#community-help
1 messages · Page 8 of 1
I did all the obvious things. Desktop mode, messing with user agents, etc. The problem is, I think, that it just doesn't appear below a certain size. And it's not delivering a mobile version, it's delivering a scaled, reactive desktop version. Most of the time this is good. In this case it is not.
What's the difference between downloading DALL-E vs using it on the browser?
The biggest difference is that when you download using the Download button, there's a watermark on the image. While it takes extra steps, downloading using the inspect feature to find the URL allows you to download the image without the watermark, albeit in a less than ideal WEBP format
Yeah nah, it would not turn out as great.
Dall·E will currently (by default) generate every image in 1024x1024px,
which is roughly 27cm or 10.6 inches.
You can attempt to upscale the image with a programme like Gigapixel, but most of
the time the outcome won't be good enough to print at such a large scale.
Printing at the default size however, perhaps scaled down slightly, works great though!
Some people even have already printed their generations on T-shirts or sweaters e.g 
Can you ask for different aspect ratios to the AI kinda like midjourney?
No, for the time being the images will always be 1024x1024px in size by default.
You can, however, 'combine' or upscale them!
Have a look into #inpaint-outpaint as well as these resources!
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y by @spring wyvern
I'd recommend you to look at the whole section beginning on page 55.
The topic 'combining' specifically is on page 63.
This whole notion-site is extremely helpful!
https://bit.ly/dictionairi
https://bit.ly/dictionairi-in-outpaint <- For In/outpainting
https://bit.ly/dictionairi-upscale <- For Upscaling
by @rugged wing

ty
Is there a site that expands images for me since it’s harder on mobile?
Nvm, you can do this with the bot in #dall-e-bot
Is it possible to "pan the view" of a generated image upwards? I am having issues with DALLE 2 cropping heads off my subject matter on some really cool images.
Hi - This happens due to the square shape of the outputs compared to the format of some of the trained images. You can follow the instructions Smile posted above to add to the image and work towards completing it to your liking.
Oh thank you! I'll give that a try!
Has anyone found an alternative to Tencent Arc Face Restoration? Or could anyone confirm this works for this? https://arc.tencent.com/en/ai-demos/faceRestoration
It hasn't worked for me since at least beginning of August, and I haven't found a good replacement for it, either.
Hello I just got access to dall e and was wondering if this access will end after some time or is it unlimited?
I shared my first image, how do I see that image and the view response? Also, If someone sends me a message answering a question or making a comment, where do I see that message?
The "Share" button creates a link that you can share to allow other people to view your image. There's currently no comment system. It simply links to the image.
Once, you click "Publish" it will look like this
ok, thank you I havn't seen that when I shared
I don't have an answer to this for you right now. If you give me some time I can look into it further
I will look on the Dalle website
Is there a statute of limitations on the real person rule, like could I do Abraham Lincoln because he lived so long ago?
Are there any rules/legalities I could be linked regarding how this program may be used for productions, creative projects, or sales? Is the art under creative commons licensing, and does the user payments alter this in anyway? I'm curious the limitations of usage regarding saved work.
You have full commercial rights to your images
The terms of service will have more information
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'm also curious - and understand if it can't be revealed - but what resources/data does DALL E use to generate its CLIP? The wikipedia page just lists "public datasets" without any specificity. The reason I ask is that the results I've gotten from some queries/prompts have been all over the place. I asked the gaudy question out of pure curiosity "what is the meaning of life?". The images looked like they came from a self help website designed by Salvador Dali himself LOL
How do I use the dall-e-bot channel to resize an image?
Question: what is the acceptance rate of access to dall-e? My friend are trying to join
Sorry new here. I could not find how can I communicate with Dall-e and spending my credits. And sorry if it was answered before, I searched it with a few keywords but could not find it. I am trying to make a skill icon (Skill icons like Divinity original sin, grim dawn or any mmo's) with Dall-e. These were the results I've used (first to last)
Fast arrow, skill icon
Fast arrow, rpg skill icon
Fast arrow, rpg skill icon, with background, colored
Fast arrow, rpg icon, with background, colored
32 bit pixel art of a fast arrow icon, very detailed, artstation, 4k, 16:9
What should I try?
is it possible generate non-square photos for example 1200 x 1000?
Not at this time
Can someone answer this?
While it is not possible to create images other than in a 1024 x 1024 aspect ratio at this time.
However, you can upscale or combine images! I do suggest you try the tools listed over at:
https://bit.ly/dictionairi-upscale
https://bit.ly/dictionairi-in-outpaint
by @wheat zodiac .
DALL·E currently generates 1024x1024 px images.
There are many AI models that can help upscale your images.
Some are paid, some are free. You can find a couple of links here.
I don't think OAI has shared any information on this, however from my experience, I have friends who got theirs accepted in a few days (that's how it has been for a few weeks now). It used to take months back when I joined.
Ohhhh ok
I need to change my pfp back lol
Extra points for something generated with DE2 😂
Lol
My old pfp was my YouTube logo
And I try changing it to something that was make with dall-e 2
ahhh, if you are using a DE2 image I suggest styling it up using the !watermark command over in #dall-e-bot.
upload the image and type !padxx where the xx is the percentage, eg !pad50
I did that
I have 50 credits free for the first month. Are they cleared, when the new moth begins?
• 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.
There we go, a full explanation from @pine void!
how to buy new credits if i live in central Asia?
Contact support@openai.com The only other way is something like revolut, but I don't think that's available in your region
thx
What the hell caused this to trigger the content filter?
Probably "juvenile" and/or "adult"
Looks like it was juvenile, thanks major
How many people got accepted into the beta?
If I want to sell a Dalle-2 picture, can I remove the logo or is that against the agreement?
No official numbers yet, but OpenAI has announced their goal is to invite 1,000,000 people to test DALL∙E
Have a look at what the product manager of DALL∙E said: #community-help message
Just a note, the Content Policy still states that you'll still be needing to disclose the role of the AI in creating the image. More info in the links 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/
Hey guys, I have 2 important questions before starting my prompting experiments. 1: Is it possible to set an aspect ratio, for things like portraits or match an A4? 2: How do I keep the subject in frame, so nothing will get cut off?
- Hello! Currently you can only generate 1024x1024 images.
- Some have been successful using modifiers such as "center shot", some used a black frame. Do check out #❌┃tips-n-tricks and the resources over at #gpt-models for more info!
Thanks!
Hey! I have 1 question. Am i get new credits every month?
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.
Hello, I've tried some upscaler but anyone has some problems. Why don't you add an upscaler directly in the app?
What do you use?
Are images immediately saved to my collections, if not, how to save to my collections...? Thanks.
Hi there! Images aren’t saved automatically to your collections. To save them, click on a generated image and press “Save” in the top right corner.
Yes!
Use !watermark in #dall-e-bot if you want to style it up with a watermark too...
Real = Not AI Generated
How do we erase watermark for a padded dall-e image? quoting "If a Dall-E watermark was present in the original image, you should erase it too!"
Hi, does anyone know how I could make an image 16:9 for a wallpaper?
i think that is composite in the dall-e bot
U click the image then click edit, then you’re able to erase it I’m pretty sure
How do I do that…
#dall-e-bot channel, input image and !composite
type !composite in #dall-e-bot
Thx
There’s no definitive list of art styles. I believe DALL E was trained on a dataset of billions of images. People have started to test and see if certain artists and art styles are recognized however, and our very own @wheat zodiac had created that wonderful site above where he catalogues them.
How do I specify which image to do “composite” on
I didn't see a save option.
By attaching the image! : )
you can attach an image by pressing the small plus symbol shown or by pasting it to your message bar.
Apart from the resources suggested above, I would recommend checking out the Dalle Prompt book by @spring wyvern over at 👉 https://dallery.gallery/the-dalle-2-prompt-book/
Thx
Anytime!
I still dont understand how do we erase the watermark for padding
The simplest method would be to try and crop the edge out?
You can do it right in the Edit tool!
oh, so we download the image - pad it and then erase the watermark
That would work just fine
cause straight-up removing it doesent work
There’s many ways to do it. That’s probably the simplest.
For image uploads, is there a way to simply apply a style to the image? The edit option requires that I erase a part of the image and the variations on input doesn’t seem to allow me to specify how I want them to vary
Got it! Thanks 😊
it worked, thanks
My account was just deactivated making character art of Star Wars characters - seemed innocuous but maybe violence mistakingly implied? Is there anyway to get my account restored?
Contact support@openai.com/
Thanks Avin - will do!
Anytime!
How do I get the Android app? My university work email maintains a separate partition which limits functionality. When I installed from the invite email, it went to the work partition and I want to reinstall in the personal partition.
There is currently no Android app for DE2, what exactly did you install from the invite? 
It must be dall-E 1
When I use !composite and follow the steps I end up with a weird image made up of 2 images very similar to the original
Is there a good way to expand the edges of this image and make it 16:9 for a wallpaper?
Once I verified, I had a prompt to "get the app"
just go to labs.openai.com if you are trying to get to generate images
it's just a web page
I think they are referring to Discord...
Dalle 1 was never released to the public.
Please refrain from sharing any personal or sensitive information in the server — whether that be yours or others’ (including doxxing). If you need to share this type of information, please only do so through Direct Messages.
Wrong prompt :/
I don't want people sharing an email here
so please don't ask people to do so. thank you
hey friends, trying to create art with specific phrases but cant quite get the order of words to show up correctly. any ideas how to accomplish that?
What is the policy of using a dall-e generated image as a profile picture? Would an attribution be needed for such cases and if so, how? Thanks!
I did that. How long should it take to return a result?
did you include a space between pad and the percentage?
it should be instantaneous
I tried both ways
Let me take a look at it
It does't look like you've posted anything in #dall-e-bot
I deleted them since they didn’t result in anything. I’ll try again and see if it works this time.
It's !pad not !ipad :p
You need to include a space between !pad and 60
D’oh! Thanks
I didn't even see that -.-
Yeah, I’m a dork, I just kept seeing !ipad
Apparently it doesnt need a space, my bad
Where can i see what the daily theme is pls?
#daily-theme and use #roles-n-badges to get a role which will give you a ping for each new theme.
Hey, can someone direct me to a guide for outpainting please? I padded an image and generated the padded image with the same search term I used for the original image and it just came back with a white box around it and the watermark removed. Figured I’ve gone wrong somewhere in the process 😅
how do i buy credits with paypal?
How often are new artists’ styles added? Do they update the database pretty frequently?
Hello!
Is there any way to get earlier generated character in a different pose?
For example, I've generated a character by prompt "pink creature", and now I'd like to generate "pink jumping creature" for one of the "pink creatures"
The only payment provider right now is Stripe.
Disclaimer: I am not a volunteer or an openAi employee, just helping out.
so im just out of luck? :(
that sucks
i was gonna buy some but nevermind i guess
not gonna make a new account on a completely different website just for this
one that ill never use again for anything else anyways
You dont have to, just enter your card details and you'll be good to go... It's just a payment partner.
Disclaimer: I am not a volunteer or an openAi employee, just helping out.
i dont wanna use my card, i wanna use my paypal balance
Use paypal to buy a prepaid visa gift card and use that to pay with stripe?
What is the difference between dalle and dalle 2 waitlist?
There only is a waitlist for Dall·E 2. Dall·E 1 was never released to the
public and therefore also never had a wailist 
Are Dalle-2 Images allowed to be used as discord pfps?
sounds.. convoluted.. i think ill just wait until they let you pay with paypal directly
Hello. I'm hoping to use a generated Dall-E photo as for an Uber Eats company. Is this allowed?
Yes, they are!
If the terms were to ever change, I assume anything created and used commercially before is still theirs to use?
Unfortunately, that sounds like a legal question, which I don't have an answer for.
Fair enough Ty tho :)
Hi, i was just approved and verified, but I'm not sure if i'm using DALLE or DALLE2?
hey is it possible to invite others atm or not?
not at this time
You are using dalle2
Thank you so much! You'll be seeing us on uber eats
Yay! I love the variety of images we can make
hey
why is it that when i inpaint it sometimes creates words instead of a motive?
i wanted to inpaint 'leafs' at the bottom but it made this:
Thanks mate!
Can you show me your prompt?
i dont remember but it must be something like leafs or stem or something
also does it matter that i changed the background of the inpaint image to black? is that making trouble maybe?
Hello, guys and gals.
My problem:
Say, I create with DALL-E an image with a particular painterly style. And it does so great. Now, whenever I want to use exactly the same style depicting a different scene this time, even if I use the same prompt describing the art style that I used before or I upload the AI-generated image (edit etc), the results look very different. As the goal is to create a sequence of thematic illustrations (meaning homogenous in style and theme)
Any ideas if there is any DALL-E 2 solution to that? Cheers.
Hi, this might be a pretty simple question, but as I navigate DALL-E I had a few prompts that were refused (with a warning), and to make sure to understand these violations and not make the same mistakes, could you tell me if I understand these right?
- I was making d&d art and had a prompt using blood that was refused, does blood fall under Shocking bodily fluids?
- I wanted to make "Shrek at the Yalta conference", refused, does it fall under deception ? And if so, would it be the same for every historical event?
Sorry if this is an easy or dumb question, but I really want to play by the rules and I wouldn't want to lose this tool just by messing around.
Thanks
I couldn't know the answer to the first question without knowing the full prompt. The second one was rejected because any cities in countries currently involved in active wars are banned.
It will block e.g. any city in russia or ukraine right now
The prompt was something of the sort " Angry white humanoid lion-man standing upright holding a sword with blood on it, digital art"
I removed the mention of blood and it got through
Thanks for the help!
Ah yeah that seems to be the case with that one. I think blood may be possible in contexts that don't involve weapons/anger but I'm not 100% sure on that
Nope I take it back, I can't even do "a river of blood on mars"
Interesting! Thank you for your answer, and for trying too
i took a jolly roger, uploaded it and removed the edges instucting DALLE to put "skeletal fists" in place. for two rolls it just gave me the same exact image i uploaded across all four pictures so i tried just plain old "fists" and it just blacked out the erased area entirely across all four pictures
am i doing something wrong?
how do i expand on existing images, like an album cover?
Can you show how much of the image you erased and what the exact prompt was?
The easiest way to do this is by using #dall-e-bot to "!pad" your image by a percentage. You attach your image to your message and type "!pad" followed by a percentage to pad the image. So lets say you wanted to add a 50% padding, you would type "!pad 50" with your image attached. Then you upload that new padded image to DALL E, and run it with a prompt of what you want to see in the new area.
okay, but if i wanted to just expand the image what would be something to type as a prompt
after the padding
How do I download images I create? When I hit the button nothing happens?
it will expand the image based on the existing image and your prompt. you can type anything.
(well, as long as it is part of the guidelines. 😄 )
Hi, I'm sorry to hear you are having trouble. You're using the download arrow button?
Yes I'm hitting the button on discord and the dall e web page and it still won't download
I don't think there is integration within discord to download a D-E 2 image from the website. Can you show me screenshots of what you're doing in discord versus the D-E 2 webpage?
The real problem is I can't even download the image from the website and for some reason I can't download dalle imagines from discord either. I can download other images from discord and other images from the internet so it seems to just be dalle
there should be no prohibitions for this discord server in particular regarding downloading images.
I'm not even sure if that type of limitation is possible.
Yeah it's really strange, like I said I have midjourney on discord and the images download just fine in that discord server,
Try a different browser
Good idea I'll try
That worked, duckduckgo must've not liked dalle very much.
wdym use that to pay with stripe?
how do i pay with stripe without having to make a whole new stripe account?
i heard someone say that i could get a temporary debit card or something like that too?
i dont wanna have to buy anything physical
im at home and its nighttime
i just wanna buy this online real quick
This is kind of outside the scope of this discord, but you should be able to get a debit card through paypal themselves, a bank, or a prepaid visa card (eg. privacy.com)
i went on privacy.com but it told me i couldnt get one cause im not american 
Does doing variations use up a credit?
Yes
Hi! How can we help?
My safari doesnt load the pictures, but now i switched to firefox and it works
no preview
Older versions of safari don't support webp, which is the image format that they use
You need to update your Safari
Or use a different browser
thanks
okay ive spent like 15 mins on this and i still havent managed to find ANYWHERE where i can buy a giftcard that isnt connected to any bank account, works outside the us, and lets you pay for it with paypal
:///
guess ill just never be able to buy new credits then
awesome
I strongly suggest that you not try to find an alternative payment method if you can't use those trusted methods, this is exactly the kind of situation that scammers try to take advantage of
just use your credit card
Not everyone has a credit card, or a debit card.
it's rare to not have any kind of card
Its also not helpful to tell someone looking for alternate payment methods "just use your credit card".
fair point
I'm just going off that stripe is handling the payment and OpenAI is trustworthy enough anyway, looking for alternate payment methods is a bit of fool's errand in this day and age 😛
this basically
Does anyone know what the name of this art style this is? It was an black sheep in the results but I liked it a lot. Emoji’s for attention: 🫠🫘🤡😭💥🏃🏻♂️🔊
Does anyone know what the name of this art style this is? It was an black sheep in the results but I liked it a lot. Emoji’s for attention: 🫠🫘🤡😭💥🏃🏻♂️🔊
you don't need the emojis, we can hear you! 😅
not sure if this counts as a style, but terms like "silhouette" (for the subject) and "limited palette digital art" could get you somewhere close. e.g.
silhouetted palm trees and sportscar on beach promenade, limited blue and yellow palette, digital art
So in illustration processes, this would normally be referred to as doing flats. Try this
So uhh in the discord rules there is stated that NSFW is not allowed, does that mean that you can’t share it or that it is not allowed to make it generate something nsfw to begin with?
Please read the content policy before you start generating with DALL·E, lest you lose you beta access.
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/
How do i make the images i get back more photo realistic instead of looking like paintings?
Typing realistic is the most basic solution but for more in depth information on styles check out: https://dallery.gallery/the-dalle-2-prompt-book/
It usually depends on how realistic your prompt is
also this ^
ok thank you i will try that and thank you for the guide.
Can you provide the prompt you're using?
One question i had was after you type the prompt and it brings back 4 variations. Saving, saves to my collection ( is that collection private? )
#❌┃prompt-help folks have also been incredibly helpful in getting precision.
Yes and yes
yes, that collection is private, pressing publish will make them public.
What is the purpose of publishing? Is it to share with the community? And is there a community board where you can see other works?
yes here is a prompt i used "photo realistic man on the run in a city chased by police car and two helicopters flying above with spotlights down on him in the style of James Patterson thriller fiction". It came back with this hilarious work. I want it to look like a real person in a real city. Not like a cartoon slapped together.
Yes. Not sure on a message board but there are channels here for sharing!
When you publish you will be given a link which can be used to share the generation (do keep in mind that your pfp and name will also be shared alongside.)
For some reason it just keeps giving me illustrated comic book style even though i tell it photo realistic.
see my image above
The best way to go about prompting is to imagine the image already exists. Don't think of it as describing an image to a computer to draw.
So what you are looking for would just be a photograph, "photo realistic" would be a term someone would use to describe a drawing of something realistic
But never a real name huh, not even a real fake thing like pepe
'photorealistic' is an artwork term; 'a photo of' may be more your style
You can also use photography jargon!
I attempted a stock market drawing prediction..didn't work well
ok thanks will try
got it.
camera settings are a huge benefit for getting better photos as well
Is there a way to feed dall e ( show them ) a real photo and have it create something like that? And if yes, do i just put the image in? Like in MJ you type /imagine prompt http://yourimage.com/image.jpg prompts here
Is it the same here?
No, but yes? There is a separate Interrogator which can indeed look at an image and give you phrases and suggestions, but it's not specifically Dall-e and there's no guarantee of overlapping training data.
It is its own separate thing
you can upload a photo and request variations of it, but you cant request any additional detail, it will just try to make images that look similar to the one you upload
that would work. So i just type in the full url?
Nope, it will be under the "upload" section
ah okay
Okay so i fed Dall-e an image ( they restricted me to a small square within that image which is kind of sad because i really wanted the full image) however, it created me variations that weren't bad. BUT the big problem is those images are just squares and part of my image is cut off from the original one i fed it. So original image dimension were like 1500 wide x 2000 high in width. I then had to select a square from that which meant it cut off a large portion of the cover.. Anyway it created 4 variations but i cant use those as the are too small
You can see the one on LEFT is original
see howmuch of the image is cropped
Yes so you will need to learn about outpainting next, that is a trick you can use to extend an image
Now if i was wanting to use one of the 3 on the right of that ( variations) as a cover i coudln't as they are too small
So here is a trick you can use to extend an image
But you will do yours vertically
okay i will view this
damn im glad I was sat on this thread
looks as though this is only working with the image itself. If i have cropped out the steps, i cant put the steps back in
so what sizing is dalle working with is it 1024 x 1024?
And why is it restricted to that? Why cant it just work from a larger image
Yep, there's definitely some experimenting you will have to do to get the regions you want and areas to extend, but yes it is 1024x1024
The model was trained on images of that size, so that is its current limitation
I uploaded a book cover that was like 1200 x 2000 but to use an uploaded image it wants you to CROP it to a square. So immediately I know its only going to produce a cropped variation of that image. How an earth can i work with a cropped variation. I.e Original image had steps at bottom, man, then washington DC at to part. The image cropped was just man and part of the building no steps. So variation ended up being basically duplicate of that with no steps and part of building at top missing. Surely Dall e cant expect anyone to work within those restrictions
People who would want to use this for book covers which are defined size and aren't square wont be able to
Most covers are like 2000 wide pixels x 3148 pixels high.
So i create a great square in dalle that is 1024 x 1024 now i have 2 problems - Tons of it is cropped off, and i have to stretch it so i end up with something i cant use. Urgh....
I love that it can create variations but its only as useful as the sizing it can give me
otherwise these 1024 x 1024 are only useful for instagram photos which is not the purpose of me using this
It's meant to be an inspiration tool and art tool, it wont work perfectly for every single use case :)
One thing you could try is shrinking down the full one so that it has transparent edges like this:
So create a 1024 x 1024 and shrinking the image within that
Is the largest size 1024 x 1024 pixels on dalle?
and then generate variations on that
crop it to your needs, and then upscale it using an alternative tool
ah i guess i will have to play around with it.
By the way why does it put a watermark on it if you are paying ?
and how do you remove that?
or is that because im on free still 50 credits
You're free to remove the watermark in photoshop if you like
but why are they putting a watermark on an image if you are paying?
I suspect that they will have that option in future versions of the app.
Its on there currently to show that the iamge was generated by an AI and not by a person
But they've stated that you are free to remvoe it
oh i see. weird
hey Toror are you able to remove that image from ( delete ) from dalle on your side
Can do
I want my coworker to be able to access our dalle2 account from his email address without using my email address. Can anyone help?
By the way did you do that image transparent edges using the tiktok suggestion?
I just did it manually by putting the image so that it fully fit within a 1024x1024 image
and then centered it
and then dalle just fills in the rest
very cool
One thing i did notice was the variations image quality was much less than the original. Why is that and is there a way to improve image quality of the variations or at least get them to be the same as the original fed to it?
Sadly Ive noticed this as well, they are always lower quality in my testing too
One user per account in this program- please contact support@openai.com to inquire about corporate accounts.
Hi,
how do you work around when editing a previous generation and the content policy trigger based on the original image (which is from Dall-E2 by the way)
note : the content policy trigger because of the image, not the words, just in case you wonder)
note 2 : I couldn't see anything that could justify that
This is still a current bug being worked on, but feel free to report your specific question in #bugs-and-issues
Hi. Can dalle produced images be used for commercial purpose ?
how should you format your prompts to get high detail? I just got access today and my first idea was to just create a cohesive sentence but I dont think Dall-E responded as well as it could have
Dall-E generate 1024x1024 images no matter what, you can add 4k or 8k or even the key word high details to generate a more sharp image. Also Dall-E respond well to professional photography term (it help generate photorealistic results)
you can start by reading this :
https://dallery.gallery/the-dalle-2-prompt-book/
thank you!
Ok I got a mail of the new pricing in 1 sept. . Can someone explain it please. Now a token is a word in your prompt ? Do spaces count? Sorry for my English
the mail is for GTP 3 not Dall-E2
GPT 3 is another openAI project that generate text, people have been using GPT 3 to generate better prompt for Dall-E2 actually and get better results
is anyone running into issues purchasing more credits? I've tried multiple payment methods but none seem to work
Ok thank you so much
Hey peeps, good afternoon. I1m getting flags whenever I enter "cumulus nimbus" cloud types lol.
Had one flag also with an image from another subject (man sitting reading a newspaper) that was originally created with dall-e 2 itself. Anyways, found the flags pretty arbitrary.
So I ask, is there a more comprehensive guide about words, terms, and input images that are not allowed, or how the flagging actually works?
Anybody have any idea how to get back to view your history of generated images if you don't already have the link? It takes you there when you generate something, but once you're out of credits, the generation button won't take you to that page anymore. I'm hoping that I'm just missing something obvious in the UI.
clic here :
That's it! Thank you, thank you, thank you. I was just missing something obvious in the UI 🙂
Just got an email about credits costing less, but I didn’t see DALL-E2 listed among those getting reduced prices? Does this price reduction affect us at all?
The email you received was in regards to GPT-3 pricing updates, not DALLE :)
These are two different products
Is buying tokens disabled for everyone, or just a problem on my end
Found my question answered above. Disregard
Is there any API to generate these images? and If not, is it against any rule for me to make my own that uses my account?
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.
Is there a way to improve quality of variations from an original image or are you stuck with whatever quality dall-e gives you?
Random (dumb) question: What happens if I close the Dall-e page by mistake while generating? Does the AI stop and gives me a worse gen?
Nope, it will either refund the token used if it didnt complete, or the generations will be available from your history
Not that I know of.
Do you know if it is planned to be able to generate different image formats.
(Having to use editing to expand the format is quite annoying, not to mention costly)
Then what is the purpose of Dall-e if its going to give you super low resolution? or worse quality than photos you feed it? Is it just for fun?
Sorry, I'm not sure I understand your question.
Is there API access for DALLE2? eg can I generate images via POST requests - if not any update on when?
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.
We use Dalle and pay them to create images. Yes? Well most people want to use those images for either PRINT or some other form. But they can't because 1. the image is limited to 1024 x 1024. 2. The variations it creates are super low quality than the original image fed to it 3. It crops out parts of an image you feed it. So what is the purpose of using Dall-e?
not sure what response you expect, maybe don't pay for Dall-E and pay a human artist instead
That's a really difficult question to answer, because there's a million and one reasons to use it
One solution. yes.
If you'd like you can post in #suggestions and let them know you feel the need for higher resolutions
pretty sure this suggestion has been made 1000 times already with the suggestion to keep style from one image to another
Yep, it has. But people should still feel their voice is heard, and there's no issue with putting in #suggestions again
I mean anyway that would be the logic developpement for the future
they are victim of their success, people really want more that what they currently have mean people are impress and want the other features to come quickly.
Possibly clueless question: I'm part of an organization on the API and I thought I read that other members of my organization automatically have access to DALL-E once I got my invite. However, one member just tried to log on and is unable to access DALL-E without an invite. Will my org members have to get their own invites?
It's true that the poor quality of the variations is quite astonishing, and takes away all the interest of these variations
yes exactly. At least in MJ you can improve the images. Though MJ has some major flaws too.
@olive crown if you'd like to discuss other AIs please use #ai-discussions.
I just read it, thanks, I kinda panicked because I closed the window by mistake
No problem ❤️
is dalle learning as people use prompts? or is it updated by devs every so often to include more things?
The model DALL E uses to generate images is static, so it doesnt learn as people use prompts. I am unsure if they plan on updating that model in the future or not.
why DALL-E won't accept this?
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/
How many violations can you get before suspension?
I was not suspended. but i wanted to draw this and wrote this on my first try
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.
If you wish to discuss images generated by other AI, please use #ai-discussions
sorry
Can anyone buy credits right now?
Having the same issue, I guess we can just wait
Is there a way to make it generate in a 16:9 aspect ratio?
No, currently DALL E can only generate in 1024x1024
you can use patch-e.com to make a mosaic with multiple generations
See the Terms of Service
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/
If I generate an image, then pad it to be smaller, I can keep the exact some prompt for it to add to the border?
The center will be the same and then just keep growing?
Yup. It helps to vary the prompt as you go farther in, otherwise the edges slowly disappear and it makes it harder for DALL E to continue generating additional content
Is the picture zoombable then?
Can I upload a picture and give it additional prompts?
Could you explain a little bit more? Do you want to add more to an existing image?
If so, you can use #dall-e-bot to "!pad" your image, and then upload it to DALL-E
No, it's just an expanded image
For example
Like if I upload this image, and I put in a prompt something like at sunset, birds flying around ect, is that possible?
Well, unfortunately you can only edit a square portion of that image. But yes, you could erase the sky or parts of the sky and add that prompt
Okay great
Jpc you’re the goat answering all these. Thank you man
Aw, thank you ❤️
Hey guys, does anybody knows how long the credit buying option will be disabled for?
According to #announcements, it was restored about 4.5 hours ago
Thanks for answering Jpc, I'll check the announcements channel more often, but unfortunately I still have the credits buy option momentarily disabled
Ok nevermind I refreshed and it worked 
Hey guys and girls, I want to change the password of my Dall-E account but can't seem to find such an option anywhere. I know that things are new and rough, but this is a really important thing for many people. Google sent me a notification that my passowrds have been compromised and I really need to change it.
Maybe I could do this through the "forgot password" process?
if you've followed these steps with no luck: https://help.openai.com/en/articles/4936828-how-can-i-reset-my-account-password
Yep, I came to that solution and it worked 🙏 But still, I think it's important to have a password change option on the account management page
If we use dall-e for commercial use does the dall-e logo on the Connor needed?
no, but ToS still indicate you need to acknowledge the role of AI in producing the art. so the watermark is the simplest way to do that
How do I enable beta?
do yall think "X character as judge in Nuremberg trial, 1945" would get flagged
Is there a way to access older prompts in my history? I can see my recent ones but not the older ones
not necessarily but it's in bad taste, and it's been done (and warnings issued over) here in the past. the nuremberg trials represent a slim light of justice for a horrible, race-based atrocity and it's really not a good subject to make memes about
try your browser history!
That wouldn't help much as I typically use the mobile app, so the prompt I'm trying to find wouldn't be in my browser history
i thought the same - but even though it's a separate "app" they still were logged in my chrome mobile browser history
Not really a question about Dall-e, but since we were talking about it, is there a discord for GPT-3?
Does inputs in other languages works?
Some people have said yes. Try it!
any plans to make the price points...not silly and ridiculous?
Is there a way to upload and image, that only has the line art and flat colours, and generate renders of it?
Hi, is there a way to buy credits with paypal?
Currently the only payment option is through Stripe
Ok, thanks
How can I setup OpenAI through python. The instructions are confusing
Hi there! I'm not sure what you're asking, could you elaborate please?
When I do localhost:5000 it’s just a blank screen
I need to know what you're trying to do. DALL E is only accessible through their website. The API is not currently 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.
I’m trying to use the da Vinci text complete. I actually got it working but how do I do something other than pet name generator?
in the playground, you have presets you can select
An API for accessing new AI models developed by OpenAI
is it true that paid credit expire after some time ?
• 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.
Yes, they expire after 12 months.
it will show the expiration date
it says on the checkout page
Paid credits will expire one year from today.
Terms
you are right, I missed it when I paid
that would definitely change my purchase in the future
cause there is no valid reason for this
you can put a note in suggestions, would be a good idea
well I actually discover this when reading the suggestion channel lol
Is DALL-E 2 open source? If not, are there any implementations that are open source?
And what hardware is needed to run training/inference?
DALL E 2 is not open source. To my knowledge there are no current plans to open source it
Ok, thanks for answering
I found that there is an open source implementation called DALL-E Mini. Is that developed by OpenAI, and does it work well?
what should i do if i try to reset my password but don't get an email?
reach out to support@openai.com
DALLE mini was a crowd sourced version of text-to-image technology and is not developed by or associated with OpenAI. They eventually renamed that model/site to "craiyon"
I have a question. I have been creating some ai art and got pinged twice accidentally for violating the rules. How many times will this happen until I'm banned? I'm not trying to break any rules
As long as you are not intentionally trying to bypass the filter, you are safe :) I have gotten 50+ warnings myself from innocent prompts, just be wary of the words you use that might be considered bad
Thanks! I've been posting some errors that weren't even remotely close to being supposed to be filtered.
once I switch the word blood with red liquid, is it something I should avoid doing ?
I have a question about a painting I generated, want to know if the prompt/idea is against content policy
What's the prompt?
Yes, trying to get prohibited words by using things like red liquid can still jeopardise your account if it comes under review.
Ok I'll be careful then
How can i use the padding thing to make an image larger?
Does anyone know of a way to implement specific color theories into the generations?
that's the closest I have gotten
It works like that... figured it out lol
so ive seen people generate like longer images, where two are combined to make a longer one and is seemless does anyone know how to do that?
Sorry for not responding until now- Thank you! That helped a lot!
like if i wanted to add more to one side of the image, like for a wallpaper or a banner , how would i do that?
Does anyone know why I cannot save my created images I made on DALL-E to my phone? I was able to earlier today but now I get a message saying, “Couldn’t save this image”.
Try another browser
How do I expand an image, i think its called inpainting?
Outpainting*
Do I have to like, pad and then select?
Yep! Go into #dall-e-bot and use the !pad command to shrink it by a percentage and the upload it back into DALL E with a prompt!
oh thx
So can I like put a credit in the beginning of the my video that I'm making?
i tried generating an image of a mob boss with custom cuff pins and i got a warning for content policy? is it a bug or can i not generate mob bosses :o
What was your exact prompt
a mafia boss wearing a tailored black suit with cuff pins with a laser engraving of a cup of steaming coffee and a coffee colored tie on his neck
i replaced mafia with mob and it seemed to go through but i just want to make sure im not gonna get in trouble about it
i changed it just now to see if it worked yea
I take images produced by Dall-E and put in videos?
Yes, you can use DALL E images in your videos
thanks
but one question, do i need to mention that they were made by Dall-e?
Yes, according to the Terms of Service, you must disclose that they were made by DALL E. The easiest way to do that is to leave the watermark on the image.
why is the screen always white? the sight is fully loaded
May just be a bug on that browser, have you tried another one?
not yet
ill try
oh it worked
thanks lots
Is Dall.E available on mobile or something?
Yep, you can access it on mobile as well.
by visiting https://labs.openai.com/ using mobile browser
I am trying to generate jewelry designs by uploading images I have. The designs generated are not so good though - would the AI improve if I keep generating them?
Hi can we use our works without watermark for Commerical Use ?
yes, but technically the ToS still stipulates that you disclose the role of AI in creating the images. the watermark is the simplest way to do that.
no - the model is static. but if you have data sources that you feel could improve the output you want, you can tell openAI about it
• While model updates aren't frequent, we can definitely consider adding licensable data for it in the next training round.
• Please tell us more about it through this form.
Thanks for Info !
So if i remove watermark i need to say Somethig like created with Dall-E right
Something like that
Okey , thank you again !
One more question Can Dall-e only do 1024x1024 works ? Or can we choose size ?
outputs are only at one size. if you want a wider aspect ratio you can composite images
How Can i ? Btw Sorry for a lot of question
hiii I have seen posted on the discird a link to a website wich is processing image like this and i lost it , by any chance someone remember?
Hi does the Dall E dataset not have Genshin Impact? I tried and got very random things
PAID credits DO roll over to the next month right?
• 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, free credits expire after month after they were granted – but you’ll get 15 new free credits.
• Paid credits will expire after 12 months from purchase.
When I was doing the thing on my phone it gave me an option to add it to my home screen
Like accepting the invite
Idk if that's an android only thing tho
You can add it to the home screen on iOS through Safari. But it only brings up the webpage.
Hi, is there any other way of buying credits? I am trying to buy credits but i do not have a credit card. And in my country especially with my bank its an huge hassle. i tried 3rd party websites. but that also proved a hassle.
I do have Paypal
Currently, the only method of purchasing credits is with a credit or debit card.
I only see credit, not debit. or perhaps i am confused about the definition but i thought debit is just the standard bank related card. that we pay our things with in stores here.
Hi, I have a friend who recently got access to Dall-E 2 and they quickly encountered this message. We've both looked through the Content Policy and neither of us can figure out what the image seemingly violated.
This was the image they allegedly submitted.
Looks like a gun
yeah that's what we were wondering, but the content policy doesn't mention anything on guns
It is in there somewhere
paint the tip red.
Like you can't do "shot" in a prompt
or perhaps its the trousers looking a bit like... heh....
oh yeah
that also
I tried Ctrl+F and typing "gun" and get no results
I mean it's related to violence
to it is in that grey area
if only it was 2010 , all this would not be a problem 😄
but i do not blame google. they will get a huge wave of drama if they did not have a policy up in current times.
yeah I'm not upset about it or anything, just wanted to try and figure out what the issue might have been so we can avoid it in the future
i think its the trousers lol
why not try to color it or change things
in fact let me do it for ya
eh, it's not my prompt but I wouldn't want to risk testing violations with it
dont worry you can request unbann
but if you do not want to risk it. then simply dont.
apparently they still get the same issue
i guess you have to ask the person who made the filtering code.
#bugs-and-issues might get more attention
so igor tested the image for you?
perhaps its still the gun thing 🤷♂️
i would try it for you but i do not have credits right now. still in the process of getting a credit card.
this maybe?
but i do not want to be soliciting a ban on your account.
yeah I think I'll stop for now.
perhaps let somebody try who has not had a warning yet.
🤷♂️
it would probably be an interesting subject for one of the devs.
to refine their filters
I've cross linked it in #bugs-and-issues if that helps
probably 🙂
im in no way complaining about pricing because 3.2 cents per image is a fair pricing, but i heard OpenAI API is reducing dramatically in price. will DALL-E credits also reduce? mainly curious as again, the pricing is fair
What is the copyright policy on DALL-E 2 generated images in general?
It states on its website iirc.
You can use them for commercial purposes but you can't mislead people into thinking you actually made/created the art yourself
the generations you generate I mean
This a link to the Terms of Service https://labs.openai.com/policies/terms
So, can I use a DALL-E 2 generated image as profile picture without the watermark if I say that it's generated by DALL-E 2? For example, if I write in the biography "Profile pic was generated with DALL-E." (I'd write it in German because the account is German-language)? Same goes for any other platform where you can describe yourself (including Discords "About Me")
Long story short: Can I use a DALL-E 2 image without watermark if I credit OpenAI?
I don't think you need to specifically specify that it was made by Dall-E
You just can't purposefully trick them into thinking it wasn't generated.
Yes.
More FAQs answered here: https://help.openai.com/en/articles/6468065-dall-e-content-policy-faq
I always say, who my profile pictures made if it isn't obvious (that's why I draw most of my profile pictures for myself so I have to credit no one)
So it is okay if I use a non-watermarked version as profile picture if I credit OpenAI?
Think of it kinda like being a Lord in the UK.
I am technically a Lord as I own a plot of land, that grants me the titleship of lord and I am allowed to put it on information like a Drivers Licence if I ever got one, and am allowed to tell people I am a Lord.
However, I'm not legally allowed to purposefully make people think I am associated with the House of Lords in London as that is a different thing altogether.
So that's a yes, I guess
Yes you may.
Thanks 🙏
The image looks like it's holding a gun, thus the content warning.
Thank you. The rules were a bit unclear if guns violated it or not, since you can have a gun in an image without it being used explicitly for violence.
I think I read here that the history log is limited to 50 images. is it still true?
How to make character perfect because in some images eyes are not generated well
dall-e is still messing up on eyes.
Yes
I'd recommend #❌┃tips-n-tricks message
Edit out the eyes and keep doing it until it looks good
I am getting broken image links on my thumbnail and expension renderings, I can only see the images if I download them first, Safari Version 15.6.1 (15613.3.9.1.16, 15613) on an up to date but older macbook air running 10.15.7 .. the developer menu says the image error is: WebSocket connection to 'wss://nexus-websocket-a.intercom.io/pubsub/5-EP_UEBwWurUAvmmTbnDbODpsKDKIB_tbyTE2fbQ18lqbJ2PhibJgtjxwmAvcZlCMeUtOG-IIXpeP7aBWfIRq_M_QTzsUE5NPBLUI?X-Nexus-New-Client=true&X-Nexus-Version=0.9.0&user_role=undefined' failed: WebSocket is closed before the connection is established. Is there some setting I need to change as I haven't had any other websites have this problem?
While it's true that the history is currently limited to 50 images;
you can always check your browser history for urls with either labs.openai.com/e/ or labs.openai.com/
You'll find your old images that you haven't saved to your collection via the links! 
fyi this was on my first attempt at using dall-e 2 so it has never worked.
Safari has previously caused issues for some Dall•E users.
Switching to a different browser should resolve the issue 
This should not be necessary, though.
Are there any other errors in the console? The particular one you shared should be unrelated to image rendering. I also just tested Safari 15.6.1 and it's working from me, so it might be a separate networking issue.
Hey guys I'm new to this! I have a copyright question: Am I allowed to use images I created with DALL-E commercially?
Hey everyone, sorry for bothering. Can I ask what are the things that reduces the credits? I know that generating an image does reduce it but are there anything else I should be aware of?
Yes.
Any time you generate images, you will lose 1 credit.
Variations, generating new images, or generating images using the edit feature.
So those will cost me 1 credit more? Thank you.
Just 1 credit.
I am sorry, to remove any misunderstanding I will ask one more time, really sorry. When I make an image, it costs me 1 credit. Then, any alterations or editings will not cost anything right?
No. Every time new images are created, including alteration or edits, it will cost you 1 credit
Thank you so much.
Hello i wanted to ask about one thing. i wanted to generate an image with text* hyper-realistic vergil from devil may cry 5 walking badass during the thunderstorm* on an image of Vergil holding sword I took from the site https://devilmaycry.fandom.com/wiki/Vergil?file=Vergil+DMC5.png which is not a fan art but a file from the game itself. I got a message that it may violate the policy, what should i do?
MAIN DmC MAINFileGalleryQuotes Vergil is the son of the demon Sparda and the human Eva, the elder twin brother of Dante, and the father of Nero. He is one of the main characters in the Devil May Cry series, introduced in the first game as the secondary antagonist, one of the main antagonists of Devil May Cry 3: Dante's Awakening and the final an...
Could you please post the exact prompt?
• Our filter aims to detect generated text that could be sensitive or unsafe. The filter will make mistakes and we have currently built it to err on the side of caution, thus, resulting in more false positives. We're working on improving our filters, so this should become less of an issue in the future.
• While we're not sharing the exact thresholds, a few warnings here and there won't lead to an account suspension.
i got a question about DALE E 2 can you generate a picture on higher resolution? like 1920x1080
Unfortunately, you can only generate images at 1024x1024 at the moment. You can, however, use a tool like patch-e.com to get creative
ohh thanks 🙂
What’s wrong with this?
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/
Can't generate real people
Shucks
Can it generate real people with them being in like a style that isn’t real life? I’ve done one before of Elvis Presley but he’s not exactly human in it
yeah uh why is it hard to create two anime characters
tried to make the fella have a short ponytail hair and the girl has a straight hair but it wont really wokr out lole
how do i get a specific character?
I guess, unless the person is real, like Justin Bieber, you can type out their names, Like Rick Sanchez or something
That's what I'm guessing
I tried but no luck. and no they are not real
Use #dall-e-bot to shrink it by a percentage using the !pad command. So for example, if you wanted to shrink your image by 50% you'd use "!pad 50". Then upload that image to DALL E, enter a prompt, and hit generate
how do i get good anime faces?
Hi there, I am a newbie and just started today. How can I change the resolution?
just with the word anime? that isnt working
I mean, I think the results I have speak for themselves
Or better, how can I change the image aspect ratio?
that is good but how?
Currently, DALL E can only generate images at 1024x1024
The prompt is included in the image
thank you very much. Hmm. okay
do i just describe the setting of the image for the prompt?
can I also enter a prompt directly inside discord and generate an image from here?
or do I have to go to the dalle website?
Yeah, or whatever you want to add. It takes into account the prompt and the image
DALL E isn't available through Discord.
Is dall e continuing to learn based off the images and prompts its making? Or is this version of the project released as like a static state of intelligence, that we are all exploring and creating data to put twords dall e 3?
No, DALL E was trained on a set of static images
Is it possible to purchase more credits using paypal?
Not at this time
Is there a recommended amount of image reduction when doing an outcrop process from a close up (head only) to full body? I figure the close up should be reduced to then use something like patch-e to expand, but curious on the reduction amount for the starter image since I don't want to lose too much quality.
really depends on your own idea of the final composition you want.
There are tools for upscaling images such as cupscale.
Will the generated images always be available to mee or do I have to save them?
Use the Save button to add anything you want to keep to your collection
Anything not saved will likely get deleted at some point in the future to free up resources
I have spent points because of the verification
technically it didn't have to happen right?
It says that that will happen so to do something you wanted to do with the prompt
I see, yeah it seems that it deletes your history beyond 50 generations,
You can get to older generations if they're still in your browser history, but there's no way to browse them directly from the site 😩
I see, thanks for the tip!
Make sure you save them though, I definitely wouldn't rely on them sticking around forever!
Verification is how we know you have a dall-e account. If any prompt that had the ignored number at the end was free, that would be easily exploited. You still get an image you wanted out of it. 🙂
Hi, just got accsess to this amazing AI. There are some pictures I would like to use for a print on demand service to sell them on clothing (Shirts, Hoodies etc.). I used the discord search but couldn't find a clear answer.
So is this scenario covered by the Terms of use image for "and otherwise commercialize them"?
There's nothing stopping you from selling things with generated images on them
correct
If you believe your account was deactivated in error, how long does it take for it to be appealed once you email support?
no idea. be patient, i believe they have a ticketing system for support inquiries so your case will be examined at some point
Is there a prompt to ask for a Rube Goldberg type of incredible machine?
Can I use the images created by DALL-E commercially?
Yes.
I recommend reading the policies & terms found in #old-server-rules under Rule 0.
What is it with the faces? I asked Dall-E to create a foto of a viking wedding - well it kind of is but what is it doing with the faces 😆 I gues it is on purpose but I saw good generated faces in other generates, so is it me?
it's not "on purpose." most small details are inaccurate in DALL-E, but our brains are hardwired to be very good at identifying facial details - it's social evolution. so our attention hones in on the faces and we struggle to "read" them when the details are wrong
you can get realistic faces; usually with individual subjects using inpainting and the term "portrait", or by filling the frame with them (and then outcropping if need be). but there's nothing deliberate about it, it's just an example of how the model still struggles with perfect details
This prompt gives me a content policy warning. Why? (No, it's not the word "high"):
powdery steep skiing slope between high trees into the valley, sun beams, blue sky, glistening snow flakes
hmm. it may be picking up "powdery" and "high" together if you know what i mean? part of the filter is contextual AI
removed word by word, and the trigger was "glistening" 🤔
very weird
Hey guys, how do we do the image as an Instagram story composition?
You can find tips on how to expand your image in #inpaint-outpaint & #❌┃tips-n-tricks
Sorry for not responding yesterday - I got really tired and only seen the message today. I will sent the prompt now
Hi! The error is not with your prompt, the error is with your uploaded image. It's seeing at as a realistic face so it will give you an error.
Is there anything I can do to upload it and not get the error? Or do I have to just find another image on which the face does not look so realistic?
You may try that, yes. But that may be circumventing the policy. 😅
Is the sharing of art that includes realistic human faces prohibited?
If it's generated by DALL-E, that's fine! Check out #photography
First timer here, why would my hummingbird be so blurry in all four generated images? https://labs.openai.com/s/7OvmheSk91M6Pjr2Qv2kvroQ
Are you wanting a result like a photo or a painting?
It kind of went more toward the photo side of things without me specifying anything, with the hummingbird being the blurriest part of it. Just strange.
It gets a little confused on what to focus on sometimes. You can try close up photo of at the start to help.
Try #prompt help for more assistance, and maybe check out #photography
Okay thanks, those are great.
the model is probably trained on a lot of blurry photos of hummingbirds because they're bloody fast 😅
Probably. I already got a decent photo after that having checked the photography thread after my question.
hey I know it's probably staring me in the face but how do you add things to your collection?
If you click an image, there is a save button top right
Not to be confused with the download icon
thank you very much
How does one use the inpainting feature? I don't see it.
I can help with this! You can start inpainting either by pressing "Edit" on an existing photo, or uploading a new photo and doing the same
Thanks @short salmon !
Is there a way for me to invite a friend? We've so far been heavily using MidJourney, and I'm excited to try out Dalle2.
Currently, the only way to join the DALL E 2 beta is by joining the waitlist
quick question, is it better to use punctuation and proper grammar or is simply writing out words to include in generation the best?
hey how do I know im using dall-e 2? I got an invite a while back signed up and whatever but when using the generator thing it does not say dall-e 2 anywhere.. unless im just being stupid? lol
Yep, you're using dalle2. dalle2 is the model, but goes under the branding of just "dalle" 🙂
The original dalle (1) was announced in 2021 but never released to the public
ahhhhh i see, thank you!
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/
A stupid question, I know, but
My credits will refill on August 29th, however, I planned buying 115 credits on August 29th
I retain the remaining paid-credits, even if the free credits refill, right?
• 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 after month after they were granted – but you’ll get 15 new free credits.
• Paid credits will expire after 12 months from purchase.
Yes, the paid credits remain
So I did understand it correctly
Thanks
what credits do
Credits are what you pay with for your prompts :)
1 prompt costs 1 credit 
and when my credits end?
i can do something else to get other credits or i have to pay
hey does dalle understand any color codes? or is there a way to get a consistent color on every prompt? When i write white background for example i always get different shades of white, sometimes warm white, sometimes a cooler white etc.
As of now, you have to pay for credits or
wait a month until your free credits replenish 
got y
Other question
I edited a generate image in Photoshop to remove some extremely ugly things (the eye for example was just a huge, blue blob)
Can I use the edited one to create variations of it with DALL-E 2?
I also removed ugly parts from the image (but not the watermark)
Yes and no to hex/color codes.
Some people were successful when they included them in their prompt, while others were not.
Perhaps you can get more help over at #❌┃prompt-help :)
Absolutely!
If you merely want to edit parts of the image you can always use the 'Edit' function of Dall•E.
Oh and, #edit-inpainting would be a great channel for you to explore! 
Can I upload a before-after image?
Left the generated one, right the edited one?
If you need help with editing/inpaint, please post
your image(s) and question in #edit-inpainting.
In case you want to showcase your work, go through the other
channels on this Discord and decide which one is ideal for your creation! :)
I accidentally misspelled my last name on my account. How can I fix that?
Oh no! Sorry about the trouble. Do you mind emailing support@openai.com? That team will be able to help with any issues related to your account
I'm not sure why this prompt, which is related to a TTRPG I am running, came up as violating community guidelines ~ can someone help me understand where the line I'm tripping over is?
Here's the prompt: "Shoujo-noir image of a nonbinary Mafia sword-duelist. On the lonely streets of a tragic, starving city. Blue hour, lit by neon."
I tried it first with "genderqueer" instead of "nonbinary" and that came back as violating guidelines. While annoyed (by my assumption) that my actual IRL gender violates guidelines, it does include a reclaimed slur so, sure, ya, fine, I get it. But replacing it with "nonbinary" came back as also violating guidelines, so now I'm confused?
Was it my use of "Mafia" or "starving"?
Please take a look at the content policy if you haven't already.
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/
The issue is most likely mafia. But you are also not supposed to generate images of suffering
Which a starving city would definitely qualify
Okay, thank you for answering!
Yep! Happy to help
Thank you!
Hi there 🙂 I hadn't tried genderqueer but dall-e isn't great at non-binary images sometimes. Drop that in #bugs-and-issues for that part.
also try dystopian for the city description
by the way, genderqueer works. 🙂
rad! It musta been Mafia then and I just jumped to (what I would say is a very reasonable and understandable) conclusion. I appreciate you letting me know I can use my gender as a term, thank you!
Are you allowed to generate images for other people?
like people suggesting prompts for you to use? yeah. as long as it abides to the content policy.
Yeah pretty much, but what about asking people for prompts?
From what I understand is, you can get suggestions from people, generate images, and share those images with them. However, if you were to offer to generate other peoples prompts in exchange for money, that would be against the Terms of Service
Ight thanks
Can we upload an input image to define art style, then describe the subject? Would be lovely to be able to define like a painting style by providing one as an example
does DALL-E train on copyrighted material
can copyright owners opt-out?
maybe Blizzard sent a Cease and Desist to Dall-E
This is a question for OpenAI. If you'd like to reach out, I suggest asking using support@openai.com
It's not
This is a little difficult to answer. There's a couple of ways to accomplish this. The easiest would be to use the title of the image in the prompt. For example: "A cat, in the style of "Starry Night" by Van Gogh.
Ok - well, I knew that, the hard part is when you don't really know how to describe the style (or it can't just be named like that)
The second way is a little bit fancier, and would take at least 2 tokens. You can upload half the image you're trying to recreate (please make sure you have permission or rights to do so). Then write the prompt you want, generate the image, and basically do the reverse.
It might be less than half. I would use 1/4 of the image
and see if that works
@short salmon Ah, so you can erase a part of the picture and tell dall-e what to draw in it?
Yes, it will match the art style of the portion of the image you include
well, if you think about it, dall-e knows what pikachu is, so he must have trained on copyrighted material
As I said, I am just a volunteer. I have no knowledge of what was in the training data.
Gm, so whats the deal with the watermark? How do you remove it?
Yeah me too
Dalle appears to be down at the moment
probably has hand cramps poor thing.
are your prompt history and corresponding images saved after you log out? Or only when you add them to your collection? (a.k.a do I need to frantically save everything to collection or only the best, since the rest can be retreived later)
Your generations are primarily saved under a unique labs.openai.com/e/ address created for each prompt you run. Saving these addresses lets you view your generations at any time (OpenAI are not deleting images from their servers at the moment). Logging out and logging back in will not affect these, as they are tied to your account.
If you want to save and index all of your generations, I'd recommend bookmarking all of your /e/ addresses. OpenAI might have plans to let you view your entire prompt history at some point, but we don't know anything for sure. Hope this helps 
(The history sidebar will also keep your last 50 prompts for quick access)
All clear, thanks! I'll have to do some bookmarking on my own then. Glad the history is kept. It's nice to sometimes go back to a previously explored topic to see how prompt alterations / edits help evolve the image, while only saving the final beautiful image to the gallery :)
Some kind of dedicated history tab would definitely be a nice addition.
is there a way to change aspect ratios?
Currently, DALL E 2 only outputs at 1024 x 1024. However, you can get a little creative to work around it. One super easy way is by using the website patch-e.com
beware though you will spend alot of credits 😛
#inpaint-outpaint message cost me like 40 i think
I did a bunch of these, they cost only like 3-4 credits
how 😮 did you upscale?
Gigapixel AI
I mean tbh I'm fine with the sizes before. I just wanted to see what upscaling would do
yeah its not bad, though i think you can get more detail in with building through patch e but ye thats expensive... 😄
this was my initial image
this is what it became, though a bit too chaotic :/
I think its straight up fire

I love the mixture of the muted grays and the colors in the sky
Is there a problem if I write something with celebrities, for example, Justin Biber doing a song with Beethoven?
Heya, just got access to DALLE, it doesn’t seem to do existing 2D characters well when trying to generate them, is this for copyright reasons?
Generally celebrities will generate a warning
Hi! My discord verification ended up using a credit when it said it wouldn’t
is there a limit to the uses of the bot, similar to mildjourney has 25?
Yeah, you get 50 credits and sign up and 15 free per month. Can also buy more credits if you choose
• 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 after month after they were granted – but you’ll get 15 new free credits.
• Paid credits will expire after 12 months from purchase.
@wild crag
Yeah that’s a good price
Probably
Im looking through the content policy but i cant see anything about it so idk if it breaks rules or not, just double checking.
if i wanted to make a logo my twitch channel, what are the rules about that? do i have to submit one of those applications or is it free to use as long as i creddit openAI?
double checking but is the public thing dall E 1 or 2?
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/
Yes.
Policy still says mentioning that it's made by an AI is still needed. No need to submit any applications 🙂
More info in #old-server-rules, under Rule 0, regarding the commercial uses.
so to clarify, using it for a logo for social media is fine as long as i creddit OpenAI?
Correct. But please, just to be sure, have a read at the Content Policy & the Terms of Service. All located in #old-server-rules Rule 0.
aight thx
hi guys,
how do you add DALL-E to your own server?
You can’t right now
so the only way to use it is online for now, right?
correct
thank you
Does the count reset at the 1st of every month, or every 30 days from when you got accepted?
• 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 after month after they were granted – but you’ll get 15 new free credits.
• Paid credits will expire after 12 months from purchase.
ok so a question (yo #community-help channel)
I make an AI generated image, it uses the credit
does it still use a credit if I edit the image?
with that edit button
yep
ah dang
you can use the edit function for small tweaks but also essentially to create entire new images based on a small amount of stimulus - computationally it's not necessarily any different, and that's what the credit's for
seems reasonable
Hey guys, does Dall-E wipe the recents at some point? Should I be saving all of the images locally to avoid losing them? Or it doesn't happen until you explicitly click the "new" button?
there's a limit of 50 sets i believe
that said: your browser history will have a record of the links, and the links tend to remain active indefinitely
Thanks!
Vivid painting of a male medieval knight wielding a double-edged axe and wearing heavy chainmail and a helmet in front of a castle tower in the middle of a day with a clear white sky
Wtf, how does this prompt violate content policy.. It cannot possibly be the axe unless DALL-E is "racist" towards certain category of weapons, since it has no issue creating pictures of people wielding swords, lances, maces and any other category of medieval weaponry
please don't diminish racism by equating it to your question about how the filter processes weapons. it's not the axe:
https://labs.openai.com/s/tGv2yshhMPnGPYQb2t17E9G9
part of the filter is contextual ai. maybe it puts a few elements here together and decides this is violent or military imagery - i'm not sure. whatever the case, when you get a policy violation it's usually for the best to change your subject for a while.
Sorry it came across that way, it was an improper use of the word, thus the quotes.. "discriminatory" was probably more correct. Thanks for the heads up, must be the combination of factors being judged too "violent" in some weird way.
So I have a game that I'm working on, I'm not part of a company nor do I own a company yet, simply a hobbyist/aspiring indie dev.
I want to use this amazing tool to generate graffiti that I can place around the various city's/levels.
My end goal is to sell the game when it's complete.
So my question is, would I be able to publish my finished game without breaching any copyright rules ?
Yes.
I also recommend you to read the policies & terms found in #old-server-rules under Rule 0. 
Thank you, I will do that just now 🙂 Looking forward to the future of this tool I'm happy to be apart of it.
Is it possible to contribute to Dall-E 2 on Github or somewhere else? Or maybe it's not open source? 😅 Would be lovely, since I'm studying computer engineering and would love to be active! 😊
Can someone tell me how this bot working?
It's not Open Source in the traditional computer sense. OpenAI is open in that they freely publish all their research - leading to open source versions of their tech (e.g. StableDiffusion).
which "bot"?
DALL-E bot
Ah, ok! Thank you for the reply!!
#dall-e-bot is just a tool designed for this server to help with DALL-E edits and profile pictures. It can give a transparent buffer to your images (!pad), help you stretch them into a wider aspect ratio (!composite) or put a watermark on your profile pic (!watermark)
So I can't use this bot like the DALL-E web to regenerate photos
Thanks
Does anyone know why this prompt is a content violation? I'm not sure what to remove
Hi, I want to ask about using images generated by DALL-E. I want to make a game in future, and I created an image of a phoenix ( https://labs.openai.com/s/ouSteRdA5y0yoAvbBY0OAMt5 ). Can I use it in game with crediting DALL-E? (My game can make me money.)
Yes.
Hi, i want to generate a nfc tag with a justice scales symbol, but i dont know why i cant put the scales in the tag. Prompt: "High quality photo of a white studio. Overhead view photo of a nfc black keychain, in the middle it has a balance scales logo in gold yellow. It is a product for an attorney. Strong focal point in the center. Product design, natural light."
Hello, I have a question: are the images I generate with Dall-E 2 automatically public and visible to everybody, or is there a way to have them private and visible only to me?
might want to post this in #❌┃prompt-help
they are private to you by default. you can share an image and choose to create a public link you can share
ok, thanks
thankss
Should I have 0 credit on DALL-E 2 if I want to get 15 credits after a month?
• 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 after month after they were granted – but you’ll get 15 new free credits.
• Paid credits will expire after 12 months from purchase.
No, you do not, you will get 15 free credits each month irrespective of whether you use the previous month's credits.
Hello, is there a way not to have the watermark on each image i generate?
Hello! Is it possible to generate images like 16:8 or of different size? Also, what is the promt to complete / modify an image uploaded?
Take a look at this 
#community-help message
thanks a lot!
You can either use an image editor to remove it or the inspect feature of your browser to download it without the watermark
You will need to convert it from webp format using an image converter if you do the latter though.
thanks, and is that allowed by Dall-E 2 terms?
Yep!
Cropping the image is also an option, in case you want to do it that way
Always 1 credit
thx
Hello, do I understand the terms correctly, when I say that every single image I generate with Dall-E 2 needs to be attributed to Dall-E 2, regardless where I will use the image?
Can Dall-E moderators see our galleries, if so, do they look at them?
We cannot 🙂
hi everybody! i waited for my dall-e invite for 2 months! i finally got it! i'm so psyched, but now, I want to use my complimentary tokens right so i wanna read and learn how it works and how to get the most of it by going through the prompt book and the tips and tricks channel, my question is: do we get kicked or lose access for inactivity, and if so, how long is that?
Nope! You won’t lose access for inactivity
amazing, thank you!
hi all i was wondering what i can do in a case like this where dall e has not captured the full shoe, is there a way to change aspect ratio to see all the details?
Not exactly. You can use outpainting to get the rest of it though.
Outpainting is the process of adding more content to an existing image with DALL∙E. Padded empty space is added around the image, and DALL∙E fills it in!
(There's much more you can do with outpainting, this is the most basic and commonly used. Check the #inpaint-outpaint channel for more ways to outpaint.)
Generate or select an image and use the !pad command in the #dall-e-bot channel.
Upload your image to DALL∙E and press 'Edit'. Do not forget to erase the watermark.
Generate it with a prompt that describes what you want to see in the new image as a whole.
thanks so much
Also
someone came up with this amazing idea @proud locust
if you start with a small black border and a blank image the rest of the way
it should hypothetically generate everything inside of the image
❤️
How to make a comic scene?
comic book? or comic strip?
Both
What results have you gotten so far using those keywords?
Does this help?
Woah So thats how you do it huh?
Which one take credit
Suprise me does
Can i use other people generated image?
No. You have rights to your images. You'd have to get permission from those people to use their images
I have an unfinished art .. i did in procreate
If i upload it and give some prompt
Can i copyright it for me?
Copyright it? It can finish the art. DALL E 2 has nothing to do with copyrights
Hey wanna ask one last thing .. Can people sell image that they created from dalle 2v
?
Yes.
Where can i see the images i created?
If you saved them, they'll be in your collection. Otherwise, the last 50 prompts are stored in the history
Where is history then?
Oh i am using mobile version so couldn't see that
Try rotating your phone into landscape
Can people see what i created ? Or is it private?
Your generations are private unless you share them
Oh landscape works thank you...
How do i delete the images i created?
You don't need to delete them. You either save them to your collection, or simply ignore it and continue generating images
Hello, do I understand the terms correctly, when I say that every single image I generate with Dall-E 2 needs to be attributed to Dall-E 2, regardless where I will use the image?
Idk if someone alredy asked this, but if after 1 month the user have a 15 credits refill, so the user can wait 2 or more months without using all the 15 credits to have more credits refill? Like, idk if this is the right word, but is it stackable?
• 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 after month after they were granted – but you’ll get 15 new free credits.
• Paid credits will expire after 12 months from purchase.
Thx
hi, a friend is asking how to gain access, and admittedly I forget how I got access... how can he?
They'll have to join the waitlist
is there a way to browse through other profiles and see their creations on the site?
Profiles and creations are private unless shared
Profiles are always private*
Images can be shared
You can look in channels like #art-and-illustrations and #top-tips-and-tricks to see some good generations made by other people
I wrote cowgirl in yoji shinkawa art style and it says it doesnt follow content policy. I dont understand does it think I am using the phrase cowgirl in a derogatory manner? Cowboy didnt give me any issue. I Tried uploading a concept drawing I made and I got the same warning. I dont understand
• Our filter aims to detect generated text that could be sensitive or unsafe. The filter will make mistakes and we have currently built it to err on the side of caution, thus, resulting in more false positives. We're working on improving our filters, so this should become less of an issue in the future.
• While we're not sharing the exact thresholds, a few warnings here and there won't lead to an account suspension.
ok thank you
hello, how can i change my account from Google Account to user/pass?
Email support@openai.com
thx, sent
@stark lantern omg, dude, congrat to your website 😄 very nice and creative
Thanks, happy to know that you like it, let's take it to #off-topic if you wanna know more...
is Mary Jane wearing a spider-man hood and no pants
yeesh. good catch
No definitive answer to these questions.
As far as I know.
thanks
Hello there! I was trying to get variations of an upload using a different art style. But I can either generate variations or edit the image which both produce results using the same art style, despite the promt using "in the style of Picasso" or "collage" for example. Is there a way of doing it?
Unfortunately DALL E is unable to do style transfer. That feature is meant for generating entirely new content in empty space.
Do you need the watermark to use your creations as a pfp
Nope
Still going to just incase lol
Not sure whether bug or intentional, but the word "furry" is tagged as possible content policy violation and cannot be put in prompts, why is this?
Lol why do ya think
