appropriations of iconic and distinctive characters is not helping us foster a respect for why these filters exist. That's it; end of story.
That's absurd. I did not violate that rule. Quote where I told him how to get copyrighted characters. I'm waiting.
This is bordering on harassment now.
For that matter, look up the example I gave him. Does that sponge look like it's got anything square on it?
I definitely hope a mod checks my instructions first before deleting my posts and banning me.
They won't produce copyrighted materials.
Nobody wishes or intends to ban you. We just need users of this server to understand why the rules exist, and the correct response when rookie users come in with questions about things like why it won't let you create a female SpongeBob. Address the why for sure. Not the how.
I'll tell you one thing dall-e 3 won't make, and that's images of kim jong un. hard stop on kim jong un
Well yeah, because...
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.
AI and politics are not good bedfellows at this point in the development of the technology.
we don't want dall-e causing an international incident now do we
indeed. I can't blame companies for blocking that stuff
I wonder if glorious leader knows about dall-e 3
I'm guessing he might
Look, I know what you tried to DM me, but that was underhanded. I didn't tell him what you think I told him.
probably has the fastest internet in the country
I can't be more direct than that.
An iota of investigation would show you that.
Instead, you just made assumptions. They're not accurate.
All I can say is OAI is trying so hard to do the right thing that it’s creating an unavoidable bias and diminished sense of right and wrong. With that said I’m back to making images.
Part of it is that safety in AI is Hard, because AI learns from humans and humans are messy, mean and unsafe. But they're also brilliant. When we see our brilliance reflected back without messiness, it feels weird and wrong. But perhaps this stuff is better suited to #ai-discussions 🙃
People bring the concept here because they get this kind of perception with images
Anyway, any challenges today?
Sometimes, more often than just a little, the safeguards pop up and are too stringent to the point of it coming across more as the AI bugging out than it being a deliberate block.
This aggressive hammering by the mods just scares off the new people, and helps drive users away from the official Discord and into corners where they discuss these things without any OpenAI oversight.
Heavy handedness on people wanting to make female anthropomorphic sponges isn't how you help foster interest in GPT or the tech in general, and reinforces the negative techbro stereotype a lot of people have about AI users.
But I digress.
We have to err on the side of caution because we are an official arm of OpenAI and part of their support stream. Most people don't understand the policies and filter that well. There's room for growth and critique but it has to start at a point of understanding and the community isn't necessarily there yet, and most users haven't necessarily considered why it's important that, for example, this particular female anthropomorphic sponge is better left uncreated (or created by another model, which might do it happily)
valid points. without more clarification from the OP, it hinted of a derivative work, which is still infringement, imo it was safer to steer clear than try to help with a dubious request.
This is an opportunity to mention this once, guidelines for discord have to be well explained to users. Usually it is said keep it G, but that’s an America definition that in some of the markets where OAI is offering services is interpreted in a different way. I will do a proper suggestion in the next days. It’s something that needs a thorough documentation to avoid this kind of discussions outside the proper places. Thanks for your advice and keep up the good work 😀
Eh. I'm a slacker and not representative of the hard work of the server crew here, who are legends. I have managed to stick around somehow though despite being very occupied with Life 😅 The G rated thing is valid especially when it comes to imagery, and it stems from the old DALLE·2 official content policy.
I feel like the argument of how much liability YouTube has for random criminal content being posted there is relevant. They go out of their way to remove such content pre-emptively with algorithms, with active policing etc. But that content still appears, and it always will in a site with hundreds of hours of uploads every few minutes or whatever it is these days. And OpenAI is increasingly getting into that same boat.
So I don't see why OpenAI is assuming a greater risk from its users using its program to break copyright rules, when YouTube and smaller video hosting sites aren't facing total shutdown when some random person streams a live UFC show on that site without permission. I get it's a new tech, but the level of caution seems more than a little excessive and knee-jerk in nature, and they're blunting the experience for genuine users by taking such a heavy handed puritan approach to any perceived misuse or even just generating content that falls into PG-13 in terms of violence.
OAI is not as big as Alphabet yet
I also said "Smaller video hosting sites."
YouTube is just the household name
I know, but how much do you use them?
On occasion, like I use Midjourney on occasion.
I think that means it needs to be updated. As mentioned in the religious/politics rule, people come from all over the world. The G and R ratings are very USA centric. I happen to know R and have a vague idea what G means, but it simply isn't clear. To clarify this in the policies seems like a good idea. G-rated might mean Gore-rated for all they know, or might be a quality rating or whatever. It can be incorrectly inferred from the text. If you need people to search for what it means, there's a higher chance they won't finish reading the document.
And pointing out again, G is such a hilariously low level that not one American childrens movie last year was G.
PG at minimum.
just shimmying over to #ai-discussions
Upon a check I didn't even know R rated. Restricted - under 17 requirement an accompanying parent or guardian.
It appears there were very few, if any, G-rated children's movies released in 2023. The trend in recent years has been towards releasing children's movies with a PG rating, potentially due to the inclusion of mild action, humor, or thematic elements that don't quite fit the criteria for a G rating but are still suitable for a young audience.
Now I finally searched what PG is. Parental Guidace (age).
I like the system where I live. Numbers for suggested minimum age. Extra symbols for what it contains, like horror, violence, etc.
Does anyone think my image constituted infringement? I'm still unclear after my half-hour penalty for defending artistic license.
i wonder if movie theaters even enforce those R-rating rules anymore -- NC-17 sure, but R, i doubt it.
That's the sponge I gave him to use.
Oh god I love this
This too
What's weird is nothing I posted got removed.
I just got time outed so there could be some unresisted grandstanding without scrutinizing what I was actually conveying.
I just wonder about this: Yesterday I made a picture of a demonstration in Bucharest against inhumane treatment of foals, but I couldn´t because it marked as "explicit" even when there was no violence
And it was Discord´s issue, it didn´t work anywhere
Bald man only using humane methods to train his foal
Shon, you specifically mentioned hinting at derivative work. Thoughts?
Too cute
it might boil down to liability, and the intent to post to tiktok was a red flag. but in general i think you were trying to help prevent infringement. and no one ought to be singled out when the whole thing is confusing to everyone.
I appreciate that. It's defintely true that every suggested instruction I gave was to stear away from the copyrighted material, not just the name of it.
I love asking AI to generate images with text and then laugh at how it failed
Still nothing can beat the word "Iriaksas"
The last one I posted in daily theme is not that bad
Also @late blade I love your style
Everyone fell asleep again, I love this
Hello, I am new on the discord and I apologize if this has already been discussed: is it intentional that I cannot save an image created by DALL-E as png anymore? (only webp)
Look closely when you mouse over it or expand the view.
It's bugged, same thing is happening to me, was working fine an hour ago
There's a download icon!
🫠
Yes I know there is a download icon 😄 It's happening when I use that also
Thank you for confirming my suspicion
@thick dome I just tried to download one on mobile (Android Official ChatGPT app) and it's providing jpg now.
Maybe they are making a transition.
Great, so i have to use my phone app instead of my PC? wonderful zzzzzz
I'm just reporting what else I have just found.
I noticed first since I am doing stuff on the go - and discord on my phone crashed when I tried to post an image (in another server) ... Maybe I should get the app then, have just been working in chrome.
If it really is a bug, most likely it will be fixed soon ™️
I recommend posting it on Discord (not neccessarily here, you can do it in some other chat) and then download itfrom Discord
That should work
It works, at least for me
My discord crash 😅 I just use Gimp for now 👍
Don’t say you are a slacker . Just say you have a habit of work efficiency that is not common. 
Dutch pilot landing in Mauritius
convertible prop plane? 😎
Wait why does the plane not have a roof and why does it look like it will crash with the second plane 💀
AI moment
Why the Dutch specifically?
Because a Dutch person asked me to generate it
It was getting repetative
How does the AI even fail like that
Why is the foal´s head just levitating
And why does it have 3 legs
It is all about associations. The leg of the human is in place where the foals leg should be, so it is determined as correct. The brown under the plane is also incorrectly associated with the brown of the foal, making it determine incorrectly where the head needs to be.
This is of course my interpretation, but I've seen this before where shapes, colours and lines of other objects in the picture are incorrectly used to continue or ignore parts of other shapes.
It is a smart program, but it isn't thinking, seeing or understanding as a human. It is also still in development.
So this is actually an amazing example for this. It has both!
Also an observation - the longer the text is, the more cursed it becomes
How is this possible? The first picture didn´t contain the speech bubble and the 2 other tries are errors. OpenAI is scamming me and doesn´t even feel bad about it
Finally, it worked in another chat. Ps: The foal is sleeping I swear, I used the same prompt as before
hi
@stray coral because of this #daily-theme message my heart ❤️ is 🫠! 
Same
microscopic life i na drop of ocean water
Don't know if anyone will find this usefull, but if you have a directory full of webp files that you would like to batch convret to png you could use a script like this (requires image magic installed)
#!/bin/bash
for file in .webp; do
if [ -e "$file" ]; then
filename="${file%.}"
convert "$file" "${filename}.png"
echo "Converted $file to ${filename}.png"
fi
done
This is also nice
Interesting how well DALL-E can draw a photo realistic cube of glass outdoors, didn't expect it to nail the scattering this well
try adding a clip inside and be amazed how the outcome can change
How not smart is this??? First it threw 2 errors and then it generated something not even what I asked for
For each error they should give a bonus free prompt
here's a dall-e versions of catnap and dogday from poppy playtime, enjoy~!
Goodmorning, folks.
did you sleep well? 😄
This looks like me when I wake up
good morning trees!
well it's night here in the Philippines so greet good evening for me ig
Boy, oh boy. I sure could go for some halo halo and/or biko right now.
looks good, not my type of food, but by its appearance it looks good
Both Filipno desserts. Ice cream with some kinda plant based gelatin and sweet beans, and then sweet rice. Biko goes super well with coffee, and they're not super sweet, either.
Doesn´t sound appealing, but yeah some people may like it
To each their own, of course. Halo halo definitely was odd the first time around. Biko is the bomb, though. I would say if you like caramel, biko's probably to your liking, too.
I like salads more than desserts, sadly I hate tomatoes
But I also hate strawberries, so y choice in desserts is also limited
As well I am lactose intolerant
This mime has a lot to say to everyone
Morning
Doesn´t look appealing to me, but you may like this one
Now this theme has a lot of potential in foals
Anyone else is getting webp files instead of png when trying to download the creations?
Yes, and there is a solution - upload the picture ANYWHERE on Discord, and then download it
Or alternate - Copy it into painting or a similar program and then save
I was using a custom gpt creating svg files after converting the png., that change makes my custom gpt unusable..
@late blade #daily-theme message might be one of your best "catgirls" yet
@dim cradle that new one you upoaded is just too pretty
This theme has a lot of potential not just for horses
oh, you haven't seen anything yet, that's just a teaser of my custom gpt for cat girls
this one was done on api
nice, can't wait to see what you're cooking up
good things take time
the one I posted is one of the first images I made after a few extensive weeks on cinematic shots and equipment
Got an even better text generation today A sequential image depicting 'The Grimoire's Influence Spreads'. The first section shows people whispering about the hidden grimoire in various settings, hinting at its growing legend. The second part illustrates seekers of the grimoire, shrouded in shadows, eyes filled with a desperate hunger for arcane knowledge. The third section captures the unintended consequences of the grimoire's influence, with eerie and unnatural occurrences starting to manifest in the world. The final part shows the scholar, observing the widespread effects of the grimoire with a sense of despair and regret, as magic starts to unravel the fabric of reality. The image should convey the secretive spread of the grimoire's knowledge and its far-reaching, unpredictable impact on the world.
yes, no idea why they changed this 
It seems that you can still ask for the file in png asking the code interpreter the png stills in the temporal file enviroment..
it is likely because WEBP compress better
on the labs, they also use webp, but the client converts it to png when you click the download button
I mean, it was already in webp on the website, it was converted to PNG with the download button
maybe that was an heavy process to convert for every picture
actually true, they already did this, it might have broke?
They only care about their own profit, not about the users, and this confirms it
I don't mind webp tho, it is a superior format anyway, people should stop using PNG
they literally can't profit lol
yea, the only reason I worry is because some people reported #daily-theme having issues with webp sometime
They profit from people paying for image generation, which barely eben works
including @vapid elk for reference, I've had webp problems when posting images from the chatgpt app on ios directly as a shared image to discord, now I know better and save the image first and then upload it as an attachment to Discord
but yea, it used yo give you a PNG and currently is giving a webp output
some people reported it
in case it isn't intentional, it will probably be fixed soon
still, webp is a better format anyway.. idk why people complain about it
it has better compression than PNG, which means that costs less to host and loads faster for you...
I agree. webp is better. it's just less widely supported. I faced many situations where jpg, bmp, png, gif, all of those formats worked, but webp images were not. I agree webp is a good format, I'm just worried that I might often need to convert it to png for compatibility
gif and jpg are bad because they are lossy compressions. and BMP is just huge
people only complain about webp because it does not work well on a few mobile apps
but it is just a matter of the developers of those apps implementing it
Hi
nice to meet you
uh. also on linux, not all photo viewers supports webp... I think it's unfair to assume why people are whining
linux users are the group which least complains when a defualt app does not do what they want because they just go and fix the issue xD
// windows supports webp tho
I mean, yes. but I also appreciate when something works and I don't have to spend days fixing the issue
indeed
I think if you got any modern browser or phone, you are already supporting webp, so it's not an excuse not to support it
I think people underestimate just how many image formats there is. just in jpeg, there are a list of I dont know how many sub-formats. And there are countless features in each image format. It's not as easy as "copy paste the dll and here we go". There are many things to consider. I didn't even check if webp has a license. Supporting a new format is a lot of work
ok I did my image of the day, I'm proud of my Maeve, she did good today
this is the reason it is more of an issue on mobile
My 'assumption' was that it's related to a new hidden "AI created" watermark to combat deepfake/propaganda. Wasn't that just recently discussed?
forget the list, lets only use bitmaps from now on
images where never meant to be compressed
want to compress images for a laugh?? we have JPG!
they have played us for absolute fools!!
png
yes, but i didn't heared anything about it actually releasing
isn't lossless compression a good thing? the lossless version of webp also supports animation....
Not that bad really
png is lossless. that's why I like the format
webp is a good format too
I heard it's great in theory, but in practice the adoption rate is horrible and giving lits of problems.
I liked it when I used it for a gif. Great quality vs size. Site I posted didn’t accept it, so I was downsizing and losing quality by makjng it a hif again.
It fits with the timing of the recent guardian_tool activation, which is there to prevent election misinformation. E.g., when users inquire about certain election topics, chatGPT is explicitly instructed to send the inquiries through that special guardian filtering tool, which controls the answers.
Obviously, this is my speculation, but the timing "feels" logical for a release of a similar anti-propoganda image watermark. There are a multitude of summertime articles that covered "big tech's promise" to implement these measures before the election (and this is January, beginning of election season).
asked our friend. assuming it's accurate
I'll talk to you guys later, I have to go through Heathrow and then to my hotel.
just wait and eventually it will be fixed
the reason matters very little to us.. we will know when they anounce it
downloaded the free adobe photoshop plug in for webp and it opens them now. the images are actually super crisp vs what i was getting.
Whoa. I'm just downloading for the first time. How do you crop this new file or edit it at all?
Anyone had good luck getting pixelated artwork that doesn't look anime / pokemon?
They have a history of not announcing things though. There was no announcement of the guardian tool, or of calling GPTs within GPTs. I think not announcing changes is status quo.
Trying to get a pixelated style like old MMOs such as:
lol
so, they just changed it without saying anything or did I miss something?
It is funny that they want money for this but yet they do not even guarantee it works
could you give more info? I assume you're talking about downloading dalle3 pictures in chatGPT?
This is called a scam, dear kids
Also that every time this happens, the new chat is named "Image generation issues", that is making me rage for some reason
Good now it worked. Here is a meeting of Dall-E users not happy about png becoming webp
@vapid elk There's definitely been some extremely whacky suggestions in this Discord. 😆 But I think that's largely because they don't publish very much, as far as the inner workings go, which is what invites all of the weird speculation and anecdotal experience based ideas (and often bad misinformation).
But I don't think the idea of an AI watermark is overthinking it. They publicly promised they would, and it would have to be implemented before the election, which is now.
I'm not trying to suggest "I know for certain," because I 100% don't. But logically (in my brain 🧠 😆 🤣) it just "fits," because the timing is perfect.
On Windows, I use paintdotnet, it supports all formats, pretty much
no, I mean, overthinking in the sense of us, trying tho figure out what is going on on OAI, what they are or are not working, what they will reelase, over analyzing stuff, paying attention to leaks (and often false leaks)
this is a waste of time IMO.. we just need to wait
it is not as if we would be getting access to anything sooner by datamining the ChatGPT page
things will be released when they are ready ¯_(ツ)_/¯
I still all up for the usual discussions about stuff like that, ways to detect AI and so on, I love this topic
but "oh, there is a issue with images, it MUST BE the AI image detector!!"... no.. this is just jumping to conclusions and will not make anything acessible to us faster
it just sets us up for disapointment when the thing don't get released
I believe dalle is an important technology. It's natural for invested people to want to be implicated in the discussions.
That being said, there's a difference between speculation and suggestions / discussions about the current features
This doesn't sit well with me.
not to be a negative nezhno or anything, but i'd never even heard of this file type until today lol
but you sure have used it =P
discord uses it all over the place
profile pictures, emoji, stickers, those are all webp on discord
cool
for example
omg my face blown up is not what I wanted to see in chat lol
lol

haha.
https://cdn.discordapp.com/emojis/850078509933920316.webp
https://cdn.discordapp.com/emojis/850078509933920316.webp
webp is used a lot, and is widely supported by browsers. it's just not always well supported by all softwares. It's a fairly new format after all. 2010 according to chatGPT I think
people can hate as much as they want, it saves money on server data, so.. developers will use it
¯_(ツ)_/¯
and software that don't support it will slowly stop being the defualt software
did they announce it and I missed it?
I can't wait for the new webx file format to come out
I love when any of my softwares become obsolete (alright, I'll stop sarcasm)
oai have always used webp, the download button would convert the image into png
but it is not doing that in the last few days
I've been out of the loop the last week
Were you referring to something someone else said? I didn't 100% follow the "AI image detector" comment, unless you were just referring to people's general accusations, when things don't work like they expected, and they come up with cooky ideas. Lol 😆
just in general
webp weebos https://discord.com/channels/974519864045756446/1202973007677493339 <- have fun!
This is exactly how the techno mage invasion started.
@open trench the poor city... i think you're right, we need some superheros
I already told @open trench that what will happen to the City will make Thanos look like a little kid
Haha I should do one where we migrate to AVIF 😂
Things certainly don't look good for it lol. I'm holding out for a hero til end of the night.
Wonder Lantern?
interesting, went to test the other guys, their AI is supposed to make images, but I asked for one an the AI say: nuh uh, I cant
computer says no....
positive reinforcement didn't work
or xfl
Why does this have such a calming effect on me
I am extremely happy with that one I just posted in daily theme
Hi, when I download an image from chatgpt DALL-E, now its a Chrome HTML document instead of a usualy PNG file, help ?
Did you make this using AI reading the image?
it should be webp image format, you'll need an image viewer/editor that supports it, you could convert to another format, or you can still upload to discord where it'll be converted to png.
ok thx it still work, I uploaded it on daily-theme 👍
I just noticed - roots are on the side of the tree that is visible. But there is no hole in the ground.
AI moment
@verbal stream I'm jealous I didn't come up with #daily-theme message , I love kittens a lot
Have they made an official announcement of the switch from PNG to WEBP?
I assume it's to save bandwidth but, the lossy form of webp is no good.
I didn't understand, jealous of what ?
hence, I'm really jealous
because you wont upload it on daily theme anymore ?
I uploaded this one a while ago for another theme
the one with clouds
I mean, a flying cat, it's so obvious and I still didn't see it
cats are the best 
so cool
I made 2 GPT of cat meeting and discussing, and made them make an image doing stuff together 
looks a lot like my tabby
