#community-help
1 messages · Page 3 of 1
OpenAI's working on an update for terms as of this moment. But for now, it's safe to follow Joanne's words. OpenAI just don't want people deceiving other people that the art is not made by an AI.
That makes more sense. I'll keep an eye on it in that case. Cheers
Like those users said and to my understanding the only thing you are not supposed to do at the moment is to deceive someone by putting forward an image from DE2 and portraying it as a non AI creation.
Very interesting issues. Sounds like #ai-discussions or #openai-chatter would be good places to raise this question, but the mods may have a more specific suggestion. You may want to take a look at the archived #biases channel that has many discussions from the alpha phase. The anti-bias features in the current system make it behave differently than during the alpha. The training data was curated, so there is selection bias in the images themselves. Hope you will post an example of what you are seeing!
Is there any way to change my name I put into OpenAI when creating my account on invite?
Hello have I correctly interpreted the terms of use in that I am allowed to sell the work created with DALL•E 2?
Yes you have commercial rights to any content that you create!
Great, thank you!
But please read the terms before you do, just to make sure!
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/
No current way AFAIK. Unless you signed up via Google sign-in?
I believe I did, because my profile picture on the website is the same as my google pfp
But I still put in a name when signing up
Speaking of the content policy, I have been wondering about a few dimensions of restrictions specifically "Political, shocking and sexual" — the shocking category seems a little vague but that aside in a broader sense political and sexual art has kind of always existed (e.g. protest artwork, nude statues or paintings, etc.). Is it the case that I can't even use the tool to generate work in those categories that's private or gated to an adult audience?
do you guys any idea how to create a photo with white background like a png file?
You could probably just say “with a white background” in your prompt and then run the image through a background remover which will turn it into a png
sometimes it just didnt work
Yep! For now, in the infancy of this technology, the filter is very strict. Even nude statues and paintings are not allowed to be generated.
Do you want help creating an output like that or converting a pre made image to a transparent?
Okay, thank you. I'll move my comments on this to the suggestions section....
You could probably just say “with a white background” in your prompt and then run the image through a background remover which will turn it into a png
Is it possible to create a sort of flip book? For example, if I took a picture dog walking, would I be able to create a series of DALL-E pictures that could replicate that photo but each consecutive picture the dog has moved a realistic amount in per frame?
If I upload an image, is it possible to ask for variations in the style of a specific artist or using different colours, etc?
Please check the #animation channel for this 🙂
As of the moment this is not possible
hey why can't you upload pictures with faces
Q: Why can’t I upload my own face?
• To prevent deepfakes, we are currently prohibiting uploads of all realistic faces, even when the face belongs to you or if you have consent.
• This is because we don't yet have a scalable way to verify whether you have consent, or whether a face belongs to you. So we prohibited all face uploads.
Click here to learn more about the Content Policy.
Does anyone know since when the 400 character limit was enabled?
Still have this prompt saved from when there wasn't a character limit :/
https://labs.openai.com/s/481YxIGZAGE3BfawAHpqSI2B
(Prompt done with gpt-3)
EDIT:
Solved this myself. Around the 25th of June the limit was reduced to 400.
Just searched for all messages from joanne and then I putting 400.
hey, is there any expert with the DALL-E 2 language here ?
You can try #❌┃prompt-help :)
Also there's a handy prompt book which tips
https://dallery.gallery/the-dalle-2-prompt-book/
i'm not talking about prompt crafting
i'm talking about the language DALL-E 2 created
basically DALL-E 2 developpers dicovered that sometimes the gibberish that DALL-E 2 creates when you ask it to create text actually means something for the AI
for example "vicootes" means vegetables and "apoploe vesreaitais" bird
is there an API endpoint yet? looking for the docs
You might be able to do this in a single image with something like Muybridge or another artist who photographed/painted animation sequences.
Is there way to upload a style (that you cannot figure out a prompt for) lets say a drawing style, but apply to another photo you upload?
#community-help message, there's no public api yet :)
Discord is the easiest way to communicate over voice, video, and text. Chat, hang out, and stay close with your friends and communities.
No, this is not possible right now
Any ideas on how to generate an image of an animal acting as a human trade - I tried "Full body photo of a Pinguin engineer drawing a plan of a very complex machine, 8K, Studio Lighting" but did not get the result I wanted - i.e. there is a human engineer with a penguin bystander instead of a penguin actually acting as an enineer?!
https://labs.openai.com/s/4DY67lEVNpxWVBgkc0Dd1qLr
Have a look at #❌┃prompt-help! :) People are super friendly on here and can often be of great help
@pine void ah, too bad. I find many styles i like and can only expand from that piece vs applying to another. Maybe future application..
Maybe, yeah! I'm sure OpenAI is working on very interesting things 👀 :)
What's wrong with this... Is it because I used the word "attack"?
I used invasion and that was ok... 🤷
Attack is the problem here, yes
hey why is it saying that there's a realistic face in the picture ?
it's all white ;^;
btw yes i have consent
With or without consent doesn't matter. Using faces of any kind are not allowed :)
there's no face, that's the problem
Please refer to https://labs.openai.com/policies/terms & https://labs.openai.com/about
not sure if anyone has already asked this yet,
- When will Dall e support different aspect ratio images?
- When can we add additional prompt for the uploaded image to create variations as we wish?
- When can dall e support higher resolution images, like 4k, 2k, instead of just 1024x1024px
Dall-E might "suspect" a face there, not entirely sure though.
okay
1: Might be a feature in the future, but as far as I know nothing has been announced. (maybe someone else knows more)
2: ^
3: There are already ways of creating a bigger image.
Take a look at https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (starting page 55 should be interesting to you)
& https://dalledictionary.notion.site/dalledictionary/DALL-E-DICTIONARY-2ec3dd6dc2184a45a9b6b4b8a021f0a6
Also have a look at #top-tips-and-tricks & #inpaint-outpaint to learn even more! :)
You’ll need to erase more of the face there
I'm just guessing, but this is probably filtered because it could be used as part of deepfakes by editing one face onto another body
The face it would impose is not a real person. The danger is the opposite
The danger goes in both directions, scrubbing a person's face from an image and replacing it with a generated one could be used deceptively to indicate that an actual photo of someone is fake (pretending the generated one is the real image that was modified to create the image that's actually the original). I'm not an expert on the logic behind exactly what's filtered and how strict it is and could be totally wrong, but that's the kind of thing I'd expect to be prevented just out of an abundance of caution 🤷♂️
Is there any way to buy extra tokens using card of unsupported country? Because Dall-e uses Stripe, which does not support many countries
Sorry, I was addressing according to the terms. I do get your point in general though
Lots of people have had luck with revolut in the interim until something else is out in place
Revolut lets you register virtual visa/mc?
So is there a built-in way to remove the watermark? I wanted to expand one of my Dall e results, however removing the watermark each time is inconvenient
You can download it without the watermark
But you'll have to convert the image
Inspect element
Does the real life persons content policy only extend to living persons or does it include dead people to?
If so, do historic persons, e.g. Julius Caesar fall under it?
Part of why I am asking is that I just saw a depiction of Marylin Monroe advertised as an example to try out.
Use the inspect elemt and download it, it will be downloaded as a Webp file, you can then convert it to a format you prefer. Let me know if you need detailed instructions on converting!
Is the watermark sufficient or do we also need to write that it´s generated by openai each time we use/post it somewhere?
You do not! You are allowed to remove the watermark as well, but dont decieve anyone by saying it's an image which wasn't generated by AI.
ah thats nice. so as long as i dont try to deceive anyone i can use the image without the watermark as i like?
even in commercial products?
Correct!!
Note: For certain types of sharing (live in-person or streamed demonstrations or social media demonstrations where you e.g. take audience prompts) you need to include a disclaimer about being AI Generated and indicating that you have reviewed it for potential harmful content, and you're also expected to look at and review the output before showing it to your audience (Basically you need to tell your audience that you are responsible for the content you are showing them, not openAI)
https://openai.com/api/policies/sharing-publication/
This only applies to (as I understand it) sharing your use of the tool, not the output of the tool itself.
Yeah I think this is when you are presenting the whole tool and not an output
I tried to, but it actually downloaded it as a html file
About how many image generations / upscales does credits provide?
1 credit = 1 generation of 4 images :)
Is there any info available about the Artist Assistance Program?
If you're interested you can apply for subsidized access right here! :)
https://docs.google.com/forms/d/e/1FAIpQLSfJzXwRXbyk3vwUtkoIVe17d4YiZ8gstikDUPCaERPqMZ8GFQ/viewform
hey guys how do i turn 2d sketch into 3d?
Have a look at #❌┃prompt-help and ask there :)
I don't think turning an already exsisting sketch into 3D art is possible though.
But you might try describing the 2D sketch and see where it gets you!
would it be possible to merge a 2d sketch and a 3d art togetherb
Why is the watermark removable with inspect element on the website?
The website applies the watermark after the image is generated, thus the original unwatermarked generation remains """underneath""". The file url can be obtained through a tool like inspect element.
Hi guys, can someone tell me if this prompt is against the terms of use?
"The panic after Annie Oakley accidentally shoots Kaiser Wilhelm instead of the cigarette during Buffalo Bills Wild West show"
I don't wanna get banned accidentally
The shooting part, basically any violence, drug abuse etc.
Please have a look at this https://labs.openai.com/policies/content-policy
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/
What payment options are offered to buy credits?
Are historical figures considered "public figures" under that content policy?
such as Abraham Lincoln for example
ugh... Please, understand that I am NOT trying to make Dall-e accessible using code. I'm using their text generating API.
You can pay with a credit/debit card :)
Thanks. Sadly I don't have a credit card
Good night!
OpenAI does not have API even for the testers of the DALL-E 2?
It would be neat to automatize continuous in-painting
#community-help message, no public API yet :)
Discord is the easiest way to communicate over voice, video, and text. Chat, hang out, and stay close with your friends and communities.
thank you
i thought that public is just for the more public audacity ahaha
anyway thanks
How can I make the ai represent a specific fauna I'm looking for better? I've first tried Moon Flower and then I used edit on the flower with this wording https://labs.openai.com/s/14mObmMFya40rwkXao0sIwyz
Consider asking about this in #❌┃prompt-help -- there may be more than one approach to try.
If I have a friend that wants Dall-E 2 and I have access, is there a way to invite them?
How do I make a extended image? Edit without deleting anything?
no
Okay, thanks for the answer
Does anyone know if it’s against content policy to create an illustration of a character that has cerebral palsy?
My friend with CP would like to see herself represented in that form… I don’t want to lose my account over it though
Are there any rules about API usage? E.g. https://labs.openai.com/api/labs/tasks
I checked the terms of service and didn't say anything saying we couldn't use them, but I'm curious if I missed anything
EDIT: I actually see this on the Discord Q&A:
🤖 Q: Is there a publicly accessible API for DALL·E?
• 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.
But... where does it say that?
Maybe try ‘anthropomorphic penguin’?
Any way to change the background of this with Dall•E
Hello Im new to this, I just wanna ask if I typed in ''Girl sitting in the middle of galaxy'' was it only gonna cause 1 prompt? cause I received 50 prompts for the inital free.
yes. 1 credit = 1 prompt
Thank you! ❤️
Yeah, but it's gonna look better if you remove the white part yourself first. The Dall-E background removal thing you do with your mouse isn't very accurate
Alright thanks
https://labs.openai.com/s/1QtNxZm5iyq4bR4LKg8bfkCs Hi team! why do you think the dog is not riding a black horse in any of the variations?
How does one remove the logo at the bottom right? I'm trying to use an image for textile patterns
How do I get to the actual page to make art stuff? I can log in but I can only find links to code generation stuff.
hey guys, is there a tool that generates image descriptions that can be used as prompts for DALL-E?
Say, I have image and the prompt generated from it will make DALL-E to produce similar results, as if I ask it to generate variations of the original image.
This is a really important question, thank you for asking! The Content Policy is a little ambiguous on this. We as a team are discussing it and awaiting clarification, but as always for now read the policy over for yourself and if you're using words that the filter blocks maybe choose a different subject for now.
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 manage to create an "infinite" zoom in? I found how to outpaint, but idk if the zoom-in will look good
Do a zoom out and run it backwards 😅
quick question: is the hash in the url of an image reversible?
thank you !
https://editor.p5js.org/Pi_p5/full/qAqoieAhx
upload to editor
download
upload to openai edit feature
click anywhere, process with the same prompt
repeat
A web editor for p5.js, a JavaScript library with the goal of making coding accessible to artists, designers, educators, and beginners.
i'm doing it with the Dall-E bot but i don't think i will manage to zoom in to the original image
Is there a way to access your previous prompts/runs on mobile like there is on desktop?
I just read an article about this here https://mixed-news.com/en/prompt-design-for-dall-e-2-photorealism-emulating-reality/
what year is a dataset?
try to put the page in computer mod
Thanks!
Oh thank you! Good tip
hello 🤍. I just received my invite. I uploaded an image to create a wide view and it worked but the image quality is not as sharp now. Is that normal? Is there anything I can do to keep the quality as the uploaded image? TY
this is a really weird question
is generating anime characters with the background flags of the countries: like Ukraine, US, etc... is ok? or it is considered as something "to influence the political process or to campaign"?
*background flags
ok so here's a little zoom out but now idk how to make an "zoom effect" like in this video: https://youtu.be/0WKckKlRM3M
How Can i make that?
Images created with DALL·E 2, an AI system by OpenAI.
Edited using Blender.
Music: DivKid | The Empty Moons of Jupiter
Stupid question i was doing some generations and mistakenly closed the window with the generator. Did i lose all the previous generations history?
near the "Generate" button
click on the history button
Thanks!
if you are from mobile use PC version of the page
Im on PC
Hey guys sorry but I'm new. Does anyone know how to change perspective in an image and I'd really like to create very large structures mostly inspired by Zdzisław Beksiński. Or eerie large cities with very tall buildings compared to a person (it always seems to get the person exactly the same size not miniature compared to something massive like I intended)
Thank you for the reply… I went ahead and ran a prompt with “cerebral palsy”, it wasn’t flagged and the AI did a decent job of interpreting the disability! Still, I’m glad it’s a topic that’s being discussed 🙂
nobody?
I'm not sure I got flagged by Joe Rogan interviewing an Alien
I think if you get flagged by something not outright offensive your account shouldn't get deactivated no matter what it just doesn't generate that particular thing (I read this somewhere don't trust me on 100% though)
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.
How much of gore isnt allowed? I want to generate realistic anatomical and surgical photos to see if it can be used for medical teachinf but dont wana get banned
Try reading the content policy there may be something about it. My guess is if you're descriptive enough with the use of medical terms and such there shouldn't be much of a problem. Otherwise try using something like anatomical figurine or similar terms which may dial down the gore.
Thank you
you can try generating a pic of the anatomy of the human body, but i doubt it would work
i wouldn't try anything further than that though
Right, thank you
no problem!
Can I please get a reply?
yup!
It could be such a cool tool for medical teaching as we wouldnt need to rely on an actual person having the said disease of that exact variation/iteration to see. Would be amazing for surgical teaching
Does anyone know how to change perspective in an image
there isn't a definite way to do this in my opinion, but try describing something in great detail, with a modifier having to do with the camera angle 
yeah, maybe in the future! it just doesn't know the human body fully yet though haha
this seems like a bad idea regardless of what's allowed
i don't think it'd be very accurate
yup
ok thanks can I ask you about these so called modifiers? how do I work with them?
just add them to the end of a prompt
ex. "high camera view", "view from the side"
you can also specify lens types, lens distance, etc.
it's so fun
oh I get it thanks
yeah -30$ already XD
same 
welcome! and sadly not really, quality gets lost in the process/it isn't picture perf yet
um, if they're not campaigning irl then it should be fine!
just my two cents as im not an employee and im not totally sure
Would it be possible to change my profile name so I don't have to dox myself from sharing my creations?
You can try worm’s eye view or megalophobia. Vanishing perspective might work too
https://labs.openai.com/s/S0ER6arjLw7c9qmQvwHDKIVg
i am wondering how this was made, i assume they uploaded an image, that took up half the screen and then dalle did the other half?
I've been using golden rule composition to frame/ center my images anyone have other centering/ framing keywords?
Hi! Can anyone give me the best way to increase the image output size?
I've seen people use isometric which looks pretty cool
Over the course of today, I have received 3 conduct warnings for some pretty innocent prompts (e.g: "the walking dead comic book art") saying I might get automatically suspended. Should I be worried about a ban?
Quoting FAQ.
• 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.```
Have you signed in with Google?
Thank you for the help.
Np
Q: Is there a way to change the aspect ratio so it isn't always square? I need wider assets with a custom aspect ratio.
As of the moment, no. As the model was trained on square images. To make a custom aspect ratio, you'll need to generate multiple images and stitch them together. There's a bot in #dall-e-bot that has the !composite command. A native #inpaint-outpaint interface is planned in the future.
Hmm ok, let me know when you come out with aspect ratios. It's kind of important for me to be able to use the assets. Unless there's something I can add to the query that would ensure the assets have some margin where the main focus isn't there, so I can safely crop it. Is there something I can add to the query to make it easier for cropping?
When can we expect the next big update for the model? I really hope you guys will fix the upscaler and how it currently generates eyes.
What is the "padding" in the dall-e-bot channel doing? Looks like it's just making the image smaller. Why are people doing that?
not sure, gl!
look out for dalle 3
to my knowledge dalle 2 will not be updated with new imagery anytime soon!
people are uploading the png to dalle and extending the boarders making a wider image
try it! remember to describe the previous prompt that you do and erase the openai logo
not at the moment :(
Is it possible to retrieve a Dall-E output that has aged out of the history view?
If they're gone forever, that seems like a really bad UX. I end up having to spend a bunch of time manually downloading images and prompts every single time, just in case I might need it later. I'm hoping that there's a way to retrieve older inputs and outputs.
same
there's two ways to do this though
- (using an image from your collection)
click on the image, and then the three dots and then edit. from there, press cancel, and then back, and you should have your previous four images generated from that specific prompt
- something else you can try with other images!
check your history tab with links starting with "https://labs.openai.com/e/", clicking on one should show you the complete spread like above (depending on how old it is it might not work)
(also the image has six generations bc it was made before the four pic change)
Thanks for the suggestions.
- Is not useful because it means I must save at least one image from every single generation regardless of whether I like it or not. This defeats the purpose of having a collection feature to begin with. And even then, I must remember that the image I'm looking for even exists with no good way of searching for it, so I might as well save every single image generated.
- This would only work if I don't clear my browser cache, and also provides no visual index to speak of, meaning that I'd have to reference items by prompt instead. Prompts are way harder to remember than images.
I feel that this is an issue we need to raise with the UX team at OpenAI, and I'd be happy to provide this feedback in a more structured form if there's a good place to direct my feedback.
whatever feedback you want to give, #suggestions is a good place for it

hey
What do you need help with?
how to i outpaint in dalle-2?
What's your understanding on the term outpaint? Just making sure we're on the same page
"zooming out images"
like
painting everything but a subject
e.g
Okay so you want that zoomed out and DALL∙E will output a largerzoomed out image?
Gotcha. Now, go to #dall-e-bot, type in !pad(number here) the bot will then zoom out your upload with the specified number.
And then go to DALL∙E, upload an image, press 'Edit', then describe the full image that you want to see.
Not sure if you've seen the resources at #gpt-models, there are visual guides there
and what will it pad out
It will resize it to 1%, I suggest !pad40 or so.
okay let me try
Yep, ping me if you need more help!
now upload and paint
Correct!
That's great!
I forgot my password, and when i send the reset request idk why but i got no email from open ai now how i can reset my password ? Can somebody please help?
Hi, I'm new to the discord and recently to Dalle2. I just have a question about the images I've generated already... I can't seem to scroll back past a few days... are those older images lost? Should I be downloading them all the time to be sure? I noticed the "my collection" only has a few images that I pressed "save" in Dalle2. Thanks for any help! 🙂
We're currently waiting for OpenAI's advise on how to handle this. For the moment, I recommend contacting support@openai.com or the live chat at beta.openai.com (not sure if it's open now though!)
Your collection will only show those images that have been manually saved by you. 'My Collection' currently supports up to 10,000 images so feel free to save! The sidebar only shows the recent 50 generations. For now, to check the rest, the only way is via your browser history.
I had the same question... What do you mean by checking browser history? As in unique URLs?
That is correct.
Hello, has anyone seen behavior like this, where Dall-E 2 delivers results that seem like components of a larger composite image?
This is quite common, yeah.
You could try and include words like "wide shot," "full shot," e.g
For 2D it can get quite tricky sometimes, #❌┃prompt-help could be of help! :)
Thx for the tip!
Am I allowed to use whatever I generate as an emote on twitch? Like is the copyright abuse or something
Are we allowed to generate things like military vehicles (tanks/planes/ships) as long as there's no combat or blood or violence in the picture/prompt? Like just pictures of a tank for example
You may use it for commercial use. But just to be sure, please read 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/
I am not sure as DALL∙E's filter is strict, only one way to find out!
Well the question is, would I risk a suspension for generating pictures of military-related stuff that don't feature actual violence?
Give me a prompt, I'll try it out for you 🙂
A humanoid cat wearing a military outfit and a WW2 helmet, digital art
i dont think it likes the terms WW2
Yeah that might get flagged, I'll still try it out, hold on
Beautiful results.
Prompt for this?
A humanoid cat with ginger fur wearing a military outfit and a world war one helmet, digital art (that one got flagged)
could try
A humanoid cat with ginger fur wearing a military outfit and a WW1 helmet, digital art
i wonder if it could make a pickehaube
Okay, first one got flagged
Probably because of the word war. But WW1 & WW2 was in the past so maybe it thinks that it's okay.
Just speculation though!
Mostly it's automatic. Not sure about the threshold. Reactivation on the other hand is manual.
I see
So would my military stuff violate the TOS or is it fine since there's no actual violence? I just like cute pictures of cats in uniform lol
In my opinion, I think those are fine tbh. But I'm not an OpenAI employee so don't take my word for it.
Alright thanks
Hi, are generated images in Recent going to disappear after some time? It’s unclear to me what happens to images you didn’t Save
(Just noticed the New button and almost wiped my entire history. Still not sure if these images will stay there if I restart my browser or computer for example)
The recents tab only supports up until 50. 'My Collections' however supports up to 10,000. But you will be needing to manually press 'Save'. To view your previous generations past the current 50, you may do so via your browser history.
Ok thanks, that’s good to know!
Hi, is there any way to change the aspect ratio of the images?
I guess the idea is you download it, pad it as much as you want with transparency, then upload it again to Dall·E to complete it, then crop it to get a different aspect ratio)
Ok thankyou
Is the word christmas suspended from dalle? Why?
What's your prompt?
How come this server doesn't show up when I open my normal Discord App. Do I have to click 'Join The Discord' every time?
I got some content policy issues with my prompt. I want dalle to create a christmas card. My prompt is:
A photo mostly red, with a fir tree made out of golden and silver glassballs, shiny silver stars, snowflakes, fir branches, vector graphic.
Any ideas why?
It's working fine on my end. Even tried with a non-volunteer account.
Cristmas and vector graphic not the problem
Let's transfer in #❌┃prompt-help
Postet it there already
Thanks for answering. I'm really not sure what to do.
Try updating/reinstalling your Discord.
I see what happened. Somehow I created a new discord account using my Google Account which is not what i wanted to do. I wonder if there is a way to merge the two accounts.
Not at the moment no, but it is in the works for the future...
What is the color reference bar in the bottom right for on my generated images? Is it there for some sort of Dall-E watermark? I'm not fussed about it but I'm just curious
Dall-E needs some sort of hand tool. It’s almost always the hands that ruins a good image. Faces can be fixed with FaceApp but there is no good solution for bad hands.
Yes it's the dalle2 watermark, you can remove it if you want to
Thanks!
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.
Refer to the above
Even the computer cannot draw hands.
What commands are available for the dalle bot?
#dall-e-bot message have a look at this :)
Does anyone test their prompts with another ai before putting it into Dall-e? If so, are there any recommended ai for prompt-testing?
(I used up my first 50 credits just experimenting and I'd like to avoid that when I get my next 15)
I haven't done it myself, but maybe someone else has(?)
To avoid getting not wanted results in the future though have a look at these two links
https://dalledictionary.notion.site/dalledictionary/DALL-E-DICTIONARY-2ec3dd6dc2184a45a9b6b4b8a021f0a6
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y
And also at #❌┃tips-n-tricks #top-tips-and-tricks! :)
Different AIs will need different prompting. I recommend to just spend your time with DALL∙E and try to learn how to communicate your ideas with it.
if I buy credits, and I get the monthly free refill,
does it use the free credits first? (probably a dumb question)
No dumb questions here! And yes, it does use up the free credits first.
I'm trying to get landscapes and such in a particular style but i can't describe it, it nailed one of my queries randomly and i can't recreate it
Where would it be best to ask the style of a particular image?
I think its best to move this to #❌┃prompt-help
Alright, thanks
the pandora box was open long ago, and our expectations are outdated. I think it's inevitable that we'll live in an environment where any piece of information is false by default no matter how real and reasonable it seems. Maybe NFT will help to separate signal from noise. But in general we need a paradigm shift in a way how we map reality and how we relate to information
AI Police 👮♂️
Will DALLE accept paypal in the future?
This may be to dumbest question of all time but how the heck do I login? I only see an API login on the Open AI website but no Dalle 2 option? I know it is super awkward but please help!
Thank youuuu!
Is there a possibility of getting Dalle-2 and GPT3 credits to make a science show? Me and group of students from my uni are going to certain festival with our stand and showing public power of GPT-3 and Dalle would be quite interesting experiment to do, is there such an option? I mean there is no clear way to do it but maybe someone here knows where should I write to get info?
Question especially to <@&998386466063396966> and OpenAi people
The blog post for the beta launch includes reference to a subsidy program, if you want to read up on it:
https://openai.com/blog/dall-e-now-available-in-beta/
I wish I could help you, and I'm sure OpenAI are considerate and interested in these kinds of cases, but it seems they're quite flat-out right now; depending on when your festival is it might be wise to consider other fundraising options. Definitely try to go ahead with it anyway!
Thanks!
Maybe print some AI art and frame them in op-shop frames to sell at your show and make your money back?
(I am obsessed with this idea. One day I will do it.)
Haha sounds great ngl but not sure if we can sell anything this way
Edit too much "haha"s
festival is practically from tommorow and it will end August 7, do you think there is a chance for OpenAI team to reply on time?
Hi! Can someone help me understand the "Public and personal health: the treatment, prevention, diagnosis, or transmission of diseases, or people experiencing health ailments" part of the content policy? Does this include all forms of medical diagnostic approaches such as the use of stethoscopes or ECG patterns? Are syringes and pills banned? Say I would like to make some illustrations of what happens at a hospital, is it allowed at all by the content policy?
This is one of the trickier dot points to give clear advice on, imo. I believe there are reasons for these rules, especially in a post-Covid era. I'm not sure, sorry. I may be able to get some extra clarification down the line.
🥲
Thank you! OpenAI seems like a very sensible company, so I guess one will find understanding as long as there are no malign intents, and that the ethical borders will become clearer as time goes on
The word war is banned and as per the content policy, generations related to any type of violence are not allowed.
Does anyone know if openAI deletes older images of users or if they archive them and we will be able to access all of our image grids in the future?
Some context of this question - I created around 800-1000 credits worth of images when I first started using Dalle, while creating them the URL histories were not saved in my browser history as I was using Private Mode in Firefox. While creating the images I would download some, but I was under the impression there was an archive of the images that I could access so I did not download A LOT of images that I eventually wanted to download. I was wrong about there being a full history I could access on my profile, so ever since I have been using with my browser history turned on.
But even if my browser is on I wonder if OpenAI will delete these older images, I wonder if they do it for storage reasons, to lower the costs of storage, anyone with any knowledge on how this works and future plans for users to be able to access their entire image history please share.
Currently they are not being deleted, but chances are that won't last forever. The ability to view your entire history has been discussed and might be something planned for down the line (I don't know of anything concrete).
I recommend reading through this discussion with Joanne
#community-help message
What is the best way to upscale a Dalle generated image?
Thanks for the info. I emailed openAI telling them my dilemma and hopefully they will be able to somehow give me access to my older images, as there were a whole bunch I wanted to download but never did, worried I will lose all of that.
There are several tools - the most popular are a Google colab notebook called esrgan
Another is a desktop application called gigapixel
Can multiple people use single account from different locations? I have a design company and my team need to use Dall e. Will it cause any issues? result in account termination? @everyone
I have seen nothing against this. However, do not tie multiple accounts to the same phone number. Also make very sure your team understands the terms and content rules.
If your account gets banned, your credits will most likely be lost.
I actually just got banned and my credits were refunded
Since Dall E gave users the right to commercialize their art, does that mean putting the art on t shirts and selling them is okay? I have some generations that id like to print and wear myself
Any reason why some historical figures names are fine, while others give a content policy violation warning?
I didn't realise there was a time limit for previews, before you have to save or download a generated image. Is there any way to recover older generations in a session?
Not a time limit, but the history sidebar only shows the last 50 generations (unless you click new session). Beyond that you'll need to have saved the /e/ url or have it in your browser history
You can post it in bugs. Historical figures should be fine
Where is the bugs section? On discord?
Under help, bugs-and-issues
Thank you, I'll have a look. DALL-E generated some beautiful gold statues in a museum display case, and it's sad to think that that art might be lost forever!
I don't seem to have that under help (or I am not seeing it). I have tips-and-tricks, prompt-help, questions, and dalle-e-bot
Does that work for you? Sorry, it’s under feedback
It worked, thank you
Going back in browser history works, and I can download the images
Do you agree with then ban?
Yes, you can sell the art. https://labs.openai.com/policies/terms
As in, did they have a case against you?
I'd like to buy some credits, but am from Serbia. any way to do it ?
Some people are having luck using revolut
thank you, im so slow lol to clarify i can literally buy my own ai artwork basically?
Why can't I login using the web browser? I've accepted the invite on my phone and did a bunch of prompts and allnwas fine, but when I try to login using the browser on my laptop I'm getting an error :(
What error? Are you using labs.openai.com url?
You can sell it, you don’t have to buy it
Tried that url, now it works fine. 🙏 Thanks! I guess I had a wrong url
It’s confusing with the regular openai site 🙂
Hi, I know this will be a dumb question but I can't get back to the page where I put in my prompts to Dalle 2. I have done a few yesterday and I have logged in to Open AI like I did when i signed up but now I can't find a like to a page where I can actually input a prompt. Any help appreciated
Therea are no dumb question, dw! :)
https://labs.openai.com
Genius! Thanks @pine void
Is there a way to feed our own photos to Dall•e and provide our own photographic material for its learning?
I am not talking about editing and processing an image, just straight out providing reference material
Not that I know of! :)
#gpt-models message *
(if I should be wrong I'm sure someone is going to correct me)
Yeah that seems super harsh! Bizarre
That’s incorrect information, see the FAQ
oh sorry! I missed that!
Ive deleted the message
Thank you. I've submitted my request!
Thanks for trying to help the system improve 🙂
Is it not possible to use the edit feature on DALL-E 2's own output of human faces?
It's a shame to let all my science/nature photos gain byte dust while it's literally effortless to provide them for training
I've tried 3 times and it's refusing to do anything with it after making me wait a minute. I didn't upload this face, it was something generated
No real error messages
The system is still in beta. Perhaps at some point, but for now, you’d have to remove the face and edit around it. They’re erring on the side of caution while it’s in beginning phases
when i search to generate 3D render of sslv rocket, dall e 2 gave random satellite model, why ?
The one that just launched?
Hi, quick question, will the history of my seraches and generated images be saved forever?
or I should download them?
no but it is gonna launch on 7th august
You should save them and or download. Only 50 generations are accessible
ah I see, okay
The ai isn’t training on things up to the moment, so it doesn’t understand it.
Am I understanding you correctly? You generated something with DALL∙E and you pressed the edit button and now it's giving an error for realistic faces?
ohk thanks
I misread, I thought it was an upload/edit. I know that doesn’t work at the moment.
Yes, that's fine, he clarified it in the second message.
It said something about "an unexpected error occurred", didn't say anything about faces. I just want to be sure because I don't want to get suspended for trying it over and over if it's not supposed to work is why I ask
Lost that in the shuffle. That seems like a possible bug there
Ah, that's a common error nowadays. Just re-edit it and it should accept the face as it's from its own generations. It should only block faces that are uploaded even it's from DALL∙E.
I tried it 3 times, first time "Unexpected Error", rest of the time didn't seem to even error but didn't output
I'll try one more time
Okay this time it worked, odd
Great news 🙂
can i remove trademark from the image i create?
Yes you can!
how?
Either with 'inspect element' in your browser or with an image editing software.
Have a look at this :)
#❌┃tips-n-tricks message (at 1:20)
thank you very much 🙂
so... i can't save anything on Dalle? Only the last 50 prompts will be ever shown, and only on the website (not on mobile), right? 😦
Therefore there is no way for me to find what prompts I've used, what work / what didn't etc. unless I download each single image and write down the prompt in a spreadsheet or something. 😦
There’s a save button on each image
The ones you save store the prompt
This is what it looks like on mobile.
In addition to the collection, if you have browser history enabled on your mobile browser (assuming that's what you're using), the /e/ addresses will be saved there
Hi, so I accidentally made a prompt that caused a warning, and it said if I kept doing it my account might get suspended.
Should I worry about this? I forgot about the rule against weapons and such, so I had the word "gun" in the prompt.
I can't see it on the browser?
You have to click on the images
You save each one individually
oh... I see. I thought that 'save' was 'save the image in your hard drive'
haha, the UX is a little terrible.. I guess that's beta! 😄
Don’t keep repeating the same type of thing - violence, war, etc are all prohibited
Of course, I am avoiding it. It was meant to be in a cartoony context, but I know it was a mistake.
I just wanted to know if I should worry, I am loving Dall-E so far and it would break my heart to lose access already
A single warning won’t hurt, but don’t keep accruing them 🙂
Okay, only got it 2 times so I should be fine?
So, if I want something with a toy water pistol, what should I do?
I guess "Super Soaker" might work.
Definitely gonna read my prompts through several times now to make sure it won't be misunderstood and flagged
Just a question. Can someone crop out a portion or use elements of the generated artwork and use it provided they give a citation saying it was generated by Dall-E 2 AI?
I believe so. You have all rights to your images
If you’ve hit a gun warning maybe find a different topic for a bit, super soaker might be okay but I’m not openai
Hey! I've run into an issue that made me waste prompts a few times; how do I remove a reference image so that when I generate a new prompt, it does not use the image as a base?
You don’t have to give a citation but just don’t mislead people into thinking it was created without ai
Okay, thanks for your help! 😄
As in you uploaded and then forgot and typed a new prompt?
Is there a way to add someone onto my account, like with their own user and login but we share credits?
Well for example, I could use it on a website and then at the bottom put an image credit note saying the image was created by AI?
Just click the dall-e logo at the top to reset it
Oh okay, so no other way than to refresh or click a recent prompt that did not use an image, thank you
am i allowed to share my account with someone who would use it for the same purpose as me?
So derivative works are one thing. If just cropping Just don’t say you took a photo
I have seen nothing that would prevent sharing but your co-user could get you banned if they don’t follow the rules
alright thank you i will make sure he doesnt do anything stupid!
also you might wanna keep an eye on your token usage to make sure that you dont run out because of too many prompts from another person.
I saw it asked before but never answered, a lot of times we are trying to follow the content rules but something accidentally triggers the warning and warns you you could be banned with more warnings, that makes me, and the other users I have seen asking it, nervous to enter more prompts, get creative etc... does anyone know how many warnings you can receive? For example, getting a warning for the word body even in a g-rated context. Or if you got banned could you appeal and show you were asking for things that did indeed follow the content policy??
There is no set amount of prompts. If you got a warning for ‘shooting’ and then for attack, and another for war it would not be viewed favourably when being reviewed
For instance, if you bumped up against body, and then against something unrelated, you’ll probably be fine. If the review team thinks you’re trying to skirt rules to get to a violation of the tos you may get your account deactivated
But if your target image is g rated, just try to rephrase. Drop a prompt in https://discord.com/channels/974519864045756446/990336929142808656 if you;d like help
imo its pretty easy to avoid it. No violence, or words that describe violence "shoot" for example. No NSFW content.
No celebrities, no real faces
Its all in the content policy
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/
Where can I find the easiest instructions for inpainting?
BT21 characters in Ufa, Russia, Professional Digital Art got blocked
USSR got blocked
countries are now blocked?
Was "Dall-E mini" trained using a different system than what the proper Dall-E 2 has used? I've used an artist's name in Dall-E mini with next to no prompt other than their name and some colors and it outputted stuff that looked very similar to their work, but doing the same thing on the actual Dall-E 2 makes it seem like it hasn't been trained on the artist at all even though the spreadsheet showed someone got similar results. An artist my friend used on Dall-E mini doesn't get recognized at all on Dall-E 2 it seems like. So were they trained differently from different things and that's why?
Will there be a different way to get credits other then credit card in the future? ( i.e. paypal, visa, ...?)
They are not affiliated
See the DALL E Prompt Book in #gpt-models
Starting at Page 56, they have information on all of those processes
Russia is in the middle of a war and probably triggering political filters
that's really sad because the city architectures, landscapes and nature is getting blocked by this
I know. You can try the name of a building or city that isn’t in the news perhaps and avoid the country name.
Thank you
yep
Any reason why it would take so long to load? I have decent internet, but the images in my recent sidebar take FOREVER to fully load. Is it a Chrome thing?
It might just be an excessive amount of users
Unsure
Gotcha. Since I got access yesterday things have been taking like all day to finally show up. But I know it usually doesn't take that long.
So, I am trying to find the limits of the filters when it comes to inspiration. And they have currently got me quite confused, so maybe someone can help me figure this out. I am clearly not allowed to rip off Star Wars in an artwork. But I should be allowed to create a Star Wars inspired spaceship, right? That is well within the boundaries of the law, when it comes to creative artistic freedom. So, if we take two massive franchises like Warhammer and Star Wars. I can easily say I want a Warhammer 40k inspired dystopian city. But I can apparently not ask about a Star Wars inspired spaceship along the lines of a battle cruiser? Is it a bug, or is Star Wars just such a big franchise that the company takes no chances at all?
It might well just be wars
I can create a medieval warhorse but not a medieval war horse
As a die hard Star Wars fan, this makes me happy
hi, I realized that older images are being removed from the dashboard automatically. If I want to preserve all my generations, should I download them manually?
How do I add stuff to my collection?
As of now this is not possible :)
If you click on an image you can click save to put it into your collection
@solar minnow
Thanks
yes, but I wanted to check ALL the images. I suppose a browser extension of some sort would be necessary.
@quick fog not sure I understand
I believe he wants to add all of the 4 generated pictures to his collection at once.
(which is not possibleby default right now either)
if I scroll to the bottom, I cannot see older images.
You can use your browser history to go back further. It’s a pain
But if you need an image you didn’t save it’s a lifesaver
that's why I was thinking that it would be doable using an extension that automatically saves all your images locally. Or maybe somewhere like S3
that's a good point, I figured they automatically delete the older ones
Wondering if we ever heard back from the team regarding cannabis related imagery given the generally wholesome status it has in many places
I don’t want content warnings so still holding off 😂
Still waiting I believe.
No prob, just keen to make some flower icons 😄
is there a list/site with all the different supported art styles?
Please see #gpt-models for style guides.
DALL-E Dictionary
- 100’s+ of FREE examples 📃
- Art Styles, Prompt Presets & more 🎨
- Integrated with the official DE2 Discord
https://bit.ly/dalledictionary
hello, can anyone answer me this, what are the rules for selling the art generated here? not just entirely from text prompts but using an artists OG artwork? take for instance i upload my design to use as the base, are the 3 resulting pics "mine" or are they "open or free use"? any kind if help on this would be appreciated, TIA
Interesting question
Let me get some additional feedback
Actually its in the Content Policy
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/
theyre my original works
If they're YOUR original works than yes, there's obviously no issue
what about art from just using the text prompt is that open use?
But if you're trying to create a variation on another artists works, that would be against the Content Policy
yeah, i understand how copyright works, but this is such a new and gray area when it comes to copywriting an AI's work
okay that answers my question then, thank you!
got it, thanks
@sick mason Is this what youre looking for?
ok great, thank you for taking the time to look it up for me
np
Are generated pictures public or private
Private by default. If you publish an image and share the link/id then it's public
Only to those you send the link to if I’m correct.
It's a real shame celebrities are off limits, although the reasoning against it is pretty straightforward.
It would be so lovely to see a v
Van Gogh painting of Edgar Allan Poe and Macho Man Randy Savage in a rap battle officiated by Mr Rogers
Question: if I’m using dall-e generated art for a story, etc that includes violence, and I don’t generate violent images but alter them to show violent scenes, does that violate the content policy if I explicitly explain that it’s an altered AI-generated image?
I’m thinking I want to make a comic entirely out of dall-e generated assets but there are two scenes in mind with mild violence
Pretty sure I know the answer to this but wanted to ask it explicitly anyway to be certain before I make it
Wombo.ai has the tool you need
I keep getting flagged for variations of "A guy wearing a loose fitting hoodie" which I am almost 100% sure isn't against any of the content guidelines. Anyone what is triggering the content filter?
What are the words that are staying the same across attempts?
You said you’re making variations
Those are the words that are staying the same
I don't want to keep trying because Im scared Ill get banned but those are what Ive whittled it down to
“A guy with a loose hoodie” works too
But loose-fitting hoodie does not work
Probably because there are a lot of NSFW images of people in loose-fitting hoodies
I dunno lol - might want to bring it up to the team
Which chat is that?
is this statement accurate? why it contradicts what is stated in the content policy or is the content policy outdated?
This is accurate. What does it contradict in the content policy?
@verbal crag whenever you get back on you’ve answered a couple questions similar to this one about the project, which I appreciate — could you let me know what you think here when you get the chance?
Might be a server load issue
A lot of people (myself included) recently got let through. I haven't told anyone because they're gonna be dogging me for prompts. Honestly though, it's not that expensive I'm just a cheap bastard.
Nice to meet y'all by the way.
sorry, I was referring to the TOS section 6 which contradicts the screenshot I shared above
Hello, I want to resize the image to 700 x 400 px for my featured blog image. Is there a way to do it properly without distorting the image quality as well as DALL.E watermark? Thank you.
so which one of these two is accurate?
openai will not resell generations you have created, but you can use it for commercial projects yourself!
Thanks for clarifying.
np 
I'm not an OpenAI employee but I'm familiar with the terms, IMO this will be in violation of the highlighted:
Why are there so many channels here? Where is the interesting discussion for people like me who are probing the capabilities and limitation of DALL E 2 in a serious-but-snarky manner? This looks like a nightmare.
They all have purposes. There's a bunch of topic/medium based ones so you can share art you're interested in. Maybe #ai-discussions is what you're looking for?
Hiya. Sorry to be a bother
I read conflicting information and now I'm really confused
Can I use the AI generated image as an NFT to sell? If yes, are there any conditions required?
A statement from DALL∙E's product manager: #community-help message
The current terms however are still vague about NFTs so personally, I would wait for it to be updated.
Yaa, i thought of the same thing. I love the results but if i have to take it down, that would really hurt.
But DALL∙E NFTs have been sold, so I guess it's fine. Do note, I'm not an OpenAI employee.
Sorry i know you get this question alot and thanks for replying
No worries!
Yes, no problem, i can wait until it is clear.
There are yet no offcial terms for removing the watermark and then selling the art being created, right? :)
Official terms for the watermark, not yet. #community-help message
Joanne has shared that they are updating the language for the terms regarding crediting/attributing DALL∙E.
Selling the art however, official terms for that are already up.
Yeah I saw the statment from Joanna, I'd rather stay true to terms though. :)
Thanks Nukke!
Question: I was wondering if there was a list of specific terms to avoid, as I was inspired by a quote from an art gallery in Cambridge, and wanted to use it to generate art, but got an automatic notification that the query may have violated the content policy.
Quote was "Without water, the landscape is dead, especially at night" - Pierre-Henri de Valenciennes [context in museum was above several 1700s/1800s paintings that prominently featured water in landscapes]
Hi! OpenAI have not been releasing specific terms to avoid as I'm guessing this is to prevent people to circumvent the policy. I would just infer to the best of my ability from the content policy to avoid more warnings.
Hello everyone, this may have been asked before (but I don't know how/where to find the answer)... How does one "reset" dalle so that previous prompts don't affect new prompts?
I'm not sure what you mean. Every prompts are "new", previous prompts will not effect DALL∙E's succeeding outputs.
Yes, they do. I tested it. I'll show you what I mean.. (a sec)
First: "sign warning of reindeers on the road"
After that: "sign warning of animals on the road"
I suppose this feature "helps" people home in on what they are trying to achieve but sometimes it is undesirable.
I generated your second prompt, I have never generated reindeers, this is the result:
Now it generates reindeers for everyone! 🙂
This is a clear sign of bias for DALL∙E, more information here: #announcements message
Please flag the results so it reaches the team! If you're specific though, it generates this:
sign warning of chickens, on the road
I stand by what I said just now:
Every prompts are "new", previous prompts will not effect DALL∙E's succeeding outputs.
I have even better example where it used no less than three "concepts" from my previous prompts even though I didn't specify them in the latest prompt.
"world championship wrestling match between santa claus and kangaroo" ... I had used "boxing", "reindeer" and "sumo" in previous prompts.
You can see all of those in this one.
This is undesirable if I want to "start from scratch".
Here are my results. Seems close to me. I have never generated boxing nor sumo before this.
Seems like it's interpreting Santa's belt for Sumo Wrestler's attires. And it thinks "Wrestling Match" is the same as "Boxing".
Again, this is a perfect example for bias which can be easily avoided by being thorough with your prompts and specifying more.
In this case, the Content Policy may be interfering with results as wrestling may be interpreted by DALL∙E as violence.
Not yet.
ok that would be a great feature when its implemented
Indeed, hopefully soon!
Actually, "boxing" is apparently violence (I got "something went wrong" message) and wrestling is OK. Yet, Santa and kangaroo had boxing glowes on when I specified "wrestling".
So even the rejected prompts seem to affect later prompts.
I was an anti-virus software developer and researcher in my previous life and did a lot of penetration testing on our own software.. so I tend to notice these kind of things.
Perhaps this should be reported to OpenAI/Dalle team. Although using this loophole is not very straightforward and the results are unpredictable.
What I originally wanted to produce was "world championship boxing match between santa claus and jesus"... but "boxing" and "jesus" both seem to be forbidden words or concepts.
I got Santa Claus to fight with reindeer and kangaroo, though. I wonder if by using some cunning trickery to circumvent the content generation rules I could make Santa box with Jesus. I think it could be possible.
I actually predicted "jesus" is no-go, but I hoped "boxing" would be okay. It wasn't.
However, I could try this "forbidden" prompt first and then try to generalize it a bit for the next prompt and it just might produce something close to what I want.
Definitely possible! But that would be technically circumventing the policy. DALL∙E does not only use a "block list" but also has a contextual understanding, but yeah, it's not perfect yet. Careful about getting too many warnings though, it might lead to a termination of your account.
I recommend you read the blog post about how the team handles bias for DALL∙E. Quite an interesting read.
Ah well but I think they'd like to hear about this "loophole", in case they haven't thought about it.
The restrictions are not applied to the previously rejected prompts even though they affect what future prompts generate.
I don't think it is a big deal because it is quite difficult to create "forbidden" images even knowing this.
However, they'd may want to "forget" rejected prompts entirely.
And I'd really like a way to manually make the UI forget my previous prompts, they may interfere with the generation sometimes.
link?
@verbal crag And the fact that you got "reindeer" warning signs suggests that us users are perhaps not separate. We could perhaps figure out a way to test it...
Feel free to test it. I'm curious if you could prove me wrong as I've already mentioned that your previous prompts does not and should not affect your new prompts.
is there a way to look at images i didnt save to collections
I'm not sure there's nothing to this, as I've noticed a couple of incidents of "style transfer" seemingly across prompts, but discounted it as coincidence and it never came up again. This was a while back. Why don't you document it all and post in a discussion thread like #ai-discussions, or if it's demonstrative, #bugs-and-issues ?
Through your browser history; your last 50 sets are also available on the right pane in a desktop browser, as long as you haven't cleared them.
thank you!
got this repeatedly: It looks like the uploaded image may violate our content policy.
Further policy violations may lead to an automatic suspension of your account.
Yet I've had no problems with a previous series of photos of a dog wearing sunglasses and a hat.
Any ideas?
Kangaroos also box. It's iconic.
It's detecting a photorealistic face, which is against content policy.
But its just a dog
😅 uploading your dog is fine, but to protect your account, or at least save the hassle of a manual review after losing service, stop teasing the not-so-smart AI with an image it doesn't like
👍 good advice, thanks.
Hi there.
Any chances that DALLE will accept payment from Ukraine visa\mastercard?
Well I just showed you how several of my previous prompts affected the latest one (boxing glowes, reindeer antlers, sumo wrestling attire) even though I specified none of them in the latest prompt. But getting (rein)deer warning signs by using more general "animal" could be due to moose/deer/reindeer warning signs being way more common than kangaroo/elephant/alligator warning signs. 🙂
I did note when I specified "reindeer" warning sign the background was yellow (because real reindeer warning signs have a yellow background) but when more general "animal" was specified the background was white.. which is perhaps more common warning sign background color globally.
Where can I find the rules for the AI?
Does anyone know why I can't upload photo from gallery to dall e? When i try I get a loading screen and nothing else
Can you try from website from PC? Works great for me.
Is it against policy to upload a drawing, or illustration of a celebrity, or does that not fall under the same rule as real photos?
I ll try when I get home thanks ❤️. I was wondering if it is normal for android to bug on that stage
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/
Or you can try VPN, maybe it's problem with connections to DALLE servers.
Thanks
Just curious: any official info on how many users have currently access to dalle?
None that is up to date. Based on our server numbers, it's rising fast, and the stated aim is one million users.
How to refer to someone famous? I've seen the Byonce examples like:"Photograph of huge inflatable model of Beyoncé, photograph from Macy’s Thanksgiving Day Parade"
but it doesn't work for less famous people. example: for a specific youtuber with more than 40 million followers.
any tips?
That is against the Content Policy I'm afraid.
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 does Dall-E handle content policy when generating images of characters? Example I saw some generations of Walter White from the show Breaking Bad. Since Walter is played by the actor Brian Cranston isn't that just a generation of Brian Cranston? Also how far does it go, of I tell it to draw Thor is it going to spit out an image of Chris Hemsworth? If it does would that be against content policy?
You can't generate prompts with real people, but you can with fictional characters. That's the current policy 🙂
Interesting, I guess it would be a case by case basis but it seems like this would be pretty easy to exploit. I guess OpenAI would have the rights to make the decision on whether or not somebody is exploiting this
They would review your prompt history if you got banned to determine what your intent was. That said, some of the policies are being continuously revised - this is beta and still very young as a product
Awesome thank you for the reply!
Hi there. I generated a (non offensive Pixar-esque) image containing a UK politician and have since re-digested the code of conduct and feel it is at odds with the image I have generated. I have removed the image from my gallery but the published link still works. How can I get that removed to ensure I am remaining compliant?
Technically all images remain available, even unpublished ones, because in cases of gross misuse OpenAI need a record. A single representation of a politician probably won't get you in too much hot water, as long as you've shown you're using the account for good and not evil since then... 😇
Much obliged, many thanks!
There were a few (non-published) sketches in the style of Quentin Blake (illustrator) too. But these were not published and were also since removed from my gallery by myself. I have learned my lesson and appreciated being able to digest code of conduct to a greater degree too!
you can absolutely reference living artists in your prompts to evoke style - don't worry about that!
yes indeed 🙂 thank you, the problem was that the sketches were of a UK politician!!! 😕 anyhow. I hope my recent generation history helps to redeem myself
how was DALL-E's satire? biting?
should not promote, will self-destruct in 3... 2...
everything quentin blake is rather impressive... quite challenging to mimic his irregular style, IMO
Is this normal? I've read the Policy and it doesn't match with my prompts..
"bomb"
he's illustrated a bunch of children's books, right? some pratchett and especially dahl?
Okay, gotcha
I was aiming to create something like this: https://youtu.be/ulCdoCfw-bY
To support Kurzgesagt and learn more about Brilliant, go to https://www.brilliant.org/nutshell and sign up for free. The first 688 people that go to that link will get 20% off the annual Premium subscription.
OUR CHANNELS
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
German Channel: https://kgs.link/youtubeDE
Spanish Channel: https://kgs.link/youtubeES
HOW C...
Describe what you want to see without the word "bomb". Bomb would give it totally different visual connotations anyway
Okay, will do
Thank you for the help!
good luck!
Been wanting to represent an astronaut Skeleton. But it always ends up placing the bones all over the suit, rather than a skull underneath the helmet. Wasn’t sure if there was a better approach there
skeleton (encased, trapped, stuck etc...) inside an astronaut suit?
Is it against TOS to write the name of a person to see what the ai generates for example Jimmy Fallon, Pewdiepie, Tom holland ect?
I have found you have to be artistic in your phrasing
I feel like more of a curator of an infinite landscape of potential images
Yes.
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/
See the Content Policy, particularly the point about public figures.
Ok thx for the help
I read the Content policy and am wondering if I can type my own name if I don't share it with a famous person or are names in general, a no go
Like for example "Karl Venice Litenhuag Olsen"
Make a channel of ideas for prompt, so that I can give you the ideas that come to mind, since I don't have credits for those who do have to use them, I'm not petty, the one that flows
When posting publicly or selling the work do we still need to keep the coloured squares Dall-e adds at the bottom right corner?
obviously still fully crediting Dall-e
I got a warning for a prompt for a fantasy character concept which had "with dead trees in the background"
I assume the word "dead" triggered the filter.
Is there a way I can avoid getting accidentally banned for prompts that follow the content policy, but that the filter mistakenly picks up as a violation?
How about if you want a face type — like if you want a Tom Holland type or a mads mikkelson type
You can remove them. 🙂
Thank you for the info 🙂
Is this something I could ask the team about to be sure?
I guess I can just email them, right?
Oh don’t mean to contradict anyone
Just want to find a way to do this project without breaking the rules
Hey so can I change my name in OpenAI account? (sorry if this question doesn’t belong here, i have no idea where should I ask)
To be clear, the generations would not break the rules and I would not represent the altered versions as original dall-e generations
@dry python I apologize the email you'll want to contact is dalle-policy@openai.com
Thank you @short salmon — always appreciate the help you guys provide here
I’m sure it’s a lot of work keeping up with all the folks pouring into this server
@hazy pagoda — hi! not sure what you meant by a filter. Can you explain more?
basically i have a giant excel full of descriptions that i insert as a prompt
i pass them into the OpenAI content filter to see if it warmfull, but i never know what is flagged
this is like a D&D description, obviously a lot will be flagged
"fight dual sword with +3 points blood damage"
there is any recipe to avoid this ?
Is there any chance the programm will ever take Paypal as payment xD Credit cards are always difficult for me as a german
Unknown atm
Isn't there a way to take an image, and say "turn elephant around" and the AI will generate that same image with the idea you've added?
Not for something like that, unfortunately
oh i found it
you upload an image, then erase the part u dont want, then type your aleration
You could try expanding it a bunch and using the edit brush to generate something like a set of animation cels
Muybridge has a lot of animal animations which might serve your purposes — not sure if this’ll help or not
thanks!
hello, I was wondering which part of this prompt violates the content policy: "a full shot photograph of a clown, poor lighting, terrifying, haunting, creepy"? I looked through the policy and I can't seem to figure it out
"a full shot" is getting flagged :)
oh, how would I imply that I want a full shot with regards to the camera distance?
Maybe try "a wide shot" instead?
it's been flagged too ("a wide shot photograph of a clown, poor lighting, terrifying, haunting, creepy")
Or put it at then end like this
photograph of a clown, full shot, poor lighting, terrifying, haunting, creepy
instead of saying "a full/wide shot"
I'm trying to get a distant/wide shot though, would that work?
Give it a try, if you don't have any credits to "waste" I can quickly do it for you :)
also, lets move this to #❌┃prompt-help
sure!
try specifying lens distance
ex. 15mm, 35mm, 50mm, etc
#top-tips-and-tricks message this generally doesn't work for me....
And I wouldn't expect it to based on how edit is supposed to work?!?
I just tried it for you, notice the erased picture beside the prompt, that's the only thing I did not erase.
The circled part are the only pixels I did not erase.
You're welcome!
when applying for access, there was a textbox to enter my twitter username in iirc, if my twitter username has changed since signing up, is there a way to change the twitter username associated with my account? (assuming it has any significance what-so-ever post approval that is)
additionally, I've noticed some users have distinct user pictures as opposed to just their first initial that can be seen alongside images they've shared on the dall-e website, but I cannot find a way to change my user picture on the dall-e website, where do I change my user picture?
- I'm not aware of any way to change the twitter username associated with your account, and you've already have access so I think that does not matter. I may be wrong though!
- If you've signed up with "Google sign-in", you may be able to change your user picture via Google's settings. Otherwise, currently, there's no way.
Hey guys, how do I check how much credit I have?
A generation I just tried has been blocked with a message that reads "It looks like this request may not follow our content policy. Further policy violations may lead to an automatic suspension of your account."
I believe this is a result of DALL.E taking the "cummings" part of the artist, (Brian Cummings) whose style I am trying to use as a prompt, and thinking I'm making porn.
Have I now received a strike? How can I appeal this?
click your profile at the top right of the page and look for the number next to the colourful DALLE.E square art thingy
gottem, thanks!
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.
Please refer to the above. Also, the #gpt-models have helpful answers 🙂
Sorry for not checking, I panicked a bit :)
Thank you!
No worries! I recommend reading the Content Policy too so you can try to avoid more warnings, found 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/
how much computing power does dall-e use, what infrastructure is it hosted on?
is it possible to change my public profile name and/or image?
is there an easy way to add the Ai mark, so that it is visible in a circle?
!watermark
Hi, I just got this message for trying the following prompts : "Racoons buying food in Japan, ukiyo-e" and "racoons buying groceries in Japan, ukiyo-e", any idea why ?
coon?
Racoon itself works, swap out ukiyo-e for the Japanese characters and keep it singular for the Racoon bit. :)
OK thanks !
Probably trying to block the word coon. Maine coon is also blocked. Try trash panda? Also please post that one in https://discord.com/channels/974519864045756446/979512526633046057
I thought people have been using ukiyo-e
Mh, I don't know for certain. I just did a few prompts with Racoon and that works just fine.
Maybe it's the "Racoons" bit?
Possible. This works with the art style https://labs.openai.com/s/ZYXPtXHDi3jHlly20JtO4Job
anyone know what might be causing the content violation error with this image? is it because he doesnt have a shirt?
Must be 'Racoons' bit then, anything else wouldn't make sense
Yup, this works though. https://labs.openai.com/s/VA4IES4GTXSW9uso4R6sbDCV
There we go! :) Nice
@fast bear see above
If it’s not blocking it for realistic face, it’s probably blocking the in bed part
oh it could be the bed, this is actually an uncropped version of the original and dalle already filled in the face and a large part of the bed, here's the original (spoilered cause some might find it scary)
and yea it didnt mention the face specifically, just that it might have violated the content policy
It’s beta and the filters will be constantly getting tuned
my account has just been deactivated due to unknown reasons, i kept receiving violation errors because my uploaded image didn't follow the content policy but i wasn't even using an uploaded image, i was using the DALL-E edit.
I also send out a mail to support, anyone know how long i will need to wait to get my account reactivated?
@broken wedge Thank you for your help !
I know they are flooded, and I don’t think anyone here can have an idea about the time frame, sadly
sadlife, well i hope i didn't lose all my images
Raccoons works fine btw
I sent an email for them to reactivate my account Aug 1st 1:53AM and they got back to me at 9:33 AM the same day, but there is probably a larger influx of appeals to deactivated accounts so most likely they'll take longer than a day to get back to you.
Any news on payment methods other than credit cards (e.g., PayPal)? I was very disappointed to find out I can't use my Dutch debit card, or my Ukrainian credit cards (because "We cannot currently accept payments from Ukraine"). Why is that by the way?
No new news on the topic, sorry 😦
Hey, so, dumb question. When I first got here I found a list of links and software useful for ediiting Dall-E Images, but now I can't find it again. Can anyone direct me?
sorry if this has already been asked, but are we allowed to buy prompts on promptbase? I found a prompt on there that looks cool but I don't want to buy it and get banned or anything
can someone tell me why verifying post burned my 1 credit?
anyway
how to uncrop images
Has anyone found an easy way to create a collage of the 4 photos + prompt? in a vertical shareable format? I mean a website or app
hello i am making a little nightmares inspired horror infinitive zoom
is this allowed by content policy?
I tried generating an image of a Cockapoo dog breed, but kept getting a content warning...
unfortunately its probably the "poo" part
or the "C" part, ha. still annoying
whoops, meant to put this in suggestions or there, ha
All prompts use a credit. The idea is that the part after the "|" is ignored, so you do your verification when you intended to use a credit anyway.
can anyone get a cartoon illustration to be made in Rubber Hose Style? ive tried rephrasing the prompt "angry monkey in a boxing ring, rubber hose style" several times. what am i missing?
Hi I'm new here and this question may sound eager. Is there any known way to call an API for DALLE-2 already so it can be integrated into other applications? If there is already documentation, I can't seem to find it
Why does MF DOOM on a swing holding a microphone, comic book style cause content policy violations while MF DOOM on a swing, comic book style does not? holding a microphone look like guns?
or maybe holding a mic looks like a p*nis
is there any way to go back on my search history or do i have to manually "save" each one i want to come back too?
Have a look at this :)
#gpt-models message
Technically, as long as you have the links, you can individually bookmark the link(s) and come back to them at any time.
There's no such thing as "group save" yet.
You could try asking in #❌┃prompt-help, maybe someone happens to be able to help! :)
well this is only if I publish right?
No need to publish, simply bookmark the link in your browser. :)
(picture for reference)
oh awesome!!
why is attack on Titan not aloud?
silly question, how do i see my DM's on discord
Hello, Is it currently possible to change the email associated with a dalle-2 bet access account?
I logged in with google and I need to decouple it from my personal google account in order to actually use
Any alternatives solutions to getting an account without it being coupled to my personal email are also appreciated.
The best way to do this would be by directly reaching out to the OpenAI support.
You can reach them either via. email -> support@openai.com
or the live chat at beta.openai.com
:)
Thanks
nw
Is there a plan to make DALL-E available via discord. (Query possible via discord instead of the website as long as you have credits)
The bot already knows which users we are so that should be possible.
Does dalle know how to spell??
When attempting to make it generate art of a phrase as 3D text it seems to miss-spell every time
it does not
Thanks
is there a list of banned words or phrases?
#things-that-should-exist should be shouldn't instead?
There is no list
thank you!
is creating horror content allowed?
Hi all, I just started generating images and I noticed that the site just says Dall-E, not Dall-E 2. Am I not using Dall-E 2?
Is gifting tokens to other people a thing?
Hello, I'm coming over from using mid journey. Is the prompt language the same? is there a good resource to learn the dall-e prompts?
If I use the "!watermark" for a dall-e generated profile picture is that considered as credit or should I still add "created by dall-e" in my custom status or about me?
where do I find this? cant find on website
browser search history
You are using DALL.E 2
Hello.
The word is not spelled correctly in the image.
For example
"wood" -> "wod", "woood", "dowo" in the image.
Sometimes only half of the word is shown on the left or right side of the image.
How can I get the words to be recognized correctly?
you are indeed using dalle 2!
just keep trying i fear, that's all you can do
Does that mean I have to try several times until it shows up correctly?
yup, you can also try editing an image, erasing what is spelled wrong in it and trying to run the prompt again on it
It would be great if you could give me specific steps.
Generate an image → manually edit the image (delete or crop the wrong parts) → upload the image → give instructions in text to achieve the ideal form.
Like this?
Find an image that you generated that you would like to manipulate in the app, press “edit“, erase what you would like to change (in this case, the lettering), generate the image with the same prompt, and done.
any guesses why this prompt triggered the content policy warning?: "Screenshot of a 1992 Classic Sierra Entertainment Point and Click Adventure Game on PC. Another entry in the King's Quest Franchise. VGA. Pixel Art. Retro. Scumm."
was it "scumm"? a copyright thing?
maybe "scum-" try without
yeah, it was that
Tried to do a commentary on a news headline to see what art it would generate - got content policy warning. I think art should be able to comment on current events. But I wonder if it was considered violating the political or the shocking category?
Prompt was: "Florida clergy lawsuits say abortion ban violates religious freedom"
I see!
thx!!!
could be violating the "deception"aspect of the policy? 🤔
i'd just stir away from that sort of stuff
I mean I’m sure “abortion” isn’t a very content filter friendly word in any case…
I think I got a warning cos I was trying to use an artist's style whose last name is "Cummings"
Doesn’t need to show an actual abortion - Shrugs. I hope it won’t become a Disney-like-filter
Great! Thanks! Wonder why it doesn't have the 2 anywhere on the site 🤷♂️
Yeah there are a lot of words like that.
Its similar to the iPhone branding model, they usually leave the actual number off a lot of the marketing copywriting
Yeah it's a little confusing like that
I just filled out an application to be a community volunteer. Approximately how long does it usually take to find out if your application is accepted?
I understand why you can't edit faces for privacy and deepfake reasons but can I edit a photo that has a composition I like in other places than the face even if the face is there?
And if no is it fine if I censor the face
Like this for example
are gender identities and sexualities banned terms? i noticed the term (dont judge me) femboy was banned, is that with good reason?
a lot of imagery tagged with that might be sexually suggestive! the exact reason why "furry" is banned unfortunately
trying "transgender" or "gay" would never trigger the system though
ok 👍 thanks for clarifying
🙂
How are people doing image zooms using the bot?
I've seen it mentioned a few times, but can't find the exact steps people are taking to make those incredible zoomed out images.
Basically just resizing/zooming out your images until you're satisfied. Hard part would be resizing it in a video editing software (E.g After Effects) and animate it like it's zooming out.
okay what does it mean when i erase part of the image i wanna replace with something else and it says "Describe the entire desired image, not just the erased area"? does that mean i got to describe the entire image again or can i just describe what i want to replace the erased area?
If I use the "!watermark" for a dall-e generated profile picture is that considered as credit or should I still add "created by dall-e" in my custom status or about me?
You do not have to credit it in any way and you can even remove the watermark completely. However do not deceive anyone into believing that the image was not created using AI.
Idk where else to ask this but where did the watermark come from? I imagine it's from the developers themselves but yk a man can dream that dall-e decided to let the world know ai is coming to take our jobs
Does anyone know why i can't generate "Maine Coon sitting on a pizza" but "Cat sitting on a pizza" works?
What do you mean by that?
Try “Mainecon cat” or “Maine con”
Is Coon a slur or something?
Sorry, im not a native speaker
I asked in 'prompt help' about being unable to create an image with a badger and a hyena - it only shows badgers. Is there a better channel to ask about that? Nobody replied there.
@luc#0002 re: furry - It seems happy to create anthropomorphic animals if you describe them as dancing or whatever: https://labs.openai.com/s/x9ZPBPsZsNHloGZKeYo7lGCR
Unfortunately coon is 100% a racial slur
So don't use it as a shortened version for "racoon" either
You learn something new everyday 😅
Exaaactly 🙂👍
I'll try to have a look at it later if I get my other stuff taken care of, unless anyone else helps you solve it! This kind of thing might need some advanced technique like inpainting due to the way the model might be confusing two non-human mammals as the same subject.
Maine cat works
Hey, Ive just got access and I was wondering if there was a screenshot feature like Dalle mini? where I could have a photo of all my results on one prompt?
I doubt it, I guess you must to do it manually
The design of the web is very clear and simple, so... I don't find any button about that x)
Ah that's fine then. I can do it manually anyways. Not too much of a hassle.
I thought about in-painting, but while that might work for two characters playing chess, if they're cuddling it's going to be tricky to delete only one - and if the prompt then says 'hyenas cuddling' it's going to put them away from the badger, I think! But thanks for the suggestion - I'll try that with one of my few remaining credits
Sadly the more discrete subjects you want, the harder it generally is. Good luck though!
Well, I have some awesome badger magician images, so it's a big win in any case! :) https://labs.openai.com/s/Twnyypnoatrtfy0ohnPKZLnZ
Actually, while you're around - is there a way to make the erase brush smaller when wiping out bits of the image? It's quite a fat brush.
Yeah there is, on desktop at least. However, there's a catch: the inpaint areas actually only work in 16 x 16 pixel blocks, because that's the scale of the original noise diffusion (64 x 64). So at a certain granularity I think it gets ignored.
Any way to make pixel art more detailed? Most of mine makes large poxels
Language prompting can help (e.g. 8-bit vs 16-bit, or give it dimensions) but it's never going to produce perfect pixel art sadly and the finer you get, the more wobbly. For an AI, DALL·E kinda sucks at perfect patterns 😅
Some results I've seen are pretty bloody good though.
hello i got recently accepted for dall-e, i invited some people in my team, can not they use dall-e also with my credits ?
If you can oversee their use of your account, fine, but it's your responsibility. You can't add additional users under your name.
Suggest they sign up themselves - many many users are being accepted with their own free credits, at a short turnaround!
So if i want my team use dall-e I should give my acc or they should sign up ?
They should sign up. You're the one who has agreed to the Content Policy and Terms - you are accountable personally.
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/
That kinda worked, thanks! https://labs.openai.com/s/TAZTjcCNEOZGqntbptTvMjfL
Based on
https://labs.openai.com/s/jcCxJuZFQrPNJv47MUetUHfM
Okay, all I want is that the people in my team can use dall-e with the credits I have, you suggested that they signup themselves for waitlist, but it took months for me to be accepted. So is it possible or not ? If possible how can i do that ?
Yo that turned out super well
New users are being accepted at a much higher rate since the launch of the beta a week or so ago.
Until they're accepted, they can use it under your supervision.
Thanks, I got it now
I can’t seem to be making things with brushed aluminium or carbon fibre… anyone knows what im doing wrong?
Thx!
Hello everyone, I am trying to figure out this content violation, is the the dnd part it considers IP ? I want to make sure I understand the rules.
Might be "gypsy"
However—according to the discourse of those who study the Romani people—the word, while sometimes positively embraced by Romani persons, is also sometimes rejected by other Romani persons as offensive due to it being tainted by its use as a racial slur and a pejorative connotation implying illegality and irregularity and some modern dictionaries either recommend avoiding use of the word gypsy entirely or give it a negative or warning label.
- Wikipedia
aha, yes, noticed it. I removed it and it worked. thank you!
I hope I dont get kicked out.
If your account gets automatically deactivated and it's later determined that you weren't being malicious with your prompts or intentionally trying to get around the filter, chances are you'll get your account back
aha. thanks. in this case. I am looking in to how good it can be on generating portraits of dnd characters.
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.
Hi! Is there a way to fix the hands without inpainting, in photos? Like another ai system, service, technique or so. I generate very good photos at this point, but the hands are often a problem. I get better results with inpainting erasing large areas around the hands, but it changes too much so it's difficult to respect the original photo. In alternative, something I can add to the prompt to make the generate photo better? Thank you for your help!
I uploaded this image to inpaint (it is literally image that was generated by DALLE of a Japanese RPG hero jumping against computer)
Further policy violations may lead to an automatic suspension of your account.```
Why?
You can often get better hands by describing the subject doing something with them, or even just mentioning them in the prompt explicitly - so it's drawing on its data for hands and not just ending them weirdly and calling it a day, because it's peripheral to requirements.
Weird. DALL-E is absolutely capable of creating images (especifically faces) realistic enough to trip the filter, but this seems like a glitchy failure.
Am I correct in the assumption that creating images depicting tentacle monsters in a sexual context would be violating the terms? Not that I want to, I was just wondering if the sexual aspect is limited to humans/humanoids or a general limit on using words that describe sexual acts
You are correct
https://labs.openai.com/policies/content-policy have a read through this :)
What a beautiful answer, it helps to understand how things work in general, too. Thank you very much!
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/
"dungeons and dragons art portrait of a cloaked women with fiery claws and a prosthetic leg" generated this stunning image. but adding revolver it did not like, so a weapon of any sort is not permitted I guess ?
indeed
got it! thanks! fair enough.
Hi there ! A lil question for you all.
IS it possible to change the resolution ? Like generating a picture in 16:9, or 2:3 or "wallpaper" ?
Not entirely through DALL-E, but if you look in #top-tips-and-tricks you'll find some handy guides on how to composite for wider images using the edit function. This is also a quick write-up I did specifically for 16:9:
Thanks 🙂
I can also wholeheartedly recommend you to take a look at these recourses!
#❌┃tips-n-tricks message
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y (starting page 55)
https://dalledictionary.notion.site/dalledictionary/DALL-E-DICTIONARY-2ec3dd6dc2184a45a9b6b4b8a021f0a6 (great for an overall overview) :)
Yes of course, that was the first thing i did actually 😄
Can I access Dall-e through an API? Are there docs specifically for that usecase?
#gpt-models message
take a look at this :)
got it.. thanks!
How can i choose a image that i want to !watermark
Upload the Image to Discord an do !watermark as subtitle
Ok!
Hey, i wanted to ask what are the rules/guidelines for modifying an image i generated by dall-e (as in, using it as a base), and also the rules/guidelines for using images generated as references for your own art 🙂
Or if its your existing profile that you wanna watermark, type !watermark in #dall-e-bot and it'll work!
You can do anything with a generated image (including removing the watermark) as long as you don't deceive anyone into thinking its was not created by AI.
What payment processor is OpenAI using for DALL-E credits?
So if i use it as a base reference and then modify if completely (so, used dalle as a sketch) i still need to say i used dalle as a sketch? 🙂
Due to the complex nature of the question I think its best if a volunteer answers this! @verbal crag @past geyser
That would be Stripe!
Thank you!
Do they state that somewhere?

I am not very sure if its documented but I heard Major mention that it is Stripe.
Alright, thanks :)
#openai-chatter message for reference :)

It's informally fine to remove the watermark even if you don't modify it.
As in, discussed here by OpenAI staff but a grey area in the policy.
Unfortunately I can't seem to add my debit card to buy credits. Worked fine on other Stripe payment systems.
No way to use paypal as payment ?
Stripe does accept debit cards, is there a specific error? If so please share it in a way that it does not share any personal information and I might be able to hep! 🙂
Not yet, no :)
While this is highly requested, Stripe is the only payment processor as of now!
I don't get a specific error, just that it had problems adding the card.
Is there any way to recover old results? The list of recent results on the right hand side seems to only have about 50 or so - I didn't realise they would disappear!
does it say anything other than that it failed to add the card?
You are so helpful Smile! ✨
How long can i use the beta?
nope. just says "oops! we had trouble adding this payment information."
One moment!
For what it's worth. I'm from the UK and using a Visa card (even though it is debit, but I can't see it not working either) so I can't see a problem there.
There's no 'set-date' for the beta to end, at least none that's known publicly (or to me). :)
Therefore, as long as you want rn!
I do know that the bank tends to want confirmation via app or SMS, so maybe that's where the payment system is tripping up.
I dunno though
Hey is there any way to change my dalle password? I may have accidentally posted an image with it visible somewhere...
hmm, unfortunately, I wont be able to assist you here! Since this is a problem with a payment method which involves sensitive information such as card details, I suggest you directly contact OpenAi support through support@openai.com and they will help you move forward.
🙌
Yikes! Did you use a sign in partner like google or Microsoft, or did you use the direct login method?
The direct login
does anyone know what this CAP thing is? I'm trying to buy some credits
I have never needed this information before for online payments with this card
I think they just called cap on the card lol jk
Not sure, I'm only familiar with CCV
yeah me too
it's silly, i admit