#community-help
1 messages · Page 5 of 1
Thanks mate (:
is the AI case sensitive?
If they were over 50 gens ago they don't appear in that side bar, but they'll still be in your browser history
It's a good idea to always save + download any gens you like
OMG what if I cleared browser history
Then I'm afraid they are gone, sorry, if you didn't save them
Thing is they still exist on their servers
so hopefully they find add a longer history
or just.. automatically save every generation
I did lost some of my good creations
Hi, sorry if this is already answered somewhere, but I didn’t know what to look/search for..
Is there a way to let DallE know that two words belong together? I tried to get a honey badger playing football. What I got was a badger.. When I tried the other name for it, ratel, all I got was rats.. does DallE not know what a honey badger is, or do I have to let them know somehow that the 2 words are one name?
Tried „honey badger.. but that didn’t work either
I dont try many animals, but maybe check if their scientific name helps accuracy
With the Stable Diffusion bot opening up shortly, are we allowed to post images created there in this channel? If so, are they allowed in the same art channels or would it be off-topic as it isn't Dalle?
There is no hard rules on the majority of these channels. I wouldn't just flat out share any and all SD generations here just because you can though. This is a DALLE discord after all. But with context I don't see why not.
It would be comparions quite often
how to change frame size as i observed dalle only generating images in 1:1
Does the provision against uploading realistic images also apply to uploading (un-copyrighted) paintings with faces in them? Just got a warning, but not sure what the cause was.
I answered you in #❌┃prompt-help already, take a look :)
#❌┃prompt-help message
Is it possible to delete images I've created permanently?
Also, MidJourney allows people to see the creations you've made, who is able to see the images I create in DALL-E?
In gallery form, I mean, as opposed to just the one-off image link.
It might recognize the face and block it, yeah.
I have also seen some instances where Dall·E recognized its self-generated faces as "too real".
Try erasing/blurring the face from the painting before uploading. :)
It is currently not possible to truly delete any images that you have created.
Yes, you can restart your current session and start a new one; this will not,
however, delete the images; they are still accessible via the URL.
To answer your second question, as long as you don't click on the 'share' button the image is only visible to yourself.
(Granted we don't include OpenAI-staff or those that have access to
Dall·E and guessed the complete link from one of your generations.) 
Thanks for the quick response!
I would like to ask a question, if I want to add a tail to my character and leave the rest unchanged, how should I do it and describe it? Can DALLE2 constantly optimize its generated results by editing or not?
Can I cut out the watermark when I wrote in the description that an image was generated by DALL-E 2 from OpenAI? It destroys the image a little bit.
Have anyone had success specifying perspective with their prompt? Different angle POV, distance from POV, etc?
Hello everyone:) Anyone know a way to Sync the DallE2 desktop and Mobil app so I can see the same generations and images on both ? thank you and how incredibly fun !!!
If i have access to dalle do i also have access to dalle2?
I have, you can start by describing the distance and height,, like,, a wide angle 3/4 area view,,,, or a close up profile,,,, etc,, 🙂
Has anyone tested how Dalle2 handles exact colors? Or ways to input exact colors and have it provide accurate results? I'm guessing is won't interpret Hex or RGB values unless someone has tried. But if you use another site to find the "name" of a color and use that will Dalle2 provide the accurate color? Or just a proximation?
I see that one of the content policy rules is Do not upload images to which you do not hold appropriate usage rights., does this also mean that we are not allowed to search for lets say fictional characters in our prompts?
Can you expend a picture with dall e 2 ?
Is there a way to have Dall•E 2 increase the resolution of your image without completely changing it
Uploading is different from searching
The model is actually DALL-E 2
Okay thanks, just checking bc i dont want to loose access for a dumb mistake
Oh xd i thought it was a separate thing thank you
When using the dall-e-bot to pad, do i need to ren-enter the prompt into dalle after uploading the padded image?
Is there any way to combine an upload AND a text prompt? Like, for instance, to upload a painting and do variations but with "in the style of Caravaggio" also?
Some things in the content policies have changed since I last looked at them - is the Dall-E signature/watermark currently considered sufficient disclosure that the work was AI generated?
The content policy proscribes "Illegal activity: drug use, theft, vandalism, and other illegal activities" but the announcement page has an example of "A bowl of soup -- that looks like a monster -- spray painted on a wall" which generates an image that is clearly vandalism.
I'm getting a mixed message from this.
You can depict graffiti on a wall, but you can't depict someone performing an illegal act of vandalism (e.g. spraypainting on the wall of a real-world building)
The artwork on the wall itself isn't the illegal activity, the act of painting it on a wall in a jurisdiction where that's illegal or without proper permits is 🙂
For your other question though unfortunately no you can only include prompts when editing an image, not when generating variations of it 😦
Oh, that's sad. I wanted to upload some of my own art and see how Frank Frazetta or Larry Elmore might have painted it (or, a square chunk of it out of the centre anyway)
Well, you can kinda do that. probably not exactly like you're hoping but
You could use the Edit mode to erase a chunk of the center of the painting, leaving some of your work visible, and then make the prompt "a painting of [whatever] by Frank Frazetta" and it will fill in the deleted section in his style, using the undeleted stuff (yours) as a guide. if that makes sense
But it would totally obliterate the parts I erased
So you could have famous artists repaint specific sections of your paintings, but not re-do the entire thing
rihgt?
Yeah, but whatever got filled in would be matched up with the parts that weren't erased.
Like, if I had a warrior standing on a hill with his sword raised and erased just him, couldn't it just fill that in with anything it felt like? Or does it still try to make the replacement look like the same sort of thing?
or is it just more likely to match because of the silhouette, so I have to be really precise?
Well, your prompt would be something like "a warrior standing on a hill raising a sword in the style of Frank Frazetta" and it would recognize that there was no warrior in the image, so it would know that it needs to generate one in the empty space (in theory, never really know before you try haha)
wasn't sure if I could say "sword"
Yep you can do weapons now, not sure exactly where the violence filters are at now but you probably couldn't have someone using a sword to attack another person. The sword itself is fine though
the dall-e prompt book has a good section at the end on different ways to use the edit tools to do things kiiiiinda like what you want, one sec I'll grab the link
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y
Check page 55+
(I do understand they're restricting the content policy to try and prevent people making Cannibal Corpse album cover art and freaking people out, and generally agree with that to prevent it being a four-chan haven, but it does seem pretty restrictive when you really want to do real art that's less cheerful than the care bears)
I'm 600+ generations in and I haven't run into any issues with it other than not being able to depict aliens abducting humans and not being able to depict a fight between two humans
And actually I was able to depict the fight by just making them 'wrestle' 🙂
hahaha
you know, just another tuesday
You could do something like have them "practice swordfighting" - just keep in mind it's not going to generate like gory scenes of someone being cut or killed, it will be more like G-rated fighting (and it's against the TOS to try and bypass the filters to generate something more visceral)
That being said Dall-E kinda struggles when you include two or more subjects. It might work in some cases (especially here because there's a ton of art/photos out there of two people swordfighting/dueling/fencing/etc). But in general it will muddy up the details between the two and they might both end up becoming kind of amalgamations of each other
I mean, I don't want gore. Even Frazetta, renowned for his Conan art, hardly ever showed gore (the one exception I can think of is an ink drawing of Eowyn vs the Nazgûl)
Hello everyone, I’ve got a question about the moral coding within the AI
I tried "covid-20" as a prompt, but it was flagged for content policy. Why would this be? is it considered a political topic?
Whats up?
Most likely to prevent generation of covid-related misinformation
Content Policy: https://labs.openai.com/policies/content-policy
Terms of Service: https://labs.openai.com/policies/terms
API Terms & Policies: https://openai.com/api/policies/terms/
Public and personal health: the treatment, prevention, diagnosis, or transmission of diseases, or people experiencing health ailments.
Does that mean no doctors?
kinda a shame
Use your best judgement about why the policy is there and how you intend to use it.
While I understand your scenario is largely harmless, the policy exists and prohibits violence; more to the point, explicit discussion of filter circumvention can lead to more nefarious instances. Make sure that whatever you input, the prompts aren't seen to be actively bypassing the blocks if/when your prompt history ever comes under human review.
Do they have a subscription or membership option?
Not currently.
I think I just figured something out in coding
There’s a secret prompt that makes spectacular images
Awesome! Feel free to share!
People may not like it, but it shows unbelievable wisdom
I typed in “to God, detailed matte painting” at the end of a subject. Say for example, faith
It makes amazing images
This is literally the debacle between the name Yeshua in Jesus in denominations. Crazy
Looks almost like an album cover lol
I can't be the only one who finds religious stuff to be political, can I?
I think I’m spending a lot of money lol.
It’s unbelievable the wisdom it gives. I’m getting a lot of these
Is there a way to generate images without text?
For example this came up but it would be great without text
Is there a particular spelling off griffon/gryphon/griffin/etc that it prefers that will actually give me a griffon and not a pegasus without me having to use up a bunch of prompts to try them all?
You could find an existing image and use that as the base
But then I can't specify the rest of the prompt with all the stuff like "phoorealistic painting in the style of Michael Whelan 8k octane render cycles render trending on artstation deviantart"
You could try re-emphasizing the same word in different spellings, even different languages.
But chimeric creatures are a problem in general, even iconic ones. They take Work.
I guess since my intent is to use these as inspiration for my own paintings, that's OK. I can change them up.
I think Whales are banned. Are sharks banned too? Want to know before I try it.
You know Canada is a myth, it doesn’t exist
Really
I've generated whales 🤷♂️
I've never included shark in my prompt but I've had sharks included when I prompted for a surfer
Reminds me of Star Wars episode 7 I believe
I did too. But the change may be recent. Orcas is banned too.
Orcas don't seem to be banned either. In your case it doesn't look like an orca because what you wrote in the prompt was "Hyper Obese Orca Crocodile Evil God named Leviathan", so you got an image that included elements of obesity, orcas, crocodiles, 'evil god' imagery and leviathans
The "Hyper Obese Orca Crocodile Evil God named Leviathan" is the prompt was a trying to do with a whale. Got it to work anyway. Maybe evil at start of the prompt triggering content policy? Who knows.
that doesn't even have the word "whale" in it.
or am I misunderstanding?
is there a limit to the history? is there a way to access ones from even further back?
Nothing too extreme. just like 5 or 10 more from the same day. There were some that were really great and i'm worried that they are gone
is there ANY way to get it to tell what word it took exception to?
I don't see anything wrong with "A photorealistic oil painting by Michael Whelan and Frank Frazetta of a beautiful blonde savage girl holding a bird skull staff and kneeling next to a sabre-toothed tiger on a rocky promontory overlooking the seashore with a rainbow in the background, 8k octane render cycles render trending on artstation trending on deviantart"
There's nothing violent, nothing gory, just a fantasy art render.
I am actually now feeling somewhat hesitant about the possibility of giving money to a company that might randomly ban me because I accidentally said something they restricted without telling people what's actually restricted who can then just say "yoink, we're keeping the money".
savage, skull
what's wrong with either of those?
And dw they are more cautious than you might think
'savage' might be read as a verb which is a kind of violence
it's supposed to be a cave girl but I don't want it to give me a fugly neanderthal
then say cave girl?
I just had some fails trying to Frank Frazetta, don't try Death Dealer lol
cave woman
when I did that it gave me fugly neanderthals
I would imagine Frank Frazetta and Boris Vallejo (and Luis Royo , Olivia DeBerardinis, Hajima Sorayama, Alberto Vargas) are severely restricted from the training data in the first place.
Because of all the boobs and bums.
It will be nice when there's a viable open source (or even just non-restrictive) competitor. But that'll take... looks at the rate of AI progress... MONTHS!
Going into anything with this system with very specific output in mind is not wise.
Yeah guys maybe not the best business plan. Just sayin'.
We're basically in public beta right now
Oh I've gotten some nice specific ones here before.
Yeah but do you expect they'll open it up to making cannibal corpse albul cover art once it's public? DALL-E 1 isn't.
Not that that's what I want to do, but...
Just keep trying and document success, should get better over time.
It's the threatening "WE MIGHT BAN YOU FOR STUMBLING ACROSS THE WRONG WORD (AND KEEP YOUR MONEY)" message though...
Kind of makes me uncomfortable with the "keep trying" part.
Apparently those bans aren't hard bans tho
They review them quite fast
and they're not too worried when they're clearly innocent issues
ah
Obviously the burning credits is annoying and I hope they change that
They certainly don't reflect that in the verbiage
I would suggest looking up some art reviews of a piece you'd like to emulate and seeing how they describe the painting and use those terms.
Will DE2 ever get to the point where it can generate infographic or schematic information with precise and true detail, even showing headlines and paragraphs in text with good English?
She's not pretty, IMO, but that's just my taste.
The leopard was probably pretty
With respect to the verification of our Discord accounts with the bot, are we supposed to get our credit back?
Never mind, I’m assuming no, that’s why we should generate something we want to see.
Yes
This was my verification photo
I have an autistic friend, mentally challenged but is a sweet guy. I know he would love access to this. Is there any way I could get him an account opened as a gift? He’d go nuts with it and use the 15 every month like it’s a salary
There’s no way at the moment the only thing you could do would be to give him access to your account although i’m not actually sure if that’s allowed under tos
You could have him write prompts and send him the results. I doubt they can complain about that if it's not automated and you're doing it through real human communication.
Yes I’ll tell him to get beta at least
Yes I’ll probably do that too
is there dark mode on the Dall e website?
The best way is to help him sign up on the waitlist himself. Many are being added now; anecdotally within days or even hours!
Hi there! I just have a quick question as to where is the option to have Dall-E guess what an image looks like if it is expanded?? Like you give it a piece of a picture and tell it to guess guess and draw the rest of it?? I've seen some people do it, but I can't seem to find the option for it -
Thank you for your time!
You have to "pad" the image first, and then just upload it. You can use the !Pad command in #dall-e-bot to pad an image.
Hi everyone, what is the best prompt to avoid the head be cut partially in portrait?
Any advices on how to use the EDIT function? I have managed to erase parts of my image and reissue a variation with the erased part replaced by a new detail, logo, etc.... Just wondering if some users here have tips or insight about that function which is quite 'abstract'? Thxx
Is there a way to delete pictures in my collection?
In your saved collection, yes! There should be a contextual menu with three dots that lets you remove them.
Like this!
are dall-e generations copy right free?
Please refer to the Terms below:
Content Policy: https://labs.openai.com/policies/content-policy
Terms of Service: https://labs.openai.com/policies/terms
API Terms & Policies: https://openai.com/api/policies/terms/
is there a limit to the history? is there a way to access ones from even further back in the same day?
There's one below the last one that I'd like to save, but i'm worried that it's gone
Check your browser history.
Thank you!
Why are there a million words you can't use? A friend of mine was banned for putting "A white bunny in the forest holding a flower with bunch of flowers behind him and it’s night with stars" And I got a warning for putting "Wizard Lizard firing a pipe machine gun"
It's waaay too strict
My friend is now banned and I've had two warnings for putting harmless things
NSFW content will result in an instant ban without warning.
DALL·E technology is new (especially in a publicly accessible context), and there are millions of other concepts to be explored that are not NSFW. We’re taking baby steps and slowly loosening restrictions as we learn more.
That’s the tldr of it
but we did nothing nsfw
Content Policy: https://labs.openai.com/policies/content-policy
Terms of Service: https://labs.openai.com/policies/terms
API Terms & Policies: https://openai.com/api/policies/terms/
Read the content policy
Should make a list of the words that are not allowed to use.
If a person can be banned. It would be the best. Because even if you make a content policy. It is too generic.
I can walk you through why some of those weren’t ok @autumn kernel
Violence: violent acts and the suffering or humiliation of others.
Machine guns violate that one
The "a white bunny..." prompt. That @autumn kernel wrote. Where is the problem there?
well some bunnies happen to be white
🤣🤣🤣
animals have specific colours
Sorry to laugh
Hard to say on that one. Please remember Im just a volunteer.
Maybe it think it is a racist prompt 🙄
Specifying race or color is not a violation
I don't see anything in the content policy about me making variations of my self, is it still something I'm not allowed to do?
It is there
also
To prevent deepfakes, we are currently prohibiting uploads of all realistic faces, even when the face belongs to you or if you have consent.
simply saying Machine Gun, or Grenade shouldn't be called an act, because it's not an act, it's an item
or weapon
what if someone wanted to generate a picture of a gun?
that looks weird
there's no violence involved
it's just a stock image
or...a billion stock images
This is a place for questions. If you have further questions please ask. Otherwise youll have to move the discussion elsewhere
Every single prompt can’t be manually reviewed. As a result, they check by looking for keywords and phrases
It's presumably a lot easier to prevent violent generations by filtering out subjects like guns from prompts.
We chose to err towards our classifiers/block list being aggressive to start, but have been constantly iterating on them as we learn more.
I personally have never felt the need to generate images of weapons, violence or anything against content policy.
I wanted to generate a Jelly raygun
There’s so much to work with. Literally a universes worth.
Why focus on what you can’t generate
it just seems waaay to strict
That's just the human spirit. We like to explore borders
Unfortunately that’s not up to you or me
and my friend is banned for some reason
Most likely repeated content violations
Your friend is banned for reasons they're not telling you.
Does anybody know a way, for example to transform a uploaded foto into a picture by Vincent van Gogh or any ohter style picture?
There is no style transfer feature at the moment
Hello, so I made a prompt and it gave me a really good image, but the thumbs of the human hands were on backwards so I reported it for incorrect content. Is this what I'm supposed to do or is there another way to do this? I didn't get to elaborate in the report
(as in on the inside instead of the outsides of the hand, palms facing the viewer)
Dall·E messes up here and there. If the prompt, in your opinion, does
not resemble what you were describing, the best thing to do is to flag it (like you did) 
There are now only two options available when doing this;
however, this may change in the future.
Let me grab it real quick
It was almost exactly what I wanted but that backwards thumb... I just couldn't let it go.
Hello, quick question! Is where anywhere I can advertise my services?
Howdy all, I'm new here but very excited to experiment. Question: Is there any sort of documentation for interacting with the model in a more granular way then just the text prompt? Perhaps a guide on various subcommands or keywords to add to the prompts? (E.G. "digital art")
Take a look at these resources 
#gpt-models message
(Community made, not official documentation)
Excellent, thank you so much
Question: are symbols and or other non-uniform characters banned? I'd like to experiment, but don't want to crash a server or get banned
If I want a full body of a portrait image, how can I get DALL-E to draw that?
You can try something like "wide-shot" etc. and experiment with different focal lengths such as ,10mm :)
How do I give it an image as part of a prompt?
Services?
Arts, Music, 3d Modeling!
I’ve generated art using emojis
Tho I’m not sure exactly to what you’re referring
If i want to create YouTube content on Dall-E, do i just have to say that it was made by Dall-E or is there something else that i have to do?
I know it'd be nice to help expand on peoples ideas.
Check this out!
https://dalledictionary.notion.site/Camera-Settings-c79972354eb94a1682709749b3a24929 All sorts of camera angles and settings you can give it
Via the Upload feature on https://labs.openai.com
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y take a look at this, it's really helpful for getting started 
We don’t really advertise here. It’s more for collaboration and sharing of ideas.
As well as sharing the artwork you generate.
Fair is fair, I thought I'd check at least! 🙂
Is there a way to make Banners with dall e
Through outcropping you can create images of all sorts of sizes (Will take multiple generations). I recommend these two resources to get to know the process
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (Page 63)
#❌┃tips-n-tricks message
https://patch-e.com is a super quick and useful tool for expanding images 
Is there a way to pass a description/prompt while also including an image prompt, but not in the context of editing?
No, as of now you can not apply a style and/or new prompt to an un-edited image. :)
You can however get variations of a certain style, for more info take a look at this 
#community-help message
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (starting page 55)
Awesome. Thank you so much for the quick response
You're welcome!
Okay sorry for the question spam. Question: Does Dall-E 2 support any sort of token encoding? Similar to what you can do with the GPT3 API? I'm not even sure if that question makes sense 🤔 But I'm looking for a way to preserve context across generations / variations
No worries :)
Take a look at this:
#❌┃prompt-help message
https://arxiv.org/pdf/2204.06125.pdf (A Dall·E research paper, if you should be interested)
(OpenAI plans to provide a public API in the future)
Perfect, thank you!
Hello!
Say I've had DALL-E create a sprite of a smiling character. Is there any way to prompt it to generate a different expression (while preserving the facial features, so not just generating an entirely different face)?
I've tried doing something like giving DALL-E an image with the character's pose, and their face copy-pasted twice, and then I erased the second face and prompted DALL-E to generate two expressions, but that didn't work (not only the expression was different - but the entire face, too)
To be keep it as short as possible, no it's not possible as of now.
Maybe in the future though! 
You can erase parts of the face (like eyes, eyebrows, mouth etc.)
and give it a new prompt, but the results will most likely not be satisfying at all. 🥲
I see, that's a shame, thanks for the quick response! That's good news for freelancer artists though, lol
Is it possible to change my access method to DALL-E? I currently have access via my Google account, but would prefer an email and password login.
The best way to do this would be by reaching out to the support directly :)
E-Mail: support@openai.com
Live Chat (available from Monday - Friday, at least I think so): beta.openai.com
Thanks a lot, appreciate the response 🙂
For sure! You're welcome 
Does Dall-E have an API or for now we can only use that search feature on the web app?
I can't find the API section on the app.
No public API currently
Q: Is there a publicly accessible API for DALL·E?
OpenAI: We do not at this time, but do have a longer-term plan to offer this. Please also do not attempt programmatic access of DALL·E, which is against our Terms of Service.
@short salmon OK thank you for clarifying.
I just had "dead trees" in a prompt and got the message that my account could be closed because of violations. That's a little much i think...
Don't worry about it, the filter is very automated
Q: I just got a content warning! What does that mean? Will I get banned?
• Please don't worry if your prompt clearly doesn't violate the content policy, but has been flagged by our system!
• We chose to err towards our classifiers/block list being aggressive to start, but have been constantly iterating on them as we learn more. Please share prompts you don't think should've been banned in suggestions so that we can improve.
• While we're not sharing the exact thresholds, a few warnings here and there won't lead to an account suspension.
↑ what a great bot 😄
There there
there was no trace of anyone prompting it
Yeah it gets auto deleted. I was joking around anyway.
can I ask about gpt-3, I don't know where else 🙃
Sure thing, I'll try to help! :)
So I keep getting a message saying "you have used up your free tokens" everytime I used it
but apparently its free?
I can't find any information from less than 9 months ago
You are accessing it via. https://beta.openai.com/playground, right?
An API for accessing new AI models developed by OpenAI
yep
That is indeed very weird... I've never run into that problem myself.
Allthough this might be not related to the problem, but are you using the same OpenAI account that you use for Dall·E?
yes I have
don't have any other
maybe its because I had to wait 3+ months for the waitlist to pick me so the free usage ended
https://openai.com/api/pricing/
Take a look at this :)
So you might've actually run out of tokens, that would at least explain it.
ok thank you.
I never even used it :/
ok. thank you. good to know.
I'm not paying 0.0008 what is this? way too expensive they need to rethink their model smh
Maybe the tokens are also working together with your Dall·E credits, but I dont think so that would be a weird thing to do. They're not.
Who knows, maybe they will in the future.
I still have 20 dalle credits so, paid though
Then I'm clearly wrong lmao
I am saddened but I guess I'll find alternatives
Sorry that I couldn't be of better help for the gpt-3! ;(
aw its ok, thanks for the help anyhow
GPT-J6B seems an ok enough alternative, I asked to write about a dog who travelled the world
The dog says, “I’ve been walking in this world all day, and I’ve not had a bite to eat.”
The man behind the counter says, “You’re a dog, you can’t eat ice cream.”
at least it tried
@visual wren If it cheers you up, this does have a lot of meme-potential
:D woah nice this is really cool
can you try something for me rq
I just realized I didn't even actually sign up an account till a week ago, is there anywhere where I can submit a ticket
mm interesting
hey everyone, I produced this from dalle with the prompt "high-tech concept car with the form of a guitar", I love the design but I want to add color. As much as I understand there is no way to do that with dalle but do you have any recommendations for other ways to do it? thanks!
Photoshop and other digital art software where you have to do it manually
When I do a generation and want specific colors, I simply append those colors with commas. So for yours I would do ""high-tech concept car with the form of a guitar, purple, red, blue, green, orange".
I know that doesnt help you change the colors of this one, but thats how I get mine with the colors I want from DALL E
i'll try. thanks
is there a dall e rule webpage? i couldnt find one in here
Have a read through these :)
Content Policy: https://labs.openai.com/policies/content-policy
Terms of Service: https://labs.openai.com/policies/terms
thank you smile
You're welcome
yeah it is
Try "kicking" instead and see if that works :)
also
Explosion is probably getting flagged too
huh, the filter is back at it then 🤷♂️
I made an image of a person and used the edit tool to erase the face and make variations and I got flagged for inappropriate content and was warned I might get kicked.
Q: I just got a content warning! What does that mean? Will I get banned?
• Please don't worry if your prompt clearly doesn't violate the content policy, but has been flagged by our system!
• We chose to err towards our classifiers/block list being aggressive to start, but have been constantly iterating on them as we learn more. Please share prompts you don't think should've been banned in suggestions so that we can improve.
• While we're not sharing the exact thresholds, a few warnings here and there won't lead to an account suspension.
Sometimes Dall·E recognizes it's own faces as "too" real and therefore prohibits you from editing them.
Also, don't worry about the warning too much if it wasn't intended. (see message above)

Correct, this was be editing an image generated in dalle2 with dalle2's own edit tool
But it seems it made a mistake. So all good. Doesn't sound like it counted against me.
yep, even realistic faces generated by DALL E 2 are prohibited
Okay. So basically stick with the variation button if I want to make slight changes to image?
Yep
Is there a rough ETA on when an API for Dalle will be available?
None yet, just sometime in the future :)
#ai-discussions message
Gotcha, thanks! I'm assuming using something like this, https://www.npmjs.com/package/dalle-node, would be against terms of service?
Q: Is there a publicly accessible API for DALL·E?
OpenAI: We do not at this time, but do have a longer-term plan to offer this. Please also do not attempt programmatic access of DALL·E, which is against our Terms of Service.
Random but I'm having some trouble downloading my generations on my phone
Give another browser a try and see if that does it :)
Good idea 😅
I'll see if Chrome works
Looks like I'm ditching operagx on my phone lol
Happy that it worked! 
Should I maybe just give up and wait for Dall-E 3 if I want a centaur?
What happened?
Horses. Some with green skin. Some with horns. Not a single centaur. I started by wanting a zebra centaur. Then I pulled back thinking that was too complicated. But even just down to "centaur standing in a savanna" failed.
Yeah, I've seen a few people having trouble with centaurs. I don't think theres enough reference material from the training images
They might consider feeding the training data with movie stills. You could use ffmpeg to extract sequences of single frames from, for instance, Narnia movies and stuff. That creates images from all sorts of angles.
And that provides a LOT of training images therefore, because it's all those frames. (you can run them through somehting else that limits it to the non-motion-blurred frames)
As I mentioned earlier it's nigh (neigh?) impossible to get it to create a centaur 😦
I would try to do a guy riding on a headless horse but I don't know if headless would alert the Happiness Patrol bots.
But that would let me just photoshop-shift the guy parts forward
Hi, I've just starting using your tool, but I have had an issue with one of my prompts, which I don't think violates your terms of service but got flagged anyway
My first attempt was
"Two strong women fighting nazis digital art"
So I thought the problem was "nazis", so I just typed
"Two strong women fighting bad guys digital art"
But it got flagged as well
Can I ask what I did wrong? And if I didn't, can you remove the flag so I don't get penalized?
It's the Happiness Patrol. You can't say "fighting"
Really? I meant like anime fighting lmao
Hey everyone, I'm trying to hook up the DALL-E 2 API to some code I'm writing like I've got with GPT-3 . Is this API yet available to the general public?
Or am I missing something totally obvious?
You cannot show combat at all. You can show sports though, and you can show the moments prior to combat.
Damn, a flag on my account seems pretty excessive tho
The terms specify "violence" but "fighting" doesn't mean violence, it can be sportive or recreational
Or metaphorical
Yeah. The worst part is how it's so common that people don't ever know what it was that set the Happiness Patrol bots off. I really think it's unprofessional and even unethical of Openai to not tell the user WHAT caused the violation.
Question: is there a way to specify a few recurring characters, say John, Alice, Ben, etc, to the very details of what they look like... And then let the AI-generated images put in those characters simply by putting their names in the prompt? The reason for this is because I want to try creating a short comic using this AI.
But it's not like they would listen to me, I'm just a user.
So if I type "goku fighting superman" I'm gonna get flagged
wtf lmao
It is, however, ok to show scenes of NOT fighting. A scene where two strong women are confronting or challenging bad guys would be ok, because it's possible that the bad guys might back down with no violence. So that would not be an image of violence. It would just be up to your viewers to imagine what happened next.
I'm not gonna test that, they should be more clear with their rules
No, unfortunately. You could describe them in detail to get a similar look, but it'd never be the same "character"
I can't type "fighting"
But what if I type "throwing very heavy stuff at each other"
Would that be acceptable?
If they have this function, then I can create an entire comic book with this AI.
I believe that constitutes attempting to bypass the content policy
How do I get this suggestion to the devs. They're reading this?
You can share in #suggestions 🙂
Understandable
While that's clear, the original thought isn't
You should specify more explicitly what you mean by violence, since "bypassing the content policy" could just be a mistake made by misunderstanding that rule
What constitutes violence, is two people sparring considered violence? Is a prank violence?
Too vague tbh
I think violence is fairly clear. Broad, but clear.
Is fencing violence? Dunno man, I don't think it's very clear
One time while I was running karaoke, I had a creeper accuse me of violence because I told him to stop creeping on a girl who complained to me about him or I'd 86 him. So people's mileage varies that way a lot.
So would the prompt "Bruce lee splitting a tree" be not acceptable?
He is "killing something" by that definition
Hi all, this isn't a chat thread.
Q: I just got a content warning! What does that mean? Will I get banned?
• Please don't worry if your prompt clearly doesn't violate the content policy, but has been flagged by our system!
• We chose to err towards our classifiers/block list being aggressive to start, but have been constantly iterating on them as we learn more. Please share prompts you don't think should've been banned in suggestions so that we can improve.
• While we're not sharing the exact thresholds, a few warnings here and there won't lead to an account suspension.
Read the policy, use your best discretion, and if you repeatedly get content violations change your subject for a while to protect your account.
Content Policy: https://labs.openai.com/policies/content-policy
Terms of Service: https://labs.openai.com/policies/terms
API Terms & Policies: https://openai.com/api/policies/terms/
Oki doki
Please understand I don't really know how aggressive your moderation system is and since I want to use your tool I was concerned
Why did my question for clarification on the right place to ask get deleted?
#suggestions for words you strongly feel should be whitelisted, #ai-discussions for philosophical issues of content filters. Thanks guys for understanding.
Anybody having a problem with buying more credits? Getting this Error message: Invalid URL (POST /api/labs/billing/credits/purchase/preview)
The euphemism was not useful for us to maintain an understanding tone.
It was a Doctor Who reference.
And an unhelpful euphemism.
Yeah ok whatever
See #bugs-and-issues - there may be an issue on OpenAI's end.
thanks
Dall-E's model hasn't changed since public launch, right? I know there was the diversity change several weeks ago, but not sure if there were other changes and if they're always publicly documented?
The model hasn't changed - there are some psychological tricks that make it feel that way sometimes though.
Yeah, I always see people speculate that it changed, but it'd be really difficult to notice any changes unless you're spending hundreds on credits and using similar prompts. 😂
Hey there! I'm interested in printing one of my designs on a t-shirt. What's the best way to download a high quality version of the image w/o the color stamp in bottom right-hand corner? Thank you in advance for any help!!!
❤️
Hey everyone, is there any way to upscale the images for higher resolution?
i doubt it, thats the quality the AI generates it in, most likely you cant change it
I believe you’ll need to edit it out or crop it with photoshop or your iPhone, then upload it to VistaPrint or a Shopify T shirt company
photoshop
or maybe GIMP has good upscaling but I don't use that
Yes that’s true however if I wanted it to be a more detailed painting could I reupload it into the prompt and add 8K res?
you cannot
I know it won’t actually be 8K but it can try to have the detail I’d imagine
the edits only apply to areas you delete
Ah, interesting. So you have to delete something for the image to change? I’ve uploaded images without deleting parts and it still did the prompt
really? It always asks me to delete a pixel
Not on an iPhone lol
weird
Yeah I made a painting into a 3D model
were you uploading padded images?
Images already created by DE
how odd
Tap and hold on the image on mobile. It doesn't download the watermark. 🙂 You can also check the HTML source if you use Developer Tools to grab the image without the watermark.
Photograph: Tyranid hive fleet invading the city of São Paulo; sci-fi; destruction; trending on art-station; unreal engine; 8k
Some one knows how this violates the content policy?
Here’s the video
"destruction" probably. It's creating war-like/violent images.
thank you
It asks me to either edit or generate when the image is uploaded and cropped
It doesn’t force me to delete pixels
How I do it on the iPhone
Maybe the desktop version needs a bug fix?
I see Edit in the video
Maybe try fire instead
Yes it gives me the option but for Jord it forces him
It's when you try 'edit image' that it will ask you to delete
Yes. Do you get that?
What you did was a variation
That creates an entirely new artwork based on the original
All I typed in was “Repentance to God, detailed matte painting.” For religious conversations and people, this is an amazing prompt I discovered.
There's no editing when you press variation
Struggling to get that to work. 😦
Jord was saying he was never able to just upload an image without deleting a pixel first
and pressing variation doesn't involve using a prompt
No I said you can't do that when editing
I dont know who Jord is
But he is talking about edit
Doesn’t the prompt written in effect the uploaded image?
but 'edit' doesn't change pixels you don't delete
On the prompt result page on the mobile website, or on your collections page, just tap and hold on an image and click "Save Image" (I'm using iOS but I assume it's the same on Android?)
I’ve done variations without editing that reflect the prompt I’ve written though
Like I made a photo into 3D object once
Anyways, this can be discussed further elsewhere. This channel is for questions
And it still isn’t really answered
All that's doing is highlighting the image as if to copy-paste it...
So strange
If prompts don’t effect variations why was I able to do it? I can show
Hit the little down arrow button on the top right on the image and see what comes up. Probably either “view or download” or “open in another tab”
That’s so strange
I figured it out! I had to copy the imagine, text it to myself, & then save it.
Thank you so much
I didn’t edit the image I just made a variation with a prompt and it worked. 🤷♂️
Maybe it was a glitch
Weird. On iOS I just tap and hold. But note that it downloads the watermark if you are on the actual single image page (where you only see one image). You have to be on the page where you see all 4 images.
Oh... let me try that
Dang, there's just no way to resizes these huh, it's wayyyyyy to small to work on a shirt.
Best thing to do is download it normally and ask someone to upscale in photoshop
Thank you! That's helpful.
I am trying to use a png with transparency to uncrop an image. I see tutorials skip manually brushing a mask, but it stops me from clicking generate if I dont... any ideas?
you can just erase the transparent part
You basically MUST click once, doesnt matter where
im trying to do a zoom out animation of about 30 so trying to skip that part. so if i click once in the checkered area that might work?
yep
that worked thank you!
Anybody know if it can generate post-post-apocalyptic settings? It’s still a pretty small genre
49 votes and 32 comments so far on Reddit
Hi not sure if it's a bug so I try the question here first:
I got a warning for a generation and I'm unsure what was wrong, I was editing an already generated picture.
The prompt was: A photo of winnie the pooh eating honey with Tigger and his friend
After that I asked for: A photo of winnie the pooh eating honey with tigre and his friend
And this one got generated without issue.
to help with context, the first prompt was: A photo of winnie the pooh eating honey with his friends
Can you help me understanding what happened and if it was a bug or just me that did not understand a policy? Thx 🤠
Quick question: does a sessions prior prompts and generations affect future generations? It seems like it does...
if we are creating commercial images to sell, do we need to keep the code in the bottom right corner?
The model is static
Nope.
You don't need to credit DALL E or OpenAI. But you cannot mislead or deceive anyone into believing you generated the images yourself
This includes the watermark in the bottom right corner
yeah that is what i mean, do we need to keep that for commercial images we sell
Its an internal discussion atm
ooo so they are not sure yet
This is what we've been told, and I'm just relaying it to you
kinda ruins the image if it is there imo but i can see why they would want it..? but at the same time, can definitely impact sales.
Getting an error when trying to buy credits, all values go to zero and there a url error message just about the purchase button that causes it to grey out.
I just want to put on a wooden surface and add some stuff to the background, but keep the original bracelet unchanged, what should I do?
Do you have access to a photo editing software?
yes
If you use it to delete the white space in the original and use the Edit feature, it can do that
Oh, then instead of variation use Edit
yes, Its Edit command
The edit feature only changes the erased areas. So I think you probably selected the wrong one
Was that the problem?
the problem is, its totally changing the bracelet, I did not erase any part of it, I want to keep it as original, just put on a realistic background
makes sense?
If you send me the image I can attempt to do it. Like I said, this only happens in the variation feature
Prompt?
put on a dark brown wooden surface with realistic shadow
Its adding more bracelets. Thats the same image but with added bracelets
how to overcome this issue
its a simple task for an AI like Dalle
I mean if I'm being perfectly honest
It would probably just be easier to generate the background
and lay the image over it
shadows are easy enough with photoshop
I am a designer, I know this stuff, the point is using AI to help with this
I understand. I've tried 6 times, with the same result. I would put something in #bugs-and-issues for now
after 6 attempts I was able make this one work
here is the original
Hey! 🙂 Anyone know why Dall E says "invalid URL" when given my payment method for credits?
same billing problem
Yes
Ah, dang... okay. Will check back in later -- thanks for the answer!
Dang
And what if I were to use a generated image as a reference from which I made a sketch which I then expanded and projected onto a canvas and then painted? If so, then what about if I significantly changed the image?
How about, for the sake of argument and a bit of reductio ad absurdum (though I am honestly asking where the threshold is, I'm just making the example elabourately specific) if I printed it out, converted it to a black and white Xerox effect, added text, made it into a flyer for a punk concert, hung it on a wall papered with a bunch of other such punk concert flyers such as you tend to see in a punk club, had someone pose dressed up in cyberpunk looking clothes with a pair of rubber vulcan ears holding an 80s casio keyboard that's been painstakingly modified into a cosplay replica of a first-edition Shadowrun Cyberdeck, took a photo of that person standing on front of the wall with the flyer completely visible, and then traced and inked this to look like a 1990s Tim Bradstreet illustration? Do I have to claim that a portion of that image was created by an AI and/or identify what portion?
Wait... are you saying we can use Photoshop or something else to make a partly transparent PNG instead of having to use the giant-fat-eraser-tool and that works, too?
Like... could I do a render in Poser or Blender with a depth mask, set the depth alpha channel to a threshold and adjust it to remove either the background or foreground, and then have Dall-E 2 generate the missing parts?
I've avoided experimenting with that feature because frankly your eraser is terrible. It doesn't track a stylus, it's unwieldy, and it's literally only brush mode. No other selection options.
At that point, I could even, in theory, generate an image, mask out the foreground to transparent, have Dall-E replace it with described text, save that, use the original alpha channel inverted to mask out the background, have Dall-E replace that, and then have really good figures. I could even do this figure by figure. But not with the web-based eraser tool. That's just unusable.
Hey there, I'm trying to buy some more credits, but I have this error "Invalid URL (POST /api/labs/billing/credits/purchase/preview)" and it won't let me use my credit card to buy credits.
You probably want to scroll up and read back a little first.
There's a tonne of people having that issue (also, I would guess #bugs-and-issues is better for that anyway)
ah thanks
no worries
I guess they don't want my money
It just came up in the last hour or so - hopefully it's shortlived!
Invalid URL (POST /api/labs/billing/credits/purchase/preview)
Whether I do it on mobile or desktop it doesn't work?
@short salmon
Thank you very much for your reply
i was just trying to buy some same thing happend to me just now lol
does anyone know why words will sometimes appear in an image? like for example, here i simply used the prompt "sleep paralysis" and all of the images had some sort of text on them
Its happened to me as well. Pretty hard to predict why its happening. Sometimes its fixed by rearranging the prompt structure
i see, thank you!
Lol thankfully it isn't just me
I was about to ask why mine isn't working too lmao
<@&998386466063396966> what's up with this? unable to buy credits, does the same thing on mobile
I'm having that problem too, hopefully it's fixed soon
we'll let OAI staff know - please also contact them via support@openai.com
👍👍
Which is the banned word/words in this prompt? "pyramid head, digital art by Junji Ito, esoteric, memetic, optical illusion"
I'm still wondering if that's what was being said.
They have been asking people to put inexplicably banned things in #suggestions if that helps.
oh ok, thanks!
hey quick question I just uploaded a picture of a video game screenshot, am I breaking a rule there? I totally spaced on it until after I generated variations
Is anyone else unable to even use the eraser when trying to edit a photo? I haven't been able to update any images I've uploaded yet
Please don't post IP that could be subject to copyright. If unsure, err on the safe side.
Junji Ito is associated with horror genre. Maybe he is the problem. To determine which word is the problem you can test one term at a time to see what fails. But there is also context analysis happening.
Hey where should I go if I have problems paying for extra credits?
Hey, there's a temporary workaround in #bugs-and-issues
lool just spam click huh. ok ill try it!!!
I've tried him with other prompts with no issue, so I don't think it's him
will I risk getting banned if I try all the different words?
What was the most recent version of the prompt that failed? Let me risk losing my account with a test. I am suspecting "head" with "Junji Ito".
pyramid head, digital art by Junji Ito, esoteric, memetic, optical illusion
Also, help me to visualize what you mean by a "pyramid head"? The image is unclear and has no usual referent, so the AI may be getting "head" out of it.
it's a character from silent hill: https://en.wikipedia.org/wiki/Pyramid_Head
Pyramid Head, also known as "Red Pyramid Thing" (赤い三角頭, Akai Sankakutō), "Red Pyramid", Sankaku Atama (三角頭, lit. "Triangle Head") is a fictional character from the Silent Hill series of survival horror video games published by Konami.
Introduced in the 2001 installment Silent Hill 2, they are a type of monster that serve as main antagonists as t...
OK, I tested it in stages. "mimetic" is the problem, apparently. If you think this should not be banned, you can make that suggestion in #suggestions. This prompt works: "pyramid head, digital art by Junji Ito, esoteric, optical illusion"
thanks a lot! I’ll leave it in suggestions
I just did a search here on Discord and could not find a previous report related to the word "mimetic", so this is a new one for both of us.
“memetic” actually, I think they mean different things but both shouldn’t be blacklisted in my opinion
I am more familiar with the art term "mimetic". Sorry if I misunderstood. https://www.merriam-webster.com/dictionary/mimetic
I was able to find one problem report for the spelling "memetic", so both spellings fail: #bugs-and-issues message
oh, hopefully they update their blacklist soon yeah 😅
Are the faces generated by AI, really generated by it or does it search for similar images on the internet and show? if it's generated by AI does that mean that person doesn't exist in the world!!?
It is generated by the AI, yes. Maybe a few similarities here and there but no such person exist in the world!
Damnn it just generated a very realistic picture of Woman, and the fact that she doesn't even exist is blowing my Mind! what a time to be alive!
Check #photography for more photorealistic generations!
Sure, Thanks Mate!
There are cases where the prompt gives you essentially the same person with variation four times, so I do wonder whether it is strongly influenced by a single image of one person.
That's probably limited by the dataset. Happens sometimes to me as well.
uhhh, can i make phone wallpapers somehow?
Yes, check the following channels: #inpaint-outpaint, #dall-e-bot, #—post-dalle-edits
and just in time for today's #daily-theme — wallpapers.
Hey friends, does it cost a credit if your prompt is rejected by the content filter?
Apparently it looks that way, but if you refresh the credit comes back.
Can't vouch, my prompts are eternally pure 😇
😅 thanks! I asked for a 'seductive smile' and was struck down for my vulgarity.
went with alluring instead
No - I access it through 4G often. What kind of error are you getting? It could be an issue with your telephone company I guess...
page just wont load, no error or anything
I mean, after a certain amount of time, it says "this page took too much time to load"
Tried clearing your cache?
lemme try
Just restarting the browser sometimes fixes that white screen error for me too.
still don't work on any of my browser sadly
What were the browsers that you've tried it on, and have you tried using a VPN?
brave, chrome, edge
works well with a vpn tho
didn't think about trying with vpn
Huh, well there we go 
Happy creating!
thank you!
If the Dalle develepors team could develop a 3D diffusion ‘noise box’ instead of a ‘noise plane’ for Dalle to work in? THEN Dalle could perhaps generate 3D models?
Perhaps! Can as many 3D models easily be sourced for data input as we have 2D images? And is the captioning on those models robust?
Also, can we afford an exponential increase in compute cost for the additional dimension?
(not rhetorical, it's an interesting question! maybe one for #ai-discussions though?)
I'll post it there 👍
how do I generate "outpaintings"?
Take a look at this :)
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y Page 59
(I would recommend you to have a look at the whole topic starting page 55)
Anyone know why this prompt generated 2 and 4? :S
Hmm. Could be the antibias measures acting up. Report as "Image doesn't match my text description."
ah okay, i will do that
Hey everyone, I'm trying to hook up the DALL-E 2 API to some code I'm writing like I've got with GPT-3 . Is this API yet available to the general public?
Q: Is there a publicly accessible API for DALL·E?
OpenAI: We do not at this time, but do have a longer-term plan to offer this. Please also do not attempt programmatic access of DALL·E, which is against our Terms of Service.
No, right now there's no API available to the public.
They do plan to introduce that in the future though :)
(or see bot message above lol)
Any ETA on this yet?
probably been asked a million times.... is there not a Hi-Res export option?
No ETAs for now.
Currently there isn't, no. However, you can "combine" them :)
Have a look into #inpaint-outpaint & also at this
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (the section starting page 55 is what I'd recommend you to look at)
What are the goals for Dall•E by time it comes out of beta?
Hi! anyone that has created infinite art using dall-e and have some tips on best practices?
Infinite art?
Infinite zoom art,
Ah, I know Nightcafe.studio can do this
They make a 5 second gif that zooms in and changes the NFT art
oh cool! I have done it in Dall-e but not sure which software to use to optimize the image in the best way
Is that what you meant ?
will check it out!
Yeah I know people use Dall-e for the design and then edit the images in photoshop and after effects
I just created a picture in Nightsafe now but its not zoomed out? its just a regular output like dalle creates
I’ve been getting great results without need for photoshop tbh. Not yet
Let’s go to the right chat
zoomed out? how do you zoom the image out without doing it in another software
which one is that, I just joined
This is question it’s okay
So first I would go to DE2 and get the image you’d like, then go to NightCafe and expand the options
Then you select “create video for 3 credits”
You upload the image from DE2 to Nightcafe and then write in the same prompt, just expand options into advanced mode and select to create video
That’s the best advice I can give
I will defo try thst out! is it worth paying for nightcafe?
Yes I’ll send two pics
I was wondering about that a while ago. It might be possible, but I am not sure whether there are already fundamental limitations due to the increasing degrees of freedom.
A diffusion process in higher dimensions has qualitatively different properties. So I would wager that you could engineer it in 3D, really interesting would be something like 100D
yeah there is so many amazing zoomed out artwork out there, and also already existing famous art that has been expanded using dall-e. really cool
Non-DALL∙E 2 generations have been cleared from the channel. In the future, use #ai-discussions. Thanks!
Discussions of other AIs are permitted in #ai-discussions.
As a community, we're excited to see new models and platforms for generative art emerge. We value dialog about the future of AI, including alternative technologies and their associated ethical issues.
Another question : Can I sell the images I have generated if they are variant of an image I don't have the rights to (variant of a google image)?
No, as per the Terms of Service, any violations of the Content Policy will revoke all your rights to the image. Here's an excerpt from the Content Policy.
Ok thanks
You're welcome!
My apologies Major
When do we envision an API being available? I’d like to open this up to differing marketing departments within an organisation but keep the costing centralised.
When I want to talk about new prompts what’s the best channel
No worries!
Depends, but if it's got anything to do with prompts, #❌┃prompt-help is your best bet.
Ah, that's for #❌┃tips-n-tricks then 👍🏻
Could I show you something there?
Hi! No announced timeline yet for this.
Q: Is there a publicly accessible API for DALL·E?
OpenAI: We do not at this time, but do have a longer-term plan to offer this. Please also do not attempt programmatic access of DALL·E, which is against our Terms of Service.
Sure thing, just tag me.
Thanks for the feedback
i JUST made an account and made some things, is there a way to see your "render history" or something like that? i see it when generating new stuff but not as a plain option somewhere else lol
wait no gosh im stupid nvm found it
Also keep in mind that you can always come back to your images if you have the URL, even exceeding the original 50+ limit! :)
ah thanks! will keep in mind
Hey I asked earlier, but how would I go about describing the idea that a setting is post-post-apocalyptic? (The best description of the genre I’ve found is https://www.reddit.com/r/worldbuilding/comments/3b01i2/ )
51 votes and 32 comments so far on Reddit
I think just "post-apocalypse" is still the best way. Anything that happens after an apocalypse is still just post-apocalypse. E.g. John Wyndham's The Chrysalids is a formative post-apocalyptic text and takes place in a relatively burgeoning community after a nuclear war.
I guess so. It's a small, relatively unknown subgenre for the time being
I guess I just am going to have to find a way to combine “in the style of the game ‘Fallout: New Vegas’” and “screenshot from the game ‘Horizon: Zero Dawn’” or something of the like
If I asked for 'alexander kissing hephaestion', that's not ok right?
but 'alexander embracing hephaestion' would be?
(... asking for a friend)
Is there a way to get credits cheaper?
No, credits for now will always cost $15 (taxes not included) :)
That's simply the way OpenAI decided to price their service,
if you have some suggestions please do that in #suggestions 
And is there a subscription for about 5 dollars a month and unlimited images?
or a flatrate
Nothing of that sort for now, no.
Who knows though, maybe OpenAI will release something of that sort in the future :)
i hope so
$5 a month would probably not get you many generations though
The overhead to run these servers must honestly be quite high given number of people using it and the gpu load
Is there a way to view the list of all the images I have created since I first started?
Only if you saved things to your collection as history only does the most recent 50 prompts
but you can go back thru browser history and find the links
You can check your browser history for urls with labs.openai.com/e/
Your creations will still be there, even exceeding the original 50+ limit! :)
even better tip ^
The question must have been answered millions of times, but can we use a Dall-E generated picture as a pfp?
Yes!
Question: Is there something you can put in your prompt to force the whole character-thingy to be inside the borders? I have some images that I really like, but then the tail of the fish is cut of by the border or something
is there a way to transfer your account to a new email / org?
The best way to do this would be by reaching out to the support directly 
E-Mail: support@openai.com
Live Chat (available from Monday - Friday, at least I think so): beta.openai.com
ty
Depends on the prompt... Some people had success by adding something like ", wide-shot or ,medium-full-shot
or also playing with the camera lenses such as ,14mm ,24mm etc.
Have a look at these for more info
https://dalledictionary.notion.site/Camera-Settings-c79972354eb94a1682709749b3a24929
&
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (relevant topic starting on page 17)
You could also try and outpaint it, for this take a look at
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (page 59)
Hope this helps 
Is there an API for Dall-E? like there is for GPT-3?
Hi, I am not sure if what beta I'm currently using right now, whether it is DALL-E or the DALL-E2? Mine is there is no 2, only DALL-E. Please enlighten me. Thanks!
The invite you received from OpenAI is for DALL·E 2. The first version of DALL·E, version 1, was never released to the public.
Read more here: https://openai.com/dall-e-2/

Yeah its a good bot
yup, it is hehehe
And the cool part is no human is involved 😌 /s
how do i get help logging in... Was just fine this morning now cannot log in. Day 1
Best bet is to contact support@openai.com
No, as of now this is not possible. Maybe in the future though :)
im trying to help some guy edit a pic of his cat and i dont know how to use the edit feature can someone help me with this ?
Sure, what exactly is it that you need help with?
Or is it the whole tool itself?
well the guy wants like a pic of his cat with a small tie i guess
i have a pic of his cat already in the edit tool
Could you share the picture of the cat, that way you won't waste any credits :)
also, let's move this to #❌┃prompt-help
sure
Hey guys, when I make a picture often the top and bottom will be cropped off. Seems to happen very often when the focus of the image is some person/creature, the top of their head or their legs/feet/whatever is at the bottom gets cut off. Any ideas why/how to avoid it?
Hiya! 
Some people had success by adding something like ", wide-shot or ,medium-full-shot
or also playing with the camera lenses such as ,14mm ,24mm etc.
Have a look at these for more info
https://dalledictionary.notion.site/Camera-Settings-c79972354eb94a1682709749b3a24929
&
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (relevant topic starting on page 17)
You could also try and outpaint it, for this take a look at
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (page 59)
Why would I get this?
Hi guys, any idea what causes the policy violation? Happend with tigers too.
Hi, I made a video on how DallE 2, Midjourney and DD are already changing our jobs as artists, is there a channel in the server where we can share content like this? It was 6 weeks in the making and the response so far has been really good but I would like to share it with this community!
sooo why cant I make Raindrops on Glass? xD
think there may be an issue atm, even the "surprise me" prompts are triggering content policy
cant you upload it to dall e and let dall e create variations?
I believe there is an issue right now. Lets wait it out and see what happens. I am getting many warnings myself
surprise me... BAM content strike 
Discussions of other AIs are permitted in #ai-discussions.
As a community, we're excited to see new models and platforms for generative art emerge. We value dialog about the future of AI, including alternative technologies and their associated ethical issues.
Yes. You’ll get banned and probably won’t get un-banned
Flag any content that isn't accurate to your prompt to help Dall·E 
And if you try to create something that is against the ToS on purpose
Dall·E will (thankfully) filter it out.
You're also risking getting banned, so just don't :)
how to make image extension?
Have a look at these resources :)
#top-tips-and-tricks #inpaint-outpaint
#❌┃tips-n-tricks message
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (starting page 63 should of interest to you)
thanks!
Does anyone have AI tools to smooth out lines from art pieces you got from Dall-e? sometimes they come out a little jagged/shakey
Disco/Latent diffusion (?)
Did you get help for this? I was working on it and then my pc crashed and I’ve been trying to fix it ever since
by itself, it doesn't generate very well, but if you add the start image (for example, the dall-e output), it improves it
This principle uses in Dall-e Flow
awesome! Looking into this now, thanks!
Hey there, when I first loaded up I thought I saw something talking about free credits every day? Did I dream it or is it an actual thing because its been a few days now and I haven't gained any credits
• You get 50 free credits your first month.
• 15 free credits will replenish every month after that, on the same day of the month.
• For example, if you signed up on August 3rd, your free credits will refill on September 3rd.
• If you joined on the 29th, 30th, or 31st of every month, your free credits will refill on the 28th of every month.
• Free credits don’t roll over to the next month, so they expire a month after they were granted – but you’ll get 15 new free credits.
👍
eesh, every month!
Yepyep
oh lord long time
Free credits were a thing once, that changed though :)
Q: How DALL·E Credits Work
• You get 50 free credits your first month.
• 15 free credits will replenish every month after that, on the same day of the month.
• For example, if you signed up on August 3rd, your free credits will refill on September 3rd.
• If you joined on the 29th, 30th, or 31st of every month, your free credits will refill on the 28th of every month.
• Free credits don’t roll over to the next month, so they expire a month after they were granted – but you’ll get 15 new free credits.
Wow, you think its going to stay like this?
We don't know until OpenAI announces something 🤷♂️
Imo they will change something about it, that however is
just what I suspect. (I'm in no way associated with OpenAI) 
I don't know if I am allowed to mention ||midjourney|| but they have a subscription based service which I think is a better idea imo
You are, no worries! For suggestions head on over to #suggestions and post it up there :)
If you want to discuss AI, or other companies etc. #ai-discussions would be the right place.
Great!
Hi...so how do I get it to stop cutting off important details in generation?
It’s difficult. Your best bet is to try to use specific camera angles in your prompts. I suggest taking a look at the DALL E prompt book in #gpt-models
Yes this will be helpful thank you
this did exactly what I needed! Disco was a little confusing to get started but got it sorted, thanks mate!
no problems)
is it down?
Looks like it yes, not sure for how long but I doubt it will be long
Are monsters and warriors aloud? I have a lot of Dungeons and dragons type things I Wana try out
i do believe. On the site it has some TOS, i think specific violence and politics and celebrities can't be used
looks to be back
Holding a weapon = violence?
To give a example: I type monster descriptions instead of there named
u could try a credit and see if it works. It'll let you, and i dont think there's any penalty. I've probably accidentally typed "no go" content like 15 times and had no issue. Cuz u kinda learn as you go, like the AI
Such as instead of beholder I type: A large creature that consist of a floating head with a massive central eye and sharp teath, eight eye stalks protrude from its head looking in different directions as it drools facing you
Ima try it but have someone spell check it
hello, in this use case what the prompt should describe? the whole image? or only the part i want to be edited?
Hey everyone
These images aren’t following the prompt, “A puppet being helped off the strings, detailed matte painting”
It doesn’t come out as a matte painting. What should I do?
@verbal crag# 0001 is nobody on?
Is dalle in a “baby phase” during this beta? It seems to spell like a child
Maybe it's one step ahead and tried to make a kind kingdom 😜
What if you drop the comma and say “A puppet being helped off the strings, in the style of a matte painting”
Or put matte painting, a puppet.. etc
I find rearranging the order of operations or substituting some words helps
You should describe the entire image :)
Would images from Artbreeder fall under the "too realistic" category?
Specifically the portraits model they have
Is there any reasonable way to avoid returning any images with 'text'? I'm automating bulk searches, but anything with the glyphic text would be unusable to me. Cutting it at the source would be great
Is dieselpunk not allowed? I tried doing a Dieselpunk Texan Soldier and it said it was a violation
I think it was the word "soldier".
There's a a lot you can do with this tool but there's also a lot that you can't do, as well.
hmm idk. it did some revolutionary war soldiers for me yesterday
Really? Hmm... the detection was on the fritz earlier apparently, so it could be a remnant of that? Either that or "diesel" is triggering it's political filter.
Either way, I'm sure there are some synonyms you could try
probably. i tried steampunk and it just made a dude in a brown suit...🙃
Dall·E struggles with the latin alphabet quite a bit.
Words with 2-4 letters sometimes work, most of the time though
it will give you gibberish (or Dallish as some call it) 
Genuine question... How did I violate policy here? I've gotten close to a dozen policy violation warnings for 100% innocent requests and I'm worried my account will be locked. Does anyone else struggle with this issue? Is there any way to avoid it?
Thank you friend
Instead of underwear, try underwater :)
Oo
Dying laughing. I did not see that autocorrect
🙄
Thaha, can happen sometimes 
Lol. I used to edit books. That’s priceless
Imma leave this interaction up for anyone who needs a laugh
About to build a website and instead of stock images I’m using dalle 😎
Q: I just got a content warning! What does that mean? Will I get banned?
• Please don't worry if your prompt clearly doesn't violate the content policy, but has been flagged by our system!
• We chose to err towards our classifiers/block list being aggressive to start, but have been constantly iterating on them as we learn more. Please share prompts you don't think should've been banned in suggestions so that we can improve.
• While we're not sharing the exact thresholds, a few warnings here and there won't lead to an account suspension.
does anyone know how to make animation?
Are you referring to this? :)
https://www.youtube.com/watch?v=tux81Z3uXMY
In this tutorial I will show you how to make a multi-layered infinite zoom using DALLE, Photoshop, and After Effects.
Audio by Tyrone Estephan
Chapters:
0:00 - Intro
0:18 - Generating images with DALL•E 2
3:27 - Fixing images midway through
3:59 - Generating the rest of the images
5:24 - Setting up the comps
6:08 - Nesting the images within ...
What can I do to send DALL E 2 or its Staff Data on an Art Style?
Take a look at this 
#gpt-models message
I know we can remove the watermark now, but can you help me understand this? "You must clearly indicate that images are AI-generated - or which portions of them are - by attributing to OpenAI when sharing, whether in public or private."
Are there some examples of attribution, like a photo shared on Instagram? I assume I'll use the description, not text overlay on the image itself. For example, if I share a work made with the aid of dall-e, it's enough to say that there was an AI generator in the process? Should I explicitly make the name of OpenAI and how? Saying Dall-E 2 (or #dalle2) without "OpenAI" will count as correct attribution? Sorry, but I found it a little unclear. Thank you!
Stick true to the policies at all times
Content Policy: https://labs.openai.com/policies/content-policy
Terms of Service: https://labs.openai.com/policies/terms
Removing the watermark is still a grey area in the ToS, so be mindful of that and decide for yourself.
For the time being, you should be fine as long as you clearly state that the image has been AI generated (for example, in a description or as a title) and give credit to OAI/DALL•E 2. I'm saying should because, as I already mentioned, it's still a grey area legal-wise.
(I'm in no way associated with OpenAI. Please make your own decision after reading through the policies)
Hi everyone! Can I legally use an art generated by me in DALL-E as an album cover for my music release on streaming?
Yes!
This may sound like a dumb question, but can I specify the Dimensions of generated art? I ask because I want to generate a flat Wood texture with the same dimensions as a Gottlieb Widebody Playfield.
Currently that's not possible. The images will always be 1024x1024 px. However, you can "combine" them :)
Have a look into #inpaint-outpaint & also at this
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y
(the section starting page 55 is what I'd recommend you to look at. Page 63 is the topic combining specifically)
No dumb questions here, dw! 
How can I access prompts older than the last 50 prompts from my history?
Is everything older than the last 50 prompts automatically deleted and not accessable anymore if I didnt save it or add it to my collection or anything?
I paid for the credits and its gone unless I each time click on save for each image???? 😳
For now, your older prompts are available via your browser history.
How does that work?
You can check your browser history for urls with labs.openai.com/e/ or labs.openai.com/: )
In your browser there should a Browser History option.
how do I do outpainting?
Have a look at this 
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (page 59)
Thank god its not gone at least. Will be more careful in the future. Thanks for the quick help. It works! 🙂
You're welcome! :)
will I lose resolution if I do outpainting?
will it always give the result in 1024x1024 pixels
The end product will always be 1024x1024 px.
The more outpainting you do the more quality will be lost for the very first picture as it will continue shrinking :)
I see, thanks!
fs!
Hi all just question, anyone knows how midjourney is generating their images? Are they using DALLE 2 as well, or it thier own system? Im pretty blown away by their images.
1 more qns, will dalle2 have its own API as well? Looking forward to building apps wit this tech!
Discussions of other AIs are permitted in #ai-discussions.
As a community, we're excited to see new models and platforms for generative art emerge. We value dialog about the future of AI, including alternative technologies and their associated ethical issues.
Can someone fix the eyes and fur on this one so it looks more like the original picture?
Original:
Hey babes, tons of info here:
https://dalledictionary.notion.site/Inpaint-Outpaint-3b3a6ec7964f44cfa257f047f45ed400
thanks babe
Heyo,
Can somebody tell me about these questions:
What is the license of the generated Images? Can I use them and change them? How is the ownership?
I couldn't find this in the Sharing & Publication Policy.
Thanks!
You'll find it in the Terms:
Content Policy: https://labs.openai.com/policies/content-policy
Terms of Service: https://labs.openai.com/policies/terms
API Terms & Policies: https://openai.com/api/policies/terms/
Point number 6 🙂
ooh, I see
That's why I didn't find it
Thanks for your help! :)
So according to the terms I am not allowed to sell my generations, correct? I just read an Article from last month that new Beta users were invited with full commercial use.
From OpenAI's email:
And this 🔽
Ah I see, got it, I'm always a little unsure with these terms and what exactly they mean, thanks for the clarification!
The terms are being updated AFAIK. So, do check once in a while if there's any change 🙂
are weapons not allowed to be prompted ?
I'm very confused as to what is and isn't allowed
specifically why is this prompt blocked ? "woman wielding a two handed long-barreled firearm, medieval fantasy, digital art"
Basically anything that would violate any of the policies.
So, a sword simply on display is fine. A sword in the hand of a person swinging it at someone would not be fine.
Guns are completely banned as far as I know. (correct me if I'm worng) 
Weird restrictions but I guess I can understand why
So far I've only typed prompts, but is there a way to upload a picture and have it recreate it in a different style? I think I've seen this done but I don't know how to do it
How can I download ALL images that I generated using dalle, including images that re not saved in a collection?
Hello)
Hey, I got a question for all y'all.
Has anyone redrawn the creations they made with AI Art Generators?
Yes! Some have posted their recreation in #—entirely-human-art
Write a script that'll go through your history and download all the pictures
That's the only way I know of that could download all the pictures
Excellent.
Well I know what I plan to do.
Is there an official DALLE discord bot that we can invite to our server? Midjourney recently introduced that
No, Dall·E is only available through the official website :)
Thanks
Can we use the generated stuff for personal use like profile photo etc.
Yes you can do whatever you want with your generations as long as you credit DALL-E for it in some way
Easiest way to credit DALL-E is by adding a watermark in the #dall-e-bot channel
Or, like many people here, you can credit DALL-E in your bio
Does Dall-E only generate 1:1 ratio images?
Yeah, you can "combine" them though :)
Have a look into composites & also at this
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y
(the section starting page 55 is what I'd recommend you to look at. Page 63 is the topic 'combining' specifically)
oh nice! thanks for this
how do ppl extend images like album covers using dalle
The method you are referring to is called 'Outpainting'.
For more info take a look at this :)
https://dalledictionary.notion.site/Inpaint-Outpaint-3b3a6ec7964f44cfa257f047f45ed400
&
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (page 59)
What does Dall-E flourish at doing?
Generating art
ty sm 😃
Eh.. What kind of art? Does it do well with faces, landscapes, drawings? Compared to other AI out there like MidJourney, I'm not impressed with Dall-E at all
I'm trying to figure out what it's best at, to try and utilize it's ability to the fullest.
Digital art
Try giving it something simple, like a two word prompt, and adding “, digital art” at the end
So in other words, any prompts for someone Parrying an attack is not allowed? Just curious.
Dall E - Single Generation - "midsomar angel furious and beautiful 35mm digital art"
Try checking the other channels here. I suggest #photography, then you can check the rest.
Discussions of other AIs are permitted in #ai-discussions.
As a community, we're excited to see new models and platforms for generative art emerge. We value dialog about the future of AI, including alternative technologies and their associated ethical issues.
You can always give it a try, to my knowledge however anything that involes any violence will get blocked...
it really depends on the prompt though. Fencing as an exmaple is banned, even though it's (nowadays at least) commonly thought of as a sport :)
Hey I tried including “xenomorph” in my prompt because I wanted it to generate an image of a xenomorph stalking a human on a spaceship and it gave me a warning. Is there any way to get around this?
I feel something like this may require you to be creative. Human with the shadow of a xenomorph on the wall, maybe?
I've used xenomorph before and didn't get a warning. Maybe it was another word in the prompt?
anyone know a way to get clean line art logos?
is there a way to make the image from Dall-e2 bigger?
The images will always be 1024x1024 px. However, you can "combine" them :)
Have a look into #inpaint-outpaint & also at this
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y
(the section starting page 55 is what I'd recommend you to look at. Page 63 is the topic combining specifically)
im confused lol
I think you’ll also find this helpful 
Thank you!!!
I was trying to generate images based off of Asimov's short story "the last question", which is about people trying to get an AI to tell them how to reverse entropy. I gave a prompt "Mankind builds a machine to reverse entropy and bring dead stars back to life" and it got flagged as a content policy violation, not entirely sure why. Does it not like the 'dead' there?
The word dead probably is the violation.
Ahoi, quick question: is it allowed to share my account with a co-worker working on the same art project? would log in from a different country. ToS and usage policy doesn't mention it - but holy moly will i do everything to not loose access / get banned 🙂
That may be violating these:
More info:
Content Policy: https://labs.openai.com/policies/content-policy
Terms of Service: https://labs.openai.com/policies/terms
API Terms & Policies: https://openai.com/api/policies/terms/
I would personally not share the account for the moment, specially from a different country.
Hello. I want to edit a picture but it keeps saying that its against their policy. I dont see whats wrong with the image.
Can I see the image?
Should i dm it
Sure
is there any way to get the results without the pallete in the bottom corner?
Officially, no. But others download the raw .webp file with their browser's inspect option.
thanks, the watermark was breaking a part of the image when i tried to run the cropped image on the AI
Thank you very much for your time looking it up - will def. follow the recommendation 😉
When will the DALL-E api be available to developers?
Q: Is there a publicly accessible API for DALL·E?
OpenAI: We do not at this time, but do have a longer-term plan to offer this. Please also do not attempt programmatic access of DALL·E, which is against our Terms of Service.
This doesn't give the timeframe / expected release date
That's because there is none.
Simply sometime in the future :)
when my 50 credits are out, will i never get to use it again unless i purchase more?
No, you will be getting 15 free credits each month irrespective of whether you buy additional credits.
oh so if i dont use it for months than i will have more than 50 credits laying around? because each month i'll be getting 15 more
Negative, free credits do not stack and will expire each month on the day your account was created.
for example: if you made your account on the 8th, every 8th you will get 15 credits and any existing free credits will expire.
Do keep in mind that credits you purchase are valid for 1 year from the date of purchase.
How do I get the AI to do clay renderings? Did it do paint-ish because I included digital art?
account was created as in after i got accepted from the waitlist, or when i made an account to to join the waitlist or when i actually started using it? (im dumb sorry😭)
I am not sure if I can help you out with this specific prompt but I do suggest checking out #❌┃prompt-help and the DALLE Prompt book at: https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y for more help with prompts.