#community-help
1 messages ยท Page 203 of 1
Or if you catch someone here with a yellow-colored name online.. you might want to ask again? Up to you if you ping them. They're busy.
I don't work here
I will try, thank you
Joanne said they can't store them forever though probably. I calculated that my generations I made every day like others came to ~10GBs of storage, so if there is 100,000 users right now then that's 1 petabyte of storage which costs 25,000$
BTW I'm into saving anything I really like so no one can just delete it on me xD.
One problem you'll find is saving things often comes in compressed quality. Save an image? Yep it can be compressed, it can require you to get a new tool or go into the web developer tools in your browser and even create a new tool to grab the full original item!
The only time it is original is when you buy things stated as HD and original, like say from ShutterStock
youtube compresses all their videos also, so even if you tries to store something really important it is not the original quality, adn YT makes it hard it seems to find which version is the original HD OF-that, too.
My input on your questions:
I download only the ones I like (otherwise others would cost storage and be in the way). Some of those I save may be poor to actually document what the AI was like in 2022. But I stick to the level of what it does really -- I know only a few times it goes really haywire and I just skip those, and sometimes partially haywire and even those I skip often.
Yes of course people will browse their generations both for just for a single caption and for looking at all them without filtering search.
I mentioned it before but the prompt is too long often for the filename and should be under the 6 images (in the collage itself). You could have them in PDF but then you can't just browse thumbnails in a folder.
$25,000 or even 100x that is not much for a company that had a $1 billion investment. YouTube stores every video you want it to, it doesn't seem to me like storing every image ever generated is infeasible.
true
i'm feeling the new change ?? i'm getting asian/ black/ diverse/ unrequested gender outputs for humans now it seems
maybe......but joanne may know tho likely
I doubt all users are using every single prompt every day though.
maybeee ๐
Plus they're sponsored by Microsoft, so I suspect they get free unlimited cloud storage on Azure
And probably discount access to GPU clusters
exactly what he said.
What does "DALL-E" stand for? Is it an acronym for something? Or just a cute name? And if I am writing something about DALL-E, do I need to call it DALL-E 2?
DALLE is a play on WALL-E and Salvador Dali
I get that, just didn't know if it also stood for something as well...
Thank you! Do you work for OpenAI? I am looking for an official answer on this, I couldn't find anything online, so I think you are 100% right
I dont work for OpenAI but I do know that what I said is the truth, though it may not be the whole truth
i would be impressed if it also stands for something
yeah! You can even cite a new yorker article that came out a week ago: https://www.newyorker.com/magazine/2022/07/11/dall-e-make-me-another-picasso-please
Thank you! I appreciate it. I have a dog named Olly and I keep calling him DALL-E!!! I might have to change his name lol
Do Dall.e users get invites they can share with others, such as in Midjourney?
nope
@neat bridge are there any plans for higher resolution images in the future once DALLE goes paid?
Or maybe we could use 4 prompts for one image with double the height and width
no immediate plans unfortunately! but definitely trying to prioritize some fun features
Probably VRAM limitations I just remembered idk how this fully works
Is it possible to have masked inpaintings? For example, paint an area of an image where you don't want modification when doing variations
when modifying a image you can choose which parts are edited and which are not
Yes but what if you can choose which parts you didnโt want to be affected. A use case would be modifying around text. Letโs say I want to create variations of the cover art on a book. In order to do so I would have to erase everything around the title to preserve the letters then redescribe the cover art. But I want to only see variations on the cover art without needing to tediously recreate the subjects. Unless I am getting it wrong here and this feature is not needed to do the modifications. That could also be possible ๐ Let me know if there is a way to do it with the current editing tool!
So you want to keep the title text, but recreate the cover art?
I think you can do that by cropping the title and pasting it in a 1024x1024 transparent image (with only the title visible)
Upload to DALL-E and instruct!
Where is the link to the discord community? Two new users don't have it in their invite email
Invite only at the moment. They usually send out Discord invites together with the Product Update emails.
Yep! But they're restructuring the server at the moment, probably preparing for new invites!
Hi,
Are we allowed to use images we found on the internet to generate new images?
Yes
Given the precedence of the recent Cosmopolitan cover made with Dall-e I assume that using Dall-e for cover art seems to be within the permitted usage - or does this require special permission?
I see what you're saying, I wish this was a thing too
i guess you could do inpainting like zigzags
and force it to recreate different parts of the original image
Legal question
Suppose I wanted to use DALL-E to make illustrations for poetry, and then publish that
Posted this in the chat channel accidentally so I'm posting this in here. Does anyone know the specificities of what is and isn't allowed? Tried putting in the word 'gay' and 'homosexual' and it said it wouldn't allow it but I see people posting Dalle 2 creations with weed and other "non G rated" things in them.
sorry dumb question, how much does dalle2 cost? i suddenly got an invoice from openai for like $600 bucks! i was not using gpt3 that much and it does not show that costs anywhere, any one else get that?
Is this permissible? How do the rights work?
DALLโE 2 is currently free.
Can you contact them for persmission?
weird, thx for the response.... well i gues they wanted more money out of me for something lol
wait wait wait. what? are we being charged???
beware of scams, just in case
verify source of email
๐คฆ i hope i didnt just get phished
lol
the odds are unlikely
uh oh
it is from mandrillapp not direct from openai
i emailed openai support... will find out if i have been hacked ๐ฆ
anyhoo, continue on... sorry to monopolize convo
now that the images per prompt is 4, is the rate limit still the same?
Hey all, I have a couple of questions about terms of use; is this the right place to get answers?
Do we lose access to the preview of DALL-E at some point?
Iโm wondering this as wel
Well
Why is it when I add an image it has a harsh edge in the iterations created?
you'll always have access as long as you're ToS-compliant!
we do expect to launch pricing soon, at which point it won't be 50 free images / 23.5 hours anymore ๐ข
oh so after the pricing is launched we lose free image access, right?
Will the commercial use be available at that point since we will be paying for it?
^ +1
working on it! we should have an update on this soon
Thank you! Iโm wanting to work with and am in talks with creating designs for clothing company and some of my generations would be beyond perfect from this amazing program.
Will there be any guidelines around embedding art, or is that covered under the existing terms of use? (I didn't see anything about this.) I'm a game developer, and I'm wondering about incorporating the results in free games (seems like commercial will be an option once commercial licensing gets finalized)
This might be a stupid question... but do you work for OpenAI? ๐
bahahaha wouldn't it be hilarious if I were some rando (but yes I do!)
๐ฎ that's very cool!
Has there been any mention of extra commands ala mid journey to request specific aspect ratios...etc
So, when pricing is launched, we would have to pay for it in order to keep using Dall-E or would we keep the access? Or would it be you can keep using it but no commerical use unless you're paying for it?
Many of my prompts get flagged for being against guidelines, but Iโm not trying to generate something toxic or hurtful. How many of these will get me banned?
will the rate limit still be in place on the paid setup? or can I generate images until my wallet runs dry? ๐
Btw is using the DALL-E generations as discord or other profile pictures allowed?
yes! check out #dall-e-bot if youโd like to add a watermark for funsies, but itโs not required
no rate limit in the paid world!
I'm in. sign me up! ๐
That's very cool! yay!
Hey guys! I'm new here - does anyone know when dalle will be taking applications or allowing for people to do commercial use with it / build products?? Curious on a timeline or where to find that info
Are you guys manually reviewing flags before removing access for violation to the tos? ๐ couple days ago I prompted "girl with a black hole instead of a stomach standing in front of a refrigerator" and had a typo that made it "standing in frot of a refrigerator" which gave me a warning for entering this. didn't even know what "to frot" means until then but now I'm kinda nervous that I'm flagged or so.
well now i know a new slang term...
hopefully so... and I hope the bans aren't actually automated
sorry for the beginner question, but are DALL-E and DALL-E 2 separate apps? or is https://labs.openai.com/ DALL-E 2?
AFAIK multiple warnings in a consecutive order can get your account automatically flagged, but it is then left for manual review to evaluate whether you were deliberately malicious in your intents or it was an accidental flag
Donโt worry too much about flags if you donโt have bad intentions, but also lookout for words that may be perceived as harmful or NSFW by the AI ๐
fair enough. ๐
I see a lot of images with things like yellow minions, spongebob, jesus, etc... are these actually allowed?
guess as long as it's fictional characters it'd be fine. you should not attempt to generate people of political status or I guess public figures too.
maybe take care with religious figures doing questionable stuff too
well, for example, one of my first prompts was "a teddy bear meeting God" ... and while one of the results was amazing and cute, I immediately got worried I violated some kind of rule about religious prompts... couldn't find one, but still
(I guess that prompt outed some biases also, since 4 of the 6 had crosses)
this should be fine. technically you could come up with something like jesus doing drugs or so which could be offending to certain people but your prompt I don't think it's a problem
I think drugs are explicitly disallowed
#misc-images message <-- the good one from that prompt
true, shouldn't be possible
90% of my prompts involve teddy bears... so hopefully anyone reviewing my prompts and sees something that shouldnt have come out will be like, "oh, this fool is harmless" ๐
I've been playing with some fictional characters and while Dall-e 2 has a strong idea of what those characters look like it seems to deliberately avoid rendering them. Is that deliberate? Midjourney and Craiyon are both happy to render likenesses of fictional characters
my worst example is robocop, which gives this style which appears to be derived entirely from a single stock photo artist (https://www.istockphoto.com/portfolio/julos?assettype=image&mediatype=photography&phrase=robot):
lol thats great
have you tried specifying that you mean "robocop movie (1987)" assuming you actually mean that
yeah, tried that
it gets more generic and shows different robots with visors but I can't tell if this is deliberate or an accident of this stock photo artist (which doesn't mention robocop in its tags)
"a detailed painting of Thor destroying a chess board with Mjolnir, colorful lightning from the hammer" ... thought about it, but Thor is from mythology and not just Marvel movies... so I was like, this is probably ok
whereas craiyon
it might have to do with RoboCop being a figure that often has a gun in his hand. if a lot of this material gets filtered out from the training data it could lead to such a behaviour in the trained model maybe
ah that's an interesting thought, though I have tried things like "behind the scenes on set" etc with no joy
it does seem to be a general issue though, I can't get a good likeness of characters, but do get things that might be described in a similar way
it knows about minions and spongebob... so dunno
and pikachu
and r2d2, but not c3po
Hey guys! I'm new here - does anyone know when dalle will be taking applications or allowing for people to do commercial use with it / build products?? Curious on a timeline or where to find that info
Spider-Man drinks a coffee at a bar after a long day of work. Behind the scenes photo from the set of Spider-Man Homecoming.
that works decently too
Almost all "Characters" i've tried dont get the character right... it feels like they replace the character name with a "short description" of it before feeding it into the bot. Homer Simpson works though
am I required to have the Dalle logo on the image?
Im not trying to fool anyone.. I just asking if I need to have it when sharing
context: I generated profile picture for a friend, and the logo gets cropped out
yeah, it feels like they have some sort of feature that tries to redirect strong responses into adjacent concepts. I don't believe it's a training issue because it definitely seems to know what these characters are. baffling
As per Joanne: It's not required(?) ๐ #community-help message
I got the same when prompting for Gabe Newel, got a fat bearded man that looks like him
oh yeah someone got the same for Richard Stallman. Maybe it's an attempt to prevent personal targetting
this might actually be a safety restriction in the model.
Do not attempt to create images of public figures (including celebrities)
from the content-policy. I'd say Gabe Newell falls under that.
for real people it says the bot was basically given face blindness to prevent being able to do deepfakes
on one of their github docs
the bot or the model?
ahh, that's really interesting, thanks
so maybe robocop and c3po are victims of that
there's also this for copyright stuff, where they dont say what they're doing but I would have to guess it plays a role in why it cant do characters well
Hi, sry I already went through the content policy and the tos, but I couldn't find an answer for the following questions:
-
Can I publish the images on a webpage that I host on github-pages or my instagram profile, as long as I stick to the content and sharing & publication policy.
-
Can I or a friend of mine use a generated picture as a profile picture on social media or within a messenger?
I also wrote an email to the support, but I didn't received an answerer yet, so I thought maybe some one here could clarify my questions:)
For social media, the answer is yes (see: #community-help message); for GitHub-Pages, I think the answer is no; the terms of use explicitly prohibit (to the meaning of) having the images hosted on a web server
Discord is the easiest way to communicate over voice, video, and text. Chat, hang out, and stay close with your friends and communities.
Are we allowed to use illustrations generated by DALL-E in literature? The terms of use say no to any commercial purpose but the DALL-E team published a blog of someone using it for their magazine cover. ๐ค
That was an insider marketing collaboration
what if I make dall-e images into animated wallpaper
and released them in wallpaper engine
you need money to buy wallpaper engine software, once you buy, you get a lot free download wallpaper made by community
Hey i am new to this discord and dalle is there any guide to get outputs the best in different styles
even though there's no official API yet....... is programmatic access disallowed?
https://pitch.com/v/DALL-E-prompt-book-v1-tmd33y was posted here before
This sounds like commercial use, and wouldn't be allowed under their current terms of service
@rugged wing thx :)
Making prompts using only math is the real chad way ๐
No problem. You can find more on #top-tips-and-tricks
One character that it doesnโt have trouble rendering is SpongeBob for some reason. Maybe he isnโt humanoid or anthropomorphic enough.
is there a list of banned words to avoid being flagged?
I'm pretty sure OpenAI doesn't release a list because then people would just bypass it by finding synonyms that aren't on it
I really do wish there was some sort of "test prompt" area where you could try out a prompt to see if it is alright or not
accidentally got a warning and I am not sure how lenient they are
yes
wait, there is?
not official
but there are lists yes
dall-e discords have made their own lists, compiled, everyone chipped in
ah, that makes sense
I got flagged for "helicopter view of illegal zamboni drag race" because of the word "illegal" ๐คฆโโ๏ธ
should've known lol
yeah I think that was the first prompt I made
well for instance u cant use Nuclear but can use Atomic
real, "D'oh!" moment
everything that has to do with weapons, bombs, filtered
so riots, shooting, counts too
ehm specific type of underwear is checked
dont ask me why
_>
bahaha
celerity names
most likely
war/politic related stuff, pretty sure
and all the obvious others
its not rly hard to figure out what words would be rejected, common sense ๐
i sh*t u not
lol
its funny that the entry has no 'author' either
like someone just happen to figure that out
and we just wend, yeah sure...
John Doe was ur name? cool!
Hi
I have a question re usage in relation to the Content Policy and Terms, could someone from the team DM me, then the answer could be shared in the FAQ..? ๐
Does anyone know how to make my saved images collection public? Like change it from private to public?
You cannot do that with your entire collection yet. Only with singular images ๐
okay thanks!
I've been thinking about this also. Perhaps "tagging" images? Allow us to add a custom tag, and then view saved images with that tag?
I also keep "interesting" images, and I also will have periods where I'm focusing on getting a particular image for a specific thing, and would group those too.
Point being, yes, something like that would be super useful.
Very dumb question but everytime I log into Dall-E, I am greeted with my first test image I uploaded to play around with.
I'd like to... not see it anymore, and just have a blank slate each time I log in. But cannot figure out how to do this. (I've tried resetting the session, logging in and out, etc... no luck) Does anyone know?
Huh, odd. I don't have that issue. Have you accidentally bookmarked the link to the saved image?
Try to delete your browser cache?
Hi, 3 questions... is there a way to mass download images, like with a third party plug in? also is there a way to see images that have fallen out of recents? finally, is there a guide to the dall-e 2 interface?
-
Yes and no.
Officially: no
There have been people mentioning such plugins, but I canโt help you with that. -
Yes, go to your browser history and just open the links of your previous generations - they are still active
-
There are many guides in #โโtips-n-tricks, check pinned, as well as #top-tips-and-tricks
You can also check the dall-e dictionary, as well as other books and guides pinned in #โโtips-n-tricks
This - knock on wood - seems to have been the case. Not sure how that wound up happening, but thank you!!!
i see a LOT less images in the recents column today compared to last week. I thought they were going to increase the number. am i missing something?
Decreased to 4
I made a userscript to do this, gives you a button to save the 6. It can't download everything you've made at once though.
when does images generated decrease to 4? (mine still do 6)
@neat bridge does DALLE collect any data with the saved/published images, such as consider those to be the ones people like? because I sometimes save my most cursed images which I can see messing that up
I've updated my userscript to download the 4 or 6 images, it also generates better text in the clipboard image https://greasyfork.org/en/scripts/446973-dall-e-2-copy-button-userscript
Copy the images to clipboard,
a friend wanted me to generate "a tanker truck fitted for the apocalypse" but it didnt work, so i tried it again without "tanker" cuz i thought maybe it thought i meant "tank", but it still violated the policy, apparently the violated word was "fitted", any idea why that word would be against the policy?
I can confirm this happened to me too twice now
use a different word then? "adapted for" "modified for"?
This doesnt violate TOS as far as I can tell, is there anyone that could shed some light on this?
replace nuclear with atomic
Is that a known thing? I'm surprised it doesnt say that anywhere
Is there any way to see how many generations you have left before you gotta wait?
they don't want stuff that can be associated with war
Yes, it comes up often and relates to OpenAI's reticence to allow the representation of weapons
@neat bridge but when the new feature diversifies the humans in a prompt, how does it know which items are humans and where to add the words? Say my prompt is:
"and he is eating kibbles now, but is not a dog but a you know what. And, the parrot was teaching the class using a bob toy."
is the bob toy a human? is the thing eating kibbles a human even if it was said to be a real dog? And where are the words injected?
we fine-tuned a GPT-3 model to try and detect whether a human is involved! although itโs not perfect, Iโm pretty sure itโll know that a bob toy isnโt a human.
which element of this prompt triggered the content policy?
@neat bridge can you help me get my cofounder access to DallE?
Hello! I would like to ask a question, I have seen that in the Dalle rules, it is not to sell, but in the webinar call, we were told that we could sell prints, do album covers, book covers, book illustrations, etc. As long as credits are given accordingly, My question is the following, I would like to make an artistic reference book and use some images of Dalle, That would be fine?
Is there a way to restyle a whole uploaded image towards a prompt? (presumably, not inpainted masking - though maybe cutting half the pixels randomly could do a lot)
I believe that function is planned. I remember reading it somewhere here in the server whereas you upload an image and only change the style of it with a prompt. Don't quote me on that though, I can't find the chat log in the server.
"Moscow Mule Cocktail" which one is the one triggering the content policy? Moscow? Cocktail? too scared to try a different way :S
It's "cocktail", confirmed
I generated some cocktail photos just yesterday! ๐ฎ
๐
Generated yesterday for the #daily-theme challenge
can you try "Moscow Mule Cocktail" ? careful as it may trigger an alert
just in case you are definitely whitelisted
I doubt I am whitelisted in any way
yep, better don't try, dont know why i said that
Or maybe I amโฆ
Instantly generated for me @pearl dew ๐ค
Mystery solved! ๐
and i'm thirsty now
I was trying to do a prompt for some aviation art with "view from the cockpit" in it, and it wouldn't let me and I was confused... eventually realized that's a naughty word and settled on "view from the pilot's seat"... which seemed to work ok
I feel like whatever is doing the word filtering should be a little smarter heh
also tried some landscape portraits and specified "hardscaped" in one and got rejected... "hard" is a naughty word ๐
@neat bridge Is there a way to implement a solution where the offending word is the whole word? Then you don't have to worry about "cockpit" or "hardscrabble" or ... I dunno, "weinermobile" (you know, the Oscar Meyer car...), but it does mean your "naughty" list will have to be expanded for compound words
Can i upload photos of myself for editing? Notice it blocks it and doesn't allow for real people images.
I wish i could transform myself in a an aniamtion keeping the proportions of my headand body
or illustration
DALLโE is currently not allowing uploads for realistic faces even if it's your own self. Refer to the content policy for more info here:
https://labs.openai.com/policies/content-policy
What's preferred here, posting the openai link to the generation, or posting the image. The former seems better since the prompt is included by default, but I mostly see people posting just the images
Is there actually a list of known banned words anywhere, was annoying when "screwing in a lightbulb" failed
this server has one
#shameless-promo message
please consider putting seconds before the same user sends messages
Hey! Which channel do you have in mind?
the general channel is that sometimes they send short messages which should be in a textual line, I hope I make myself understood
Really wish it wasnโt reduced to 4 generations per prompt, seems like too little
Is an external-facing API likely to come at the same time as charged prompts/the public release? Or sometime after?
The word "cockpit" worked for me the other day doing alien stuff.
interesting!
I'm sure this has been discussed before (tried searching and couldn't quite find what I'm asking) - but what is the deal on inadvertently writing banned words? Is there a "three strikes and you're out"-type policy, is it more about the frequency of offending words that occur and then you get banned?
I understand Dall-E's restrictions and applaud them, but from time to time inadvertently bumping into this and panicking
@neat bridge sorry if this has been shown before
i asked about this yesterday, and unless you keep pressing to make something after it tells you no, it shouldn't be a problem as long as you didn't intend to violate the rules
as far as I've been told here
Yeah sometimes you have no idea what the offending word even is. A friend tried to make "chunky chili" in an image and "chunky" was verboten apparently
Do we know what is โnextโ other than pricing for DALLE? An earlier message from the team mentioned some โfun featuresโ that may be coming soon and Iโve been curious ever since!
AFAIK Outpainting / outcrop / expansion straight from the website & super easy to use ๐คฉ
Dalle is giving back just 4 images each batch. Only me or anyone else?
I think it interpreted the "pushing the car down the road" as dangerous or something
hey, I find some of the content polices not clear, should I ask it here? is there any chance a dev may reply?
I think they mentioned that they were going to reduce the number of images from 6 to 4
I'm a dev, and I couldn't resist the urge to inspect the page.. so..
I noticed I'm able to make the requests with any http client, would my account be at risk if I do that?
in fact, I noticed that the batch size can be passed on this request.. so.... I'm pretty sure messing with it is not a good idea.. but what if I Postman'd it?
Batch sizes can currently be modified to 6, but this will likely be disabled in the coming days. Any higher value just presents an error
I've already reported this exploit as well
not that I've tried ๐ผ
beta testers are meant to find and report bugs :)
good =P
anyway, I was wondering.. after the beta, i guess it wil ltrnasition to a paid system, just like gpt, anyone knows if the the restrictions to G rated content will be lifted?
I am not an OpenAI employee so I cannot say for sure, but having a technology like this (certainly the current leader in this space) going viral for R+ or 18+ rated content is likely not something they want to be known for. I would expect that they want to keep it as PG as possible for the foreseeable future.
I have no doubt that the open source community will come up with the solutions you may be looking for, for that sort of generation and content.
Plenty of content for the Cursed Generations channel otherwise
I was thinking more on the side of being able to generate most prompts without having to deal with false positives on the filters
like being able to use names of famous people, or maybe doing scenes with some degree of violence...
i don't think it is a good idea to ever let the system be used to generated sexual content, maybe terrorism.. self harm scenes.. and this kind of content that will attract a really bad attention
i will pay extra for the celebrity pack ๐ค
They can negotiate with OpenAI for royalties for the use of their likeness ๐
that would actually be a cool idea for further monetising the system
if you really want to prompt with a celebrity, you can pay for that, and the celebity in question will have already allowed it
but I don't think that will ever happen
Hello
hello รต/
First time on discord ๐
on DIscord at all or on this discord? or both? =P
Both haha
oh, noice
What goes on in this group, do we just ask questions?
Oh ok
please don't do this! we deactivate accounts that do this.
(and thanks for asking beforehand <3)
no problem =)
thx for the reply
Please don't hardcode the batch size in the client, someone could easily ask for 6 images again. It needs to be enforced on the server.
sorry i meant making network calls outside of our product
making network calls from inside the product can easily be changed to modify the request
Hey, I just join the community, Iโm glad to be here .
I was wondering how to extend a picture with dalle, I cannot see a button for this on the web app ?
Right now people are doing this by resizing the image with transparency and Re-uploading it. There isnโt any built in feature for this yet in the web app
Thereโs like a discord bot that resizes the images for you tho to help that process I think right? But I donโt know how to use it ๐
how can I always get this "from the feet up" angle? it took me several tries to get this composition -> https://labs.openai.com/s/mz8LiyvkI8KwD2luJ6MrS23m
Hey, sorry if this isn't the right place--I've emailed and DM'd but havent been able to hear back yet-- I have a gallery exhibition and sale in NYC coming up shortly, and I have a piece I "transformed" from a Dalle output, but have no way of knowing if it's transformative enough to be okay according to guidelines, since that's subjective--Is there someone I could talk to about it?
Anyone have any suggestions for image upscaling on CPU or moderate GPU (11GB of memory)? I'd like to upscale from 1024 to 2048.
In this task, we try to upsample the image and create a high-resolution image with help of a low-resolution image. The behavior of optimization-based super-resolution methods is principally driven by the choice of the objective function. Recent work has largely focused on minimizing the mean squared reconstruction error. The resulting estimates...
Thanks! Also found this: #โโtips-n-tricks message
Can I use an OpenAI image on my linkedin cover? How does things like this work?
@neat bridge could you please confirm whether this is the case?
when I try to upload an image I can't click to upload anymore nothing happens why anyone knows?
Apparently my page isn't even loading, only a blank white page
it workes on Safari
problem is about Chrome I guess
i mean.. just try the same prompt without it
sometimes the page gets slow and things take a while to load (sometime a very long while)
I see
Yeah, that seems to be the case
When I left my browser running for a good while, it loaded
Hello and greetings to everyone! I'm JF Neto (Netuno FX), visual artist from Brazil. I would like to propose this challenge what do you think?
Hey! Currently @lucid ravine is doing a rundown of the alphabet! We are currently on Day 3: C / Cars
Feel free to join all of us in #daily-theme
So for the credits, instead of having 50 generations per 24 hours, is it now per month or am I miss understanding something?
15 free credits every month
But you get 50 for the first month
@neat bridge Congrats on going beta! ๐ I wonder what you mean with 115-generation increments for $15. Does that mean 115 credits for $15?
Yes
Do I have commercial rights to images I generated previously?
thanks.
15 free credits per month is not enough. I would accept 15 per 24h.
i wonder why they're even giving us free monthly credits in the first place tbh
Another question, where do I check my credits?
Are we still limited to using 50 in one 24 hour period? I bought more and it still says wait x amount of hours
thx
Do you think it is free to generate images for them? They need some kind of income to support millions of users.
I think the $15 price tag is fair.
I'm amused I had to burn a credit to verify my Discord.
Same lol
Same here
Fwiw I donโt think the credit is actually burned? Iโm showing 150 still for today.
I see 149
I'm showing 149 and I haven't used DALL-E in a week or so
Darn
will we be able to connect to Dalle-2 via API anytime soon?
Is removing the dall-e 2 watermark against TOS? It seems odd to allow commercial use and then be stuck with the watermark.
I was about to ask that as well
Are we still limited to 50 requests per 24hour then even though we can buy as many credits as we want?
Is there a way to do bulk downloads of our generated images?
Can I get the credit back it cost to join this discord?
Yeah, I was willing and even eager to pay for dalle, but such a dramatic drop in the number of free credits + the high price + still no API access + content restrictions... gotta say I'm disappointed
I think $15 is reasonable.
I do too
Idk, $200 per month to do what we do now is steep
Good question
imo $15 is fine
I think it's too much if you're experimenting a lot
curious about the 15 refilling credits and the 50/100 bonus โย if we used only 15 this month, would we be back up to 150 next month?
this is more of a community effort now
$15 dollars for 115 generations is a great price
feel free to remove!
we'll look into this, thank you!
What's the best way to remove the watermark? Cropping?
Or will there be an option to do that within the app?
So it's still a square image
Just to confirm the limit should now be removed though?
thank you โญ
right now in the network feed you can actually pull the image without the watermark, did it by accident earlier
is 4 generations per prompt gonna stay the standard for this price?
I think right-clicking and selecting "save as" might be the way to go.
Is there a way to view all previous generations? I know it was recently bumped from 10 -> 50 by default but is it possible to paginate through ones history?
Can anyone answer if we still need to credit dall-e with the new terms and commercial use? I assume we dont need to credit, just as you wouldnt credit photoshop? @neat bridge ?
browser history if you have it
can we now sell the digital asset ?
yup you don't need to!
yes! no more limits
was unclear in the message
Can someone from the team provide clarity on the creation of NFTs?
yes -- dropping to 4 per image allowed us to add more prompts for $15
mint it! sell it!
yes!
Thanks @neat bridge ๐ฅ
Gonna set up my manifest contract and get to it like... tomorrow
My guess is no, and the 100 โthank youโ credits are one-time
Update - Requests has been lifted, thanks Joanne
is it allowed to offer the generation in general, not on a specific example? like showing past examples so people know what to expect, whats possible
wouldn't the "Disclose the role of AI." from the content-policy still be relevant though?
THE MARKET HAS CRASHED! THE MARKET HAS CRASHED! NFTS ARE IRRELEVANT! THEY'RE A SCAM! A PYRAMID SCHEME! A WAY TO GET A QUICK BUCK OUT OF UNSUSPECTING USERS! YOU ONLY OWN A RECEIPT TO A JPEG OF A DAMN MONKEY! THEY HARM THE ENVIRONMENT!
Reject Humanity. Return to Monke.
Ofc you have no argument, cryptobro.
I don't have the time to argue with someone who already made up his mind and screams at me in caps. ๐
we will edit that, sorry
you are making an nft winery
Somehow I'm not surprised it took just a few minutes for the digital Pokemon card enthusiasts to come out
I don't have the time to argue with someone who wants to turn NFTs into wine.
correct
Wait, just saw in the terms that even credits you purchase will expire... That seems like a somewhat arbitrary restriction...
hey joanne, do we own the copyright to images previously generated with dalle2 or only images after now?
Is there a way to invite or give additional users access to an account if you will be working with a team on prompts? or will they need to request access and wait to be validated to gain access?
You own the copyright to older images too
isnt that literally like the opposite of how a winery should work
Can the NFT talk go to another channel?
So this is saved just for questions about the announcement?
preferably #the-pits-of-hell
Yes, please!
Yes
OpenAI owns all Generations
In which way?
wine is a physical object 
is it possible for dalle to be a standalone downloadable product eventually? Would be awesome to pay once and then use my own compute power to generate images as needed rather than running the calculations through openai's servers
Again, another channel please @white ore
@neat bridge As a hobby artist with aphantasia this has been a godsend. I wish Iโd had more time to figure out why some generations are fantastic and some are just plain awful. I notice that as of the 4 per prompt the quality has taken a hit, with anything with a character having half its head cut off, etc. are you still tuning this? Is there a way to tell it to include the whole head or whole body? This just got a lot harder to test as I donโt want to waste too many trying to dial that back in again.
โ For now, please transfer NFT topics to #off-topic.
And NFTs are not.
OpenAI owns all Generations - the user just gets the rights to reproduce etc. but you can not sell the ownership of the generation since you don't own it, is that correct?
I think so, too.
Why is there a time limit on the 50 credits?
The time limit is an error and will be removed
how do you... know all this stuff?
Joanna said so earlier
openai rep said it earlier
oh okay
@neat bridge Is it possible to disable the feature of extra words seemingly being added to prompts for diversity?
Sorry didnโt follow all messages. Please DM @civic haven.
dont dm me bro
technically you can disable that by being very specific to the point where added diversity words would contradict the original prompt I assume
that was rather unsolicited beleevens
What network feed are you referring to? I also tried downloading the pictures (can't right click and "save as" on the images in Firefox either)
@tight falcon Sometimes my animal prompts randomly give results with irrelevant human pictures, which is a bit annoying
report then as 'not matching the prompt'
... I did not know I could do that. Thanks!
There are resources over at #faq-and-resources-old to help you with. Have fun!
!help invite
Shows you some invite links
Are there plans to generate images by api?
in the future, yes #faq-and-resources-old message
@neat bridge OpenAI owns all Generations - the user just gets the rights to reproduce etc. but I can not sell the ownership of a generation since I don't own it, is that correct?
Thank you
Thanks for letting me participate in this! had a blast
Relax. You asked about it. Wonโt DM if youโre not interested.
That doesnโt address why all of a sudden the generations are being cropped way in. This changed on me recently. I know I can edit and submit the image back with empty space but that takes a lot of time and extra now more scarce credits.
How far off is allowed API access?
this ๐
I found it too roundabout
especially now credits are a thing
anyway to do this without losing quality?
Especially when Iโve seen variations really produce poor results. Wasnโt so bad when 1 out of 6 was good, but now thatโs going to be like playing the slots ๐
Going from 50 prompts a day to paying $.13 per prompt (especially now that it's only four images per prompt) will be hard to do! If there were 6-10 images per prompt again it would be an easier sell for sure. This one will be tough!
When will there be Dalle API?
Hi @neat bridge/anyone else from the OpenAI team, I manage a multi person organization for GPT-3, how do I add my team to DALLโขE?
I see the FAQ says: "How do DALLยทE credits work if I belong to a multi-person organization account?Both free and paid credits are shared within each org.Only the owners of an org account can buy credits for the org." but it doesn't say how to start sharing.
it's possible you're looking at thumbnails instead of the full image?
maybe\
does network always take some time to load the image?
I get this often
just read up on how it actually works https://developer.chrome.com/docs/devtools/network/, it only shows things that loaded on the page after it opened
ah ok, thankyou so much
Yeah im definitely getting WAAAAAY lower quality when I do that, you're getting the same quality for sure?
Ah, thanks! seems like the web developer tool panel. I'll do some digging
ii get 73kb compared to 1.4mb
Does the hitting the content filter consume credits?
Nope! More info here: https://help.openai.com/en/articles/6399305-how-dall-e-credits-work
You can also verify manually
do my free credits get used up before my paid credits if I own some of both? 
Unfortunately the only two options rn but I'll ask about improving it!
that doesnt really matter because there's no difference, but i'd like to know too ๐ค
there's a difference in how fast they expire
true, I didnt think about that
ok I tried and it will use up the free credits before it starts using up the paid ones
do unused credits roll over to the next month?
If its used in commercial work, do you need to include the watermark?
Nope!
Free credits donโt roll over to the next month, so they expire a month after they were granted โ but youโll get 15 new free credits.
As per Joanne: #community-help message
Wow...!
Should we stock up on credits now or will the price stay the same once Dall E is Full Access?
why not just pay to license instead of to generate? the price point is really high for a lot of working artists, especially when the AI doesn't usually spit out what you want it to within the first few tries...
if this is the case i'll probably just be using midjourney...their pricing is a lot more reasonable. honestly very disappointed.
So it will be 50 for a first month instead of 50 for a day. It means I only have chance to put only 1 or 2 prompts per day?
Having a source of income would make dalle3 better
According to Joanne, this is first pricing model and they plan to explore more. #announcements message
with this pricing model... Will we get a way to upscale/edit images in browser?
sorry if I'm the 800th person to ask this, but am I now allowed to sell DALL-E images as NFTs & Prints?
yes
well slap my sweet spot.
Yes! (oops @fresh otter )
Dall-E is more powerful than MJ, but you can't upscale and the variation button often doesn't really hit the same as MJ. If we buy credits and run 50 generations a day, sticking to our original limit, we'll spend ~$200/mo when generating every day. Of course, there isn't any way to know how many you've used since there is no counter so it will be incredibly easy to blow through all of your credits in a day unless you're manually tallying how many generations you use
Hopefully they fix all this stuff in alpha
We're about to be in Beta
Bit past alpha
+1 @neat bridge Can you give us a ballpark on when the API is coming?
DALLยทE is officially in beta
As many others have probably said it would be nice to have a subscription model instead of just credits. Or if credits were only spent getting higher quality versions instead of just generating in general.
I don't know how many times I'd ask it to generate something only to be completely let down and have to rephrase and restruction the input. Spending credits for all those failed attempts just really makes me not want to use it.
Not really a question per se, but hopefully something you'll consider as more and more people bring up the same pain points. Thanks. ๐
With the new policy do we still need to have the watermark in the corner?
The current price plan doesn't really feel like a fully flushed-out one. It feels like one put together just to accommodate for the mass requests of people wanting to pay. They've been generous with listening to feedback so I'm sure we'll see a better plan
No you don't
You can remove it
my bad. i mean the final product.
Final product probably will be! But we're far from that atm
sub model would be great. or a cheaper model for ppl making less than 100k per year
Another request here for API! My creative tools would love it ๐ฆ
Donโt know if this has been asked yet but, are there any plans for some sort of reduced student pricing?
I feel like you guys should be more clear about the details here. the terms state that ownership of the generations remains to stay at openai and NFT bros often imply that buying an NFT also comes with ownership. Correct me if I'm wrong but this wouldn't be the case, would it?
Good thing the NFT scamstorm is almost over
Hi! I'm not an OpenAI employee so I can't speak on behalf of OpenAI. I'm sure they'll release a much clearer terms & policy as time progresses. But I'll pass this on to them!
thank you. I'm concerned about this because it's a very nuanced matter and a lot of scamming is made because things like these are not crystal clear to many.
I don't want a poor soul to dish out money for an NFT thinking he then owns the rights to the picture and starts printing these to sell just to have the original guy come up and (rightfully) claim that he still has the rights to commercialize this generation.
Can you make it so the downloaded file lacks the watermark by default?
NFTs are also the reason actual hard working artists are barely getting appreciated so this is sad but OpenAI's a neutral project so we cant really blame OpenAI.
this is a massive misunderstanding of how copyright works with NFTs.
If you own the rights to the image, you COULD include it with the sale of the NFT. But that is not implicit in the sale of the NFT, and in fact must be explicitly be made a part of the deal.
If you want any easy way to remove it, the unwatermarked images are stored in the openailabsprodscus.blob.core.windows.net folder under sources
exactly. and since dall e 2 owners don't own the generation I don't think we're free to even include that
Most to none of the NFT actually promise you to "own the art" as in copyright. Most just sell you the "NFT" a digital block on the blockchain, where not even the picture is stored, only a link to the picture which is stored an a classic webserver.
The unwatermarked images are stored in the openailabsprodscus.blob.core.windows.net folder under sources, you can download it from there for now
im an "actual" "hard working" "artist" and I also make NFTs. As an artist you have nobody to blame for your success and failures but yourself
I actually just joined the server to try and get similar clarity on derivative works rights. Can we use the generative pieces as the starting point for derivative works? i.e. get something that's great except for the nonsense text, and alter text into the image, and head off to commercialize it however.
I didnโt get an answer. Sorry for my English but will be better to tell us how many prompts do we have for day and if I get 4 imรกgenes and donโt save or download does it count? We need more detail in this change
can i offer image creation on fiver? please let me know
not sure what you're referring to
i do know some nft artists i see are honest with their works, but not what im talking about. i was specifically referring to the fact that most NFT sellers are usually ones who just sell generated content that are lazy and barely having any effort
people are being promised if they hold on their generated image for a while, they'll get rich
he's referring to people aiming to attach good looking images to a digital asset that has no inherent value because it sells better that way. and that more and more people getting educated about this so this scheme is about to die out over time
good for you if you can get your gains from your NFTs, but personally people can just come to you and commission you, aka you can directly getting money without having to go thru the blockchains, and the client can actually own the art, not the receipt that says they own the art
it's a win win, i dont get why people went on NFTs instead - oh yeah, it's the big numbers they were promised
This is exactly it. There are plenty of times where some prompts just generally don't deliver great or even good results, whereas other prompts give excellent results for all four images generated per prompt. Because image generation is so experimental - even moreso now that there are only 4 images generated per prompt instead of 6 - a subscription model makes way more sense to me, at least until DALL-E is able to make consistent bangers regardless of the prompt (which, I know, is vaguely defined and possibly very wishful thinking - which is why a subscription based model seems better no matter how you look at it).
Second that. Yes, there are a few honest ones who just want to get a fair price for their art and work, but those are far outnumbered by the massive systems of scammers and opportunists who just buy "NFT" art of low-paid crowdsourced labour platforms (or even just rip them from the original artists), hype the product via Massive PR social media, and make a quick exit - leaving everyone else in the dust. (as with #Crypto in general)
coding takes effort, so hopefully i can buy credits for dalle soon so i can use it for future projects lmao
Just popping in to say pay-per-prompt is the worst possible business model you could have chosen.
for real ๐งโโ๏ธ
i think pay per prompt's nice, at least i can use them in the future rather than full on subscription imo
i dont want to be charged per month bc i forgot to use it
so how are you planning to expand to 1 million users with these major price hikes? are you gonna count the like 95% non-paying as well?
Take note:
"most" and "usually" are where you lose me. there is a gigantic community of "real" artist across multiple blockchains.
ahhh missed that, thanks!
it has inherent value to a buyer/collector on a blockchain. just like art made from recyclables might have inherent value to an environmentalist
what's the inherent value? Why do I want to own a particular part of the blockchain?
Mostly, who's really winning from it, no matter where the "traffic" comes from, are the platforms (OpenSea, etc). They don't really care if it's the x-st sNFT (as there are millions of Shitcoins in Crypto just to pump and dump)
we also have "real" artists outside the blockchains yknow
Ohhk so we cannot save up those 150 prompts
โ Hi all, kindly transfer all NFT talks at #off-topic
everything in the world is moving to blockchain...
Maybe the nft discussion should move to a different channel please
Yes, more info here: https://help.openai.com/en/articles/6399305-how-dall-e-credits-work
My DALLE2 stopped working and shows this error no matter what I prompt, " {'task_type': 'text2im', 'prompt': {'caption': 'A photograph of a weird bird man shrimp', 'batch_size': 6}} is not valid under any of the given schemas" . Any advice or thoughts?
refresh, batchsize should be 4 i think
Thank you, Yes it's an old session.
I would suggest an unlimited prompts pay tier. Midjourney has one for about $30 per month for non-corporate accounts. They also have a throttling system that would make better use of your compute.
And it won't be long before OpenAI won't be able to charge such a high amount of money as they announced today. Competition is fierce and progress is rapid.
Thanks, I'll muddle my way through the tool in Firefox instead of clogging the channel with questions.
So I got the email granting commercial useโฆ what about as nfts?
If I buy credits, request a generation and, as usual, the generation isn't what I requested, is the credit deducted?
Oh, I misread! For that, I'm afraid that the credit will be deducted (at least for now).
do we still have to credit dalle when we use it for commercial use?
also, are older images we generated are also subjected to the commercial use?
(search above for joanne s anwers) Joanne answered this earlier with not needed
Oh, yeah. Definitely gonna pay to ask for several versions of the same thing over and over again until DALL-E figures out what the hell I want AND literally pay for the privilege of doing so.
heh, jokes on you. it's not dall-e figuring out what you want over multiple tries, it's you who figures out how to best phrase it. 
Hey what's up
do the credits roll over to the next month? and do we lose the extra that we are given if they do not?
regarding the latest #announcements
I'm working on 3D image of the human brain I want to create a spinning model of human brain
I want the camera to turn 5 degrees each image for example
My wallet thinks you're very funny. I'm sure your mommy does too.
Any advice for this use case?
Does a prompt search counts as a one credit?
Learn blender, lots of free or inexpensive models out there
Yes
how to remove watermark?
โreceive 50 free credits during their first month of use, and 15 free credits every subsequent monthโ are these per day?
Per month
@neat bridge how to remove watermark? or download version without watermark
I guess per month
๐ฎ ๐ฆ
How o feel about Dall-e 2 updates and current pricing model
But I don't wanna learn blender I wanna optimize my work flow for dall-e ...you think this is incorrect?
It gives a lot of creative pressure and anxiety ๐
That mean you will only have 1 or two prompts per day first month and the. 1 every 2 days ๐ญ๐ญ๐ญ
Blender will work on any laptop? I wanted to use unreal engine 5 but my computer wasn't strong enough
Goodbye daily theme. I need to keep my credits
Word this credit thing is noooo joke
I think you;re asking a lot of dall-e. as it chops all my heads off good luck getting brains. You can use blender and then send to a render farm
Iv already generated great brains
I'll share a throw away example
Ok, have fun
It's not about fun ๐
I'm looking at some of the blender brain models my one made with dall-e is objectively better
Is this already in? Does today prompts are already credits? Where can I found how many credits do I have?
the number of credits is shown in your profile, or if you click your icon in the top right.
You guys have any good colabs ? That are guaranteed to work ?
Your profile will also break down your free and paid credits and when they will renew.
This pricing setup just completely destroyed my TikTok channel
Does anyone know if variant spellings make a big difference to a prompt? Do I need to Americanize my prmots for things like color vs colour?
Anyone know if the monthly free credits stay in your wallet if you don't use them?
@neat bridge how to remove watermark? or download version without watermark
there was something .net when download folder but to me that made no sense...
ah thanks
You can inspect the web page and download the webm directly without a watermark
the 100 free credits we got we have to use them this month to receive our next 15 free credits for the next month right ?
You can't, its 1/20th of the resolution
Do I still have to make my generations G-rated even though it's paid now?
edit not sure since they are indeed split as paid and not free, so i just need to be at 0 free credits to get 15 new free?
yes
Blech. Ok. Thanks
Just realizing it's really only 15 credits/month and frankly very turned off,.
its also still limited like before
yeah 50 * 365 * 6 => 15 * 12 * 4 ๐ฆ
+1
not even 1/day now
Does the credit system start now?
yup
Wtf
So that means, people can now come in dalle2 now? Or still have to wait to be accepted?
still asking this
Where can we sell our images?
Good thing I have 980 images saved
Went from 50 generations of 6 images a day to 50 generations of 4 images a month
Unfortunately it's an introductory 50 credits, and then only 15 renewable credits per month.
still have to get accepted
15*
Will the free 100 credits roll over to the next month?
couldn't find info about it on the faq
they are "paid" see https://labs.openai.com/account / and paid ones expire after 12 months according to https://help.openai.com/en/articles/6399305-how-dall-e-credits-work
ok seems right
but its not clear to me if i only need to spend my free credits in order to receive 15 free or also my paid ones , would be nice to get some clarity on that
It sounds like each month you get 15 free, if you use them all, then you get another 15, of you only use 5, you get topped back up to 15 the next month.
I don't think the free credits are dependent on your paid credits
^ Yes, exactly. And you would use up your free credits first before your paid credits get used.
alright , thanks guys ๐
Honestly, the new beta system is a bummer. I just got my invite last week, and having limitation from 50/day to 50/month (and 15/month afterwards) is, frankly, sucks. 50 / month doesn't even feel like a trial, that's just demo. People will be too scared to even experiment on the (currently still flawed) query prompt. This directly impacts on fewer user input (flags / favorite), fewer training data, and slower learning overall
For pricing model, I'll give it to them to experiment for, but, after the 4 images / prompt, and now largely reduced prompts, it's a bit too much. I hope they'll reconsider to revert the limitation
Can we build a custom interface for dalle-2 generations as it is now available for commercial use?
Can it now be integrated in a similar way as GPT-3 allowing us to build business use case on top of it?
I believe it's not possible yet rn, they don't open any API access atm (IIRC, CMIIW)
In the email they say that it is available for commerical use, but how would that be feasible without API access as we can't generate things in the playground and take it from there on scale
Trying to wrap my mind around it
I believe they even mentioned an org account, they might think the org will hire people to write the prompts, very clever /s
yeah im looking for API access
i have an openAI api key
and i am scripting an attempt to access the DALL-E 2 generate function via python
but it comes back with a scope error
so it looks like the publicly granted API keys are not allowing access right now
but it seems that some people had access before because there is a python package that used to work
if anybody knows how to get API access like that again, id love to hear about it
im looking to generate thousands of instances for an NFT project
Thereโs a really cute character design I generated, and Iโm wondering how I can go about using the design. Am I allowed to replicate the design (use it as inspiration for a new character that looks similar) in my own artworks and credit DALLE, or do I need to use the new credit system to be able to use/commercialize it even if I didnโt directly use the art DALLE made?
Interesting would love to know more
You used a credit generating it. No other credits needed. You can remove the watermark and use as is or create based off it. See the announcement for more info
the update is that it dont work now lol
im hoping that somehow some way maybe i could test the API out for the team? XD
The character was generated before the credit system was implemented
was wondering the same, if it is created before do we have to manually edit the image to remove it?
I've tried to dig through this discord since you asked, but sadly I can't give any definitive answer. I remember read about someone screenshotting a phrase (IDK where it came from tho) that OpenAI also grant the rights of all previously generated contents, even before the current credit system. But, as I have no source atm, take this with grain of salt (or try to search thru their announcements)
Does commercial rights include selling creations as NFTs? The "Ownership of Generations" section in the terms seems to say we have no copyright claim on generated art. If we have no copyright then how can we resell our generated art?
right now id doubt you could get banned attempting it because it doesnt let you generate using the API key
They say that attempting can get you banned.
And that they've already banned a lot of people for it. Not worth it imo!
it just returns a scope error
the thing that seems to be banned is botting
attempting to use the public website to do bulk processing
but if you use your API key for a Dall-e function call it just returns the error that you dont have the scope of permission to use that function
{
"error": {
"message": "Your key is missing a required scope: app.labs",
"type": "invalid_request_error",
"param": null,
"code": "missing_scope"
}
}
This is what Joanne said:
Please also do not attempt programmatic access of DALLยทE, which is against our ToS.
yeah thats botting
like attempting to use a script to fill in the prompts on the website repeatedly, thats banned
but if you have an API key, they either let you use it or not
right now it seems not
its up to staff to clarify but i doubt youd be banned for checking if your api key works
This is programmatic access
I don't think this is a good conversation to have here
You should email them
its a question of potato po-tah-to
if they allow API access then they allow programmatic access
Pretty disappointed in the beta credits system. Between the reduction of images per generation, and the massive downgrade in free credits from 1500/mo to 15, it has become really hard to experiment and explore. I don't think the beta needs to have as generous free usage as the alpha, but I feel 50 and 15 per month is too low. That being said I don't understand the resource constraints so maybe it's all that can be done when expanding to a wider user base
Yeah I don't work at OpenAI so I wouldn't have all the details! But posting libraries etc here is directly against their ToS so I think we should avoid those topics until there's more clarification.
where is it stated to be against ToS to post libraries here?
the python code only plugs into an API call provided by openAI
A frightened millions of users are as useless as having nobody use your services, bad move on their part
tbh them charging for the service just makes serious customers want to use it more
im just asking for them to raise throughput for devs
I'm pretty sure an API is on the roadmap
my point is not about the pricing, but the limit for the new users to experiment. Heck, I won't pay a dime if I have no idea if the result will satisfy my needs or not
Did you just get in today?
Within a week
The API is literally already out, as shown above lol
it used to work
they took away access to public API keys
but they say they will re-enable it
just posting to say i hope they re-enable it soon or would be nice enough to give me access to the API
thanks again
if its possible, DM me!
But the beta users will only have monthly 50 tries, compared to 1500 of close invitation users
i want to use about 24k instances for NFT sets
gross
Any reason why the variations only returns 3 new results? Feels kind of annoying to not even get 4 variations on the original now that we're paying for each roll.
I really, really, really wish I could pay to get more than four.
well you can just enter the same prompt multiple times but like you said, it cant be botted
if you enter the same prompt, it doesnt just give you the same thing again, it gives you results from a different random seed
Will there be a UI or account info where you can see how many uses remain?
How do we request access to the API? Is this generally available to hobbyists or only bigger companies/projects?
Hi, would you consider implementing a 'free edit' system on generations, (for refinement purposes).
Let's say that for every round of generations we make, we are allowed to make 5 free edits to a single image, if say only a certain percentage of that generation was edited (say less than 10% of the pixels).
Sometimes the images are almost perfect, but just need a little tweak here at there (eyes on faces for example). It would suck to have to keep paying the same amount only to get a failure of a refinement each time.
Will Dall-E allow more aspect ratios in the future?
top right, if you click on your account icon it'll tell you how many credits you have left. That's how many uses remain.
This does not work for me. Correction: The credit balance shows up if I am on a result page for a generated image, but it does not show up if I am on a blank page with a blank prompt prior to generating an image.
uhh I have a couple questions. so as you now own the rights to the images you generate would this mean I would be able to use one of the images Iโve generated previously (before this new update) as cover art? it would be on spotify, youtube etcโฆ as you have to credit openai, does this refer to keeping the logo in the bottom right?
Yeah, I was kinda looking forward to be able to pay for more access, but the new scheme seems like not great value for money. MidJourney, FWIW, gives you 15 hours of fast GPU time for $30 per month, and then additional hours for $5 with the option to switch to a slower mode for free. Thats about 900 images + free slower images. The same price with Dall-E gets you 230 images plus 15 free images. Thats definitely going to constrain many people's ability to explore and create.
are you by chance still on a tab that had the page loaded for a long time? refreshing doesn't fix it?
Good catch! Refresh fixed it. Thanks!
I heard some people obtained 5000 free credits, i only obtained 100, why is that difference?
huh? Everyone got 100
That person bought their 5000 credits
Oh sorry, i read fast and i misunderstood the message of that user ๐คฆ๐
750 dollars moment
I am ok with everything just dont lower the resolution , viceversa increase it and dont lower the brain machine just to let it in 1 million of people
very important!
hey, so, from now on, the metric for uses changed? instead of 50 uses, where each use returns after a day counting from the time it was used, it now has 15 free uses that reset all at once once a month?
Was including the watermark in images previously specified in the sharing policy? I've always believed it to be, but looking back on the policy prior to the latest announcement I can't find it explicitly mentioned. Basically is being able to remove the watermark a brand new thing?
tip: open the dev tools, check the network tab and get the image without the watermark, it is added in post byour browser....
or.... there could be a checkbox on the page to downlaod without the watermark..
feature suggestion ๐
no, you just have to be clear about the fact that it was made by AI
in fact, the bot here has a command to addin the watermark to images.. in case you want to bamboozle someone... i guess
damn.. it really is 15/month? well, I'll not complaining.. but it is a very big cut =/
I'll see if I get that artist subsidised access.. I was really planning into using dalle on my projects as a tool just like I do with copilot
@neat bridge Can you show me the prompts dalle creates using the new human diversity method if I were to feed it these:
"so he was sailing near a woman and the cat saw him sit down on a chair"
"it was something not like mother nature but only something I would not call real"
Is there a way to get DALL-E to not crop off part of a generation?
it doesn't crop images, but the training data was random 1024x1024 patches of existing images
so it just has a garbage sense of composition
if you tell it to make a wide shot or whatever that helps
Can you explain if images need to keep the watermark and say they were created using AI in commercial uses with the new content policy? In some cases in may be hard to keep watermark (aspect ratio) and not have an opportunity for attribution (video ad).
you can remove the watermark in photoshop with photoshop content aware fill
I have found "Zoomed out" to be helpful in that, it's super annoying
or right click on the image and inspect it, the image without the watermark is right there
I am not asking how to remove it, I am asking what their content policy is.
how are they going to enforce it? is an american company going to be looking for their AI generated content in a german tv advertisement?
besides, US copyright law says you own the copyright for AI generated media https://www.smithsonianmag.com/smart-news/us-copyright-office-rules-ai-art-cant-be-copyrighted-180979808/
further, open ai would need to submit every image to the us copyright office
Copyright is not the same as a content policy on a private platform
true
I am hoping for an official answer about the watermark as it is unclear from the current content policy vs the previous content policy
well some guy is selling them as NFTs without the watermark
well probably some guy is gonna get B&
they did explicitly say no NFTs, but they recently said you can commercialise them
maybe they want people to think they can make money off this if they buy credits
print them into physical world we can tounch
Right, they changed the content policy today and I am asking for clarification and you are complicating my very clear question.
easy as cropping
omg children, shhhh
๐ฌ
i pointed this out to them
you can also get 6 images per prompt if you edit the network request
they fkn know that
@jolly widget I'm aware of what you're asking
they need worldwide support staff clearly
yeah not sure how much the price per image will increase when they have to pay for a call centre to support a million paying customers
anyway, it's the hacker in me to seek out ways around things
Hi! Re: watermark
#community-help message
thank you!!! lotta information to take in. I appreciate it.
Indeed. You're welcome!
I considered putting my own procedural art up as an NFT, way too much work though
it's a bit hussle, in new pricing terms even more, but before the update you could have downloaded the photo, resizing the image but leaving 1024 ratio, then uploading it back to dalle2 and editing it, basically works like a content aware tool, detected the image and builds whatever you want on it, based on your prompt. i've written here about resizing, if you find it interesting... https://link.medium.com/vSH8e5qKPrb
Oh interesting solution! thank you for writing that article ๐
https://www.remove.bg/ works like a charm too
Not sure how it's going to work after yesterday's update tho๐ ๐ I don't think many people would like to "waste" credits on that, it would have been cool if we were able to do it there instead of using another third party
Guess this will be their way to limit content generation spam. Make it too expensive to spam ๐
Or u can use helper words or lenses, I'm sure they've suggested before me ๐ I generally use wide lenses, works fine
hi! I was wondering if I could use images i made previously to this version for commercial use now that we have rights for that. Or do commercial use rights only apply to prompts created from now on?
Nope, all your past present and future generations are fair game for commercial use! ๐
Hi, you have confirmed that we do not need to attribute openai or dalle for the images generated, right? Why is this not mentioned anywhere in the newly revised commercial usage policy? Kindly clarify and confirm regarding the attribution part.
the one with the watermark is a bit sharper quality however
You have to let everyone one know that Ai was used in someway, itโs in the policy,
Is it possible at the moment to provide access to members of our organization? I can provide access to OpenAI services to other members of the team but not sure about Dalle's tokens, I guess they should be part of the beta program as well.
You will have to ask them, but donโt share your Id with the organisation, if someone misuses it then they will block it, they said that in zoom interview,
Can anyone elaborate on commercial use, I mean can sell my artwork to someone
There will be A LOT of new Dalle users around! COOL! But can we have a channel for the DALLโขE 2 Core Crea Crew? So we don't get all the same questions a million times over and over again AND share our stuff and experiences there?
Is thereโs somehow a way to reclaim my credit if I didnโt like the results?. Lol
It would be good if the Terms of Use could clarify what copyright situation is if creating a derivative work from a DALL-E generation. Examples:
- Using the image composited within another image
- Editing an image to fix glitches, etc
- Creating a whole new image or 3D model based on a generation
Who would hold copyright on those works?
Thanks!
Not sharing my ID, i read this notification but it's unclear how to share these credits with them
Can I edit my NFT in DALLE and then sell it as NFT, can someone please confirm whatโs the stand on NFT, what are the things we can do in commercial license,
is there a way to upload a picture of a character/animal/person/etc. and then have DALL-E put them into different situations (eg. waiting for a bus, having a coffee, painting a house, chatting with a friend)? Or, similarly, have DALL-E create different poses like one might for a storyboard reference?
You can upload for example a character with a masked out background and prompt what you like to add.
or mask out a specific area and add something there- in your composition
thanks demosch. however I think my goal is to place a character *into * an environment, not just draw an environment around them if that makes sense
Yeah, I suppose I could do it that way, but was just curious if there was a way for DALL-E to do it. Sort of combining the automation/generation process with the compositing in one easy step
Do we get free searches to do some testing and posting here. Using the credits given to us for testing seems a little odd...isn't it..?
@tawny acorn Currently there is no way to create "a character" in DALLE and have that character appear in multiple images, like an asset in a game development system. Each prompt starts from a new random seed. This is a serious limitation for the product for applications like making illustrations for a children's book, for example, where you want the same character to appear in multiple situations.
Thanks for the concrete answer! Seems like it'd be a really handy feature for really making congruous narratives. Maybe one day it'll get there
Is it allowed to publish girls with white paint dribbing of their face?
Or is this considered sexual and not allowed?
It didn't violate anything in the creation process
Hi ๐ Does anybody know if there's a way to publicly share your DALLE collection?
I post a thousands of my images on https://www.buymeacoffee.com/aviewfromai/posts
I have achieved this by describing in detail the character, his clothes and his accessories.
I also got a unit clarifying that all the images are delivered to me with the same style (oil painting) and at sunset.
I see a lot of people mixing different art styles in their prompts. Whatโs happening when people give those prompts?
i.e. a dog fetching a frisbee, watercolor, acrylic painting, oil painting, oil pastel
mostly trying to communicate 'not shitty clipart' to the AI
is there any effective way to get DALL-E 2 to take a source image and kind of remake it as a different type of image? I've tried a few things with the inpainting to do this but, seems futile
I have seen some people applying transparent noise over a source image to kind of try to get it to regenerate the whole image while also incorporating underlying features of the original, very hacky but some interesting experimentation https://twitter.com/waxreplica/status/1548792293313978370
oooo... I hadn't tried noise
Much experimentation from the days of free dalle2 lol
does the teams thing work with dalle yet? said a team would share credits but I don't see a way to make a team
That's actually genius lol
something to keep in mind is that when doing an upload+edit, the original and the edit are sent to dalle... not sure exactly how it handles it if you pre-transparent an upload
I mean, it'll still send both, but they'll be the same so dunno
I didn't have as much luck with the noise like shown in that tweet, followed his same process, but I think it's all about getting the right balance, its perfect for giving a color scheme seed for DALLE to start with at least
well, knowing that it uploads both the edit and the original, I tried painting out the original entirely and asking for X in style of Y, and it doesn't work 90% of the time. occasionally will get something close
The bot could have a command for that...
@patent rain
Im serius, if that does not get added to the bot ill do it myself '_'
This is too good to not have it
I mean... The dalle interface could have the features of the bot... Just sain..
Tbh, dalle could even be usable via the bot
V interesting indeed
I have had good luck with "cyan tint" or "yellow tint" as a qualifier on a prompt if I am looking for a specific color palette.
Not seeing the 100 extra credits mentioned in the beta email. Anyone know who to ask re: getting those credits?
How many credits are you seeing in your bank @clever cradle ?
Has there been a confirmed change in the model? I saw a few complaints on Reddit about it being " worse " or " bland , didn't believe it but trying the same prompt before and after there's a noticeable difference for photo portraits.
It's not like you're hearing about the people who are getting perfectly decent generations and keeping them to themselves
People have confirmed that there are extra words being added to prompts for diversity, which as one might expect messes with the accuracy of some prompts
I've run some prompts from last month and they look just as good, personally I haven't noticed a difference
Yeah it might be the 4 prompts having less " good " images to choose from vs the old 6
48
dm'ed you!
yo, do initial 50 tokens expire on 1st of next month or one month after one gets them ? ๐
One month after they were received :)
you can see the expiration date here: https://labs.openai.com/account
can we sell them as NFT? or maybe sell it as a print.
Now that DALL-E is pay to use yes. You now have the rights to your generated images.
well then let see how much value does dalle nft has over other AI arton opensea
damn, there are people on opensea who were illegally selling dalle art 3 months ago, the policy was just implemented.
How much value? Apparently, this much: https://superrare.com/0xb932a70a57673d89f4acffbe830e8ed7f75fb9e0/pieces-36134
The artist's first minted work with DALL-E 2, a tool for revolutionizing humanity's access to creative expression. The work suggests a lifetime of singular, chaotic moments, all coalescing into a figure of coherence and purpose.
This piece includes a physical print, which will be signed and sent to the collector at their request.
This is really mind blowing but, not many are on superrare, I have applied 3 times and still no reply. But the possibilities are endless, these art can be printed on canvass and even T-shirt,
For example I will release on today, letโs see
i mean i wouldnt pay much attention to the value. lots of scammy shit going on with NFTs with people trading them between accounts to drive up the value
or in this case make fake offers
not saying thats what happened here ofc
@neat bridge I still havent received a response from the OpenAI support in regards to my question about GPT-3 and commercial publication.
The terms of sharing do not make it clear to me whether that is restricted to just fair use, or also commercial publication.
I know that Dall-E just updated its terms of use to allow commercial use, and I am aware that AI generatd novels have been sold on Amazon, but for all I know that could have been illegal.
I would much prefer an official answer from OpenAI on this before I proceed to use GPT-3 go help me write a novel that I intend to publish commercially and then get into legal trouble.
If you guys could star this so that someone from OpenAI sees this, it would be much appreciated!