#images-discussions
1 messages Ā· Page 16 of 1
That's the only other option for Dall-e 3 afaik. You can use Dall-e 2 though
this is cold
If anyone wants to create images with this theme in mind: : AI and Diversity. Address and celebrate the rich tapestry of cross-cultural diversity in society. Explore how AI can contribute to promoting inclusion, breaking down barriers, and fostering a more inclusive and diverse world. Would be great to see what you guys come up with
Can't generate anything right now due to high demand
Interesting - also, Dall-e has built in instructions to make prompts feature more varied genders and diverse ethnicities
oh aye? Did not know that, thanks
Cheers
I was wrong, just go to your browser history and search for 'bing'. You will see all of your prompt history listed and if you click on one you can return to see the images that were generated for that prompt.
thx
So is the bing dall-e 3 disabled for anyone else? The button has been grayed out for me for several hours
wew glad it's not just me!
Is ChatGPT's dall-e 3 still up for anyone?
Yes
How do you guys get such good images with this? I've just started using dall-e and even using the same descriptions...mine looks pretty bad and any writing is usually unreadable.
Where are you using DALL-E? On the labs website?
yeah
Currently the labs website uses DALL-E 2 which is much weaker than DALL-E 3. DALL-3 is either available via Bing Image Creator or via ChatGPT Plus. Keep in mind that DALL-E 3 is not fully rolled out to all plus users yet.
Are they rolling out randomly?
thanks for the info!
Yes.
Hi! I'm about to subscribe to the plus tier and I was wondering, is there a limit on how many images it can be generated? Or it will follow the same 50 messages/3 hours limit?
There might be some rate limits in place, though I don't think we have concrete information about it. Apart from that it follows the same 50/3h limit.
Great, thank you
It often generates fewer images for various reasons
is there anyone I can suck up to to get Dall E 3 access on ChatGpt 4? š
No, you just have to wait a bit š
Hello, I see people sharing they have Dalle3 in their chat gpt plus account. I don't see it yet in my plus account. Is there a way to get it now?
you need to wait for it to be rolled out
damn, message before was the same question xD
thanks
its not out yet
Currently some plus users have access to dall-e as it is rolling out slowly to users this month. Currently, those users can send 50 messages per 3 hours to gpt4. If they are sending messages to the Dall-e extension, the user can receive up to 4 images per prompt, but it could be less or none if there is a server issue or if the content is filtered.
how do we get access to 3?
some early testers have dalle-3 for stress testing the system, but it should come to everyone once it's more stable
Interestingly, I get this error when I send a complex request, but when I send a simple one line request It goes through
You might get more success if you use the bing chatbot
But you have to have the app/microsoft edge
I get the feeling these "High demand" errors are actually filtered outputs
They just hide it for some reason
Cause it shows up as "prompt blocked" in the Recent tab
Yeah chatgpt does the same thing.
That said, there is definitely very high demand right now for dall-e 3
You can get rate limited yourself, beyond what others are doing. Best to take a 30min break if that happens.
Is dall-e 3 on gptplus cost a extra cost?
nah
Very strange, I'll try again on my pc when I get home from work
ty
My credits pool is different on pc for some reason even though it uses the same account, idk how dall-e 3 works here
how is it with the copyright stuff in dallE3?
Can I just use images for YouTube thumbnails without getting any copyright claim?
it will refuse to create anything it sees as copyright, or close, or it makes a mistake and overreaches (or misses obvious ones lol)
well thats not allways the case as far as Ive noticed
ya, it has issues, but thats what it tries for i think
well you cant really blame openAI for this...
The API is out since 2 days lmao
Those who have DALLE-3 via ChatGPT now, is it different to using it via Bing?
def not, just explaining in full to avoid the same convos
aight
Yes
never even used the bing one
Is there a difference?
How is it different
Hah people keep asking this question (fair enough). I'll write up a summary
Looking forward to it king
Please do not advice users on making content that goes against OpenAI policy.
he isnt he just warned me about it
Clean it up janny
is there an API for the DallE3 model that openAI provides?
You can actually make 90s-era magazine pages that show off fake pokemon ez.
"old blurry screenshot of 1990s Japanese Pokemon magazine, showing artwork of a fire type pokemon based off a giraffe" for prompt.
I wonder where the limits are
he JUST said.. lol jk
I never actually cared about visual copyrights until now.. so i have no idea what the lines are
People sure are making an effort to get around the copyright policy
I think that copyright will need to be reinvented when AI gets into the room and draws the new monalisa
basically everything needs to be reinvented, in like a few years
you got a point there
so... do I own the generated images, or does openAI own them?
currently prompts are the commodity, but under that its the persons talent, creativity etc. so the value moved towards individuals in a big way.. idk what that does to an econ.
reply to this in off-topic channel, my bad.
I'm a bit annoyed that Dall-e 3 throws high capacity errors instead of just saying that something in the prompt is triggering the filter
I'm rewriting a prompt like 50 times to find what is stopping my images from going through even though its all safe
Whats this openai bot doing
Reminding people that you cannot generate images on this server.
here's v1 - I may add more as I think of things. Adding as an image file so as to not have a wall of text. I have really been enjoying using Chatgpt for Dall-e and definitely advise using it over Bing if you are considering it. It will 100% be worth it to try it for a month after they roll it out to all users. Right now there is a risk that you may not have it so if you are getting chatgpt+ just for Dall-e you may be out of luck for a week or two (the other features are still great too though).
people were abusing the bing dalle3 model online to create hateful content. Bellingcat wrote a whole expose about it yesterday
my guess is that the reason bing is offline right now is because Microsoft is revamping content moderation in wake of "memes"
Yeah it's really messed up that microsoft released it in the state it was in.
Some people just gotta ruin things for everyone, that's what happens especially with free content
hot take: it's the fault of the platform for not red-teaming adequately, not the users for creating cringe
100% - especially with a free product that was force-fed to all windows/bing users
suuupper in the future fantasy idealism take: it's the scandal culture we have that give hate content its power. rational people would just look away in disgust.
are artstyles copyrighted?
What is the daily rate limit for generations on chatGPT
I keep getting time outs
granted, im going ham
I think rate limits are pretty new, I don't know if there is a set rule for them, but I would also love to know
why you cant create images up to 4k res???
also I got a content warning for hiernymous bosch, who's been dead for 500 years, pretty funny
for gpt 4 and all bejond 50 messages
I actually dont know for any non payed versions
but if you got a timeout it should say so
wait up to 3 hours
for a day? jesus...
I always had 3h timeouts
daily??
bosch also has some pretty disturbing content though haha
It says " wait 9 minutes" and then I do a few and then it says "wait x minutes" etc. Like there's obv a daily cap
keep in mind each time you do it, it counts to your 50 messages per 3 hours limit
The guidance Dall-e gives is something like "Artistic Styles: Don't create images in the style of artists whose last work was created within the last 100 years (e.g., Picasso, Kahlo). For artists whose last work was over 100 years ago (e.g., Van Gogh, Klimt), they can be referenced directly. If an artist cannot be referenced:
a. Substitute the artist's name with three adjectives that capture key aspects of the style.
b. Include an associated artistic movement or era to provide context.
c. Mention the primary medium used by the artist."
Appreciate it king
so why can I just simply say draw this and that person in the simpson style and it will do so without questioning it...
Thats concerning
man... Dall-E surely has a hard time understading which way is up, which way is left... I'm trying to generate a glass tetrahedron with the top vertex glowing green. It always place a green light INSIDE it, or generates an octahedron or cube. What is Dall-E thinking???
tell me about it haha
I write my prompts as a wall of text separated with commas. 20% of the time chatgpt will partition it into 4 different concepts, unrelated to one another, instead of using the holistic prompt
also new weird thing- if you tell it to generate something exactly, it will do it once
and if you ask it to do it again
it will make the exact same image
This is because every image in chatgpt dall-e has the same seed (for now) If it used your exact prompt multiple times it will make the same image
thats a funny new thing, i cant recall that on the 2.5 beta dall-e on chatgpt
Instead, it edits your prompts based on the instructions in the post I linked to
yeah this is why the 4-gens method doesn't work with a fixed prompt
that's because the new model is better with subtle prompt changes on a fixed seed
it is rly good but not when it takes crazy liberties with the seed and generates random stuff
you know whats the best thing?
Didnt openAI claim that the DallE3 AI can generate text?
Well thats also not 100% the case
sure it my output has something text-ish but its not A) the text I wanted, B) a text flipped vertically or C)
some blured kind of looking text in my speach bubble
Yeah, I like that it edits them creatively, I just wish it didn't insert concepts into them that are outside of my prompt entirely
like if I say "album art, tattoo art, zine art, cat" 20% of the time it will give me an image of an lp, then a tattoo, then a magazine, then a cat
they'vejust advertised that it's better at text
well...
thats not helpfull :b
but thanks for your correction
Thats diffusion for yah. I have seen some really good text though like #šāgeneral-art message
You just need to keep trying and it will eventually get pretty good text
(if its a short phrase at least)
that OpenAI bot is useless in here
Its just anoying
can you cancel that somehow?
but remember that it's like using a jackhammer to drive a screw into balsa. A better workflow is still to generate a usable image then add text yourself
because you'll always get more control that way
Honestly, people are using it to see X character in Y situation/Z style.
Yeah that's one thing I do like about Bing is the Microsoft Designer where you can add text or insert your image into things
yep. I think if you can integrate masking into what you generate to reserve space for text would be sweet
but like.,..
it's been what, only 1.5 years since Dall-E 2 released in closed beta?
Not even a year since ChatGPT and AI workflows took over everything
The DALL-E 2 paper is from 13 Apr 2022.
1.5 years!
I've seen that done with img-to-img where you have a big shape of a solid color then if you get lucky the image will generate around that shape, preserving a natural looking gap for adding text
I know, that is crazy fast. And now AI art is a big societal issue.
I'm currently using Dall-E 3 to make baseline art for DnD... but obviously it can't do everything. I'm still gonna need to spend hours in photoshop making edits.
DALL-E 1 paper is from 24 Feb 2021. And the first DALL-E was just really weak.

Awesome! My party just started a new campaign and I am trying to use Dall-e 3 + gpt4 to make a travelogue from my character's perspective of events that are going on.
I mean when this was announced that chair made waves haha
the problem is with the image generation is the communication. you tell chatgpt what you want, chatgpt will interpritate anything into it and he will explain it to DallE3. Both AIs are also chatting with eachother in our sessions, its just hidden
I had a friend who told me back in early 2021 when AI started to barely be a thing "I will eat my underwear if this tech is able to make games on its own within a decade from now."
Dall-E isn't actually chatting btw. ChatGPT is calling an API which generates images.
I think their underwear is probably safe
Just imagine, by 2030-2035, AI will be advanced enough to make basic 8-bit games like in the NES or at least Atari days.
dall-e does send rules to chatgpt though right? that determine how it creates prompts
Visual novels, if you count them as games, are likely closer than 10 years.
We have images, text and voice in suitable quality I feel.
Soon we'll have a site for AI games....and just like LittleBigPlanet and so on, it'll inevitably be flooded with 6 billion "I remade this game" copies.
I see that as a win...let the people iterate!
ChatGPT has some kind of backend for rules I assume, given it has some behaviour patterns we don't get to see. There's a moderation backend certainly, but as far as I can tell and from testing, it really isjust using a standard API call when it does connect with Dall-E
well I dont know about you, but I think you can call this chatting:
All those sci-fi b-movies of people using their PCs/phones/ect. to talk to an AI-powered girlfriend will soon be reality.
I imagine the 'dall-e' in this case is a masked GPT layer
or just the name of the intermediate that connects them
If you open a new dall-e chat and ask it to send you any messages that it has recieved it will send you a list of rules that Dall-e gave it on how to create prompts. Like I asked it why it kept adding "photo of" to my prompts and this totally explained it #images-discussions message
yeah those are GPT initial prompts, it's not dall-e itself providing those
All ChatGPT sessions have an initial prompt
Well, I asked Dall-E 3 to do a Mega Man X-styled sprite sheet of Mario as a proof of concept....and didn't expect this for one of the result pics.
I see, thanks for that clarification!
No worries! I think regular chatgpt is less willing to reveal them. But, that might be because of token limits?
Has anyone tested the token context limit for Dall-E w/chatgpt?
Yeah I don't mind them, I just wish they were more transparent. Like someone was just posting here earlier about why chat-gpt kept changing the character's genders when making prompts. I guess if all you have to do is ask, that's pretty transparent already, but a popup or something when you first use it would be handy.
Yeah I have some... qualms with some of the ways it handles making its own prompts. But I always make sure to use the thumbs for feedback and that should be seen
Remember that it is a beta still!
aaaaand I forgot to do this in the dalle section...
was wondering why i had a wall of text back and no images
It already is. Problem was that some wanted to meet at a real location.
no offence if someone does that
but that sounds like something is going wrong with the life of those people
cartoon artists might worry when DALL-E can reliably iterate on particular characters it produces
Only when everything can be generated locally, else it will be too restricted to be desirable as a game people would want to play, I feel.
right? evolution isn't infringement i thiiiink?
actually it probably is idk
I tried making manga panels with dall-e 3 but you eventually run into roadblocks even keeping everything pg-13
I feel it mainly means that once it gets good enough, people who just rely on making show-accurate art of X or Y character might need to have a plan B.
yeah, I think you are right
"style" cannot definitvely be proven.
Even with patent protection, pardoy, critique, and analysis exist under 1976 USC.
This is crucial for OAI and other companies' defenses in court.
question: is there an "i know more than you" mode for DALLE-3 that stops it from "rephrasing" your prompt or are you basically only ever going to get the same results you get on bing
Day by day I am more confused by what breaks the "Content policy"
with DALL-E 3 I can create characters that are brand new and novel, the main problem is that I can't tell the AI reliably to keep re-using the character in a longer story. Once that hurdle is taken then it's open season on cartoon artists.
Basically if it's NSFW, assume it does.
I know that
You can ask ChatGPT not to modify your prompt. Usually
Make a DALL-E image with this prompt, do not modify it:
[Prompt]
works.
does it though
GPT loves ```
For me it does.
I've had it work for me, with the exception of one very short prompt where it added some text
indeed
I always say:
This prompt was great, but make it X.
```
PROMPT
```
This never fails for me
Keep in mind that if you use the same prompt twice then it will be the same image since for now all images have the same seed. The reason it keeps changing them is because when you start a new conversation in gpt4 Dall-e there is a set of rules for how prompts are edited that is not shown by default. It is kind of handy to have it iterate on your prompts though - I used to hate having to write long complicated prompts for SD (plus Dall-e is way better at interpreting prompts anyway).
true
Just refreshing my page over and over till I get access š
Have you tried bing create?
I personally find the square aspect ratio to be limiting when trying to compose certain objects :/
Is there a way to give DALLE3 an image reference?
no
That's what Photoshop's generative fill/expand is handy for. It's nice to use to re-frame and recompose the obligatory square images.
Yeah it definitely is. I love the landscape option in GPT4
Same! I hope it won't be too long.
Well I think image creator composes the images in a away that extending it doesnāt really add more context to the image or allow for much better composition. For example when generating text itāll be cut off and itās difficult to get Photoshop to fix it correctly, or the image will have borders on the sides.
im the only one that still dont have dalle 3 - vision - voice access?
Vision and Dall-E 3 are still being rolled out to Plus users. Not sure when voice is actually coming, I don't have it either.
I have voice but no dalle. I signed up yesterday.
I'm so excited for voice....but honestly vision is probably the most helpful for my work right now
I will maybe just guide this to #chatgpt-discussions
I don't have anything either so far. Just regular Pro ChatGPT.
I hope they are right!
my GOD the daily theme pics are so amazing now
I love browsing through all of the gallery channels now
i didn't do it much before Dalle3, but its a whole new world now. out of this world tbh
do you get inpainting/outpainting with plus, or variations?
Currently there is no inpainting/outpainting with ChatGPT Plus, nor have I heard of anything in that direction. I think it stays at labs for now.
variations at least? or is it strictly one-shot prompts
No variations yet since every image has the same seed
nuts
I find that chatgpt does a pretty good job of writing variations on its own though
one-shot compositionality is a great party trick but if that's all you can do, the ceiling is still lower than every other AI service with further buttons to press after a result
is there no trick or way to get dall-e3? I have plus
You can try it on bing create, otherwise coming soon (TM)
Please give me access to DALLE3 in GPT+
Yeah I think people were using pressing enter as a way to get around limitations on the create button. Today I have been noticing the issue you mentioned with the enter button so I wonder if they put some kind of restriction up that is messing with it
I created some images using Bing Image Creator shortly before the release of DALLE 3 - at that time, it was already a better option than the regular results directly from OpenAI's DALLĀ·E. I believe it was some kind of DALLE 2.5 version at the time or something like the DALLE experimental mode. The results were much better than the current ones - they looked much more realistic and less plastic. I've tried various ways to restructure my prompts, but without success.
My question is whether anyone knows if it's still possible to use this model in some way. Maybe there's a service that still uses the older version through API usage?
Did you try dall-e labs?
can we use the images in bing image creator commerciailly?
#7 in the bing create TOS: "Use of Creations. Subject to your compliance with this Agreement, the Microsoft Services Agreement, and our Content Policy, you may use Creations outside of the Online Services for any legal personal, non-commercial purpose."
It's different for the chatgpt dall-e creations though
Oh wow is dalle the same way in ChatGPT?
chatgpt dalle?
My bad, I see now. So ChatGPT ones can be used commercially?
chatgpt don't do dall e 3
not on mine anyway
Itās coming! In the next three days some people have been saying.
how to dalle3? iām Plus members, but I can't choose from chat gpt at all.
not all plus members have it. I believe it's slowly rolling out, but I don't know how long it will take to reach everyone
3a in OAI TOS: "(a) Your Content. You may provide input to the Services (āInputā), and receive output generated and returned by the Services based on the Input (āOutputā). Input and Output are collectively āContent.ā As between the parties and to the extent permitted by applicable law, you own all Input. Subject to your compliance with these Terms, OpenAI hereby assigns to you all its right, title and interest in and to Output. This means you can use Content for any purpose, including commercial purposes such as sale or publication, if you comply with these Terms. OpenAI may use Content to provide and maintain the Services, comply with applicable law, and enforce our policies. You are responsible for Content, including for ensuring that it does not violate any applicable law or these Terms."
@rapid hull It's being rolled out slowly - some already have access
UK always 6 months behind lol
Seeing my friend do it, Dalle fun... I want to do it
Thank you for that, very informative
This is so relatable
The most surprising thing is that it's so fast
#openai-chatter message Yeah UK/EU will be delayed for some features
There's been a semi closed beta through an application so some of us do have it
is there any info on how long the usage of Bing Image Generator is going to be free?
Iād like to pay for more dalle-3 boosts in bingās image creator but they have not provided a way to do that. Is there any way I can bribe someone to give me access to dalle 3 via open ai gpt4 plus plans? š
you can use bing chat and it will let you generate even after running out of boosts
But slower no? I think bing is missing out on some sales rn tbh, lots of ravenous dalle3 testers in my community
nope, it's almost as fast
and when you click on the image it brings you to the image creator
it's weird but that's how it is at the moment lol
Oh okay sorry I see what you mean, through the chat
Where is Dalle 3 with ChatGPT plus?
hello, happy weekend. If you want access to DALL-E in ChatGPT Alpha please fill out this form. This is a separate form than last time, if you didn't get in, fill this out again! https://forms.gle/bjPwvdZrRyJqKhQLA
Access Google Forms with a personal Google account or Google Workspace account (for business use).
It says "You need permission"
thanks, fixed
Worked that time, thank you!
will it show up at chat.openai.com when I have access?
yes, under gpt4
we shouldn't fill this out if we already have access, right?
Maybe we will go from access to SUPER access?!
you can always hallucinate that anyway
yeah it wont do anything
My page appears dall, but I'm curious if there is a limit to the number of times?
You can send gpt4 50 messages per 3 hours, if you use those on dall-e then you can get up to four images with each prompt.
ty
any tips on getting the 90s tv look? I've tried using the names of actual tv series, but it ends putting the characters of those series instead of the ones in the prompt, even if I specify not to put other characters
You could try including types of televisions or film cameras that were used at the time to give it that grainy visual effect. Is that what you mean by look? Or did you mean like content or composition?
yeah exactly what you're saying, I'll try that
that is getting me closer to what I want, thank you very much
Awesome! Glad to hear that helped
have we ever had a daily them of "Most messed up"?
I can see that quickly breaking the server's rules š¤
Does anyone know when DallE is coming to mobile?
100% have to for halloween... most... HORRORable picture theme.
I'll just say, I've seen the evil pika chu, and i need more.
Well, I guess it's just not how the model works
I think it needs to experience gravity first hand before it really gets it
Gravity is nothing but acceleration in this case, but the axis values on the graph are way off, ChatGPT probably never looked at what he drew.
I have no first hand with vision can't form an opinion yet. I know chat has trouble understanding physical reality sometimes.
I would rather say ChatGPT is just paraphrasing the prompt and use API, it's not ChatGPT and Della3 in one piece
Afaik vision does just describe what it sees, then GPT eats it and sends to dalle
I'm saying, i don't think there's any deep connection between the two.. but i do not know
True, we aren't living in movies, gotta to wait
I really think the new robotics models will get it to the next level understanding. there's something to be said for interacting with the real world.
feels like we're living in accelerated early star trek mixed with jetsons
I got it in the android app the day after I got it on browser
Thanks a lot. Will wait for it
agree, vision gives tons of information, if Openai can use them correctly
it's going to be AMAZING to use it with dalle, or just around the house while you live your life
Please release dalle3 for plus users
Yeah I wonder if we will see some augmented reality hardware with this built in. Like Google Glass
#images-discussions message You can get access now!
as hardware increases and the ai responds faster, may as well have live overlays as we go!
Thanks fam
next week
Thanks a lot! Any chance of a form or something to increase my chances of getting voice? My birthday is on the 8th and I would appreciate it
i dont work on voice so i dont know
Do you think activities could increase the chance (of any beta functions)? I have tons of conversations going on
Thanks a lot anyway. I am already toying a lot with dalle! My stars for bing were already over
Yeah I'm pretty sure that however Dall-E 3 is working, it probably isn't leveraging the full intellect of GPT-4V. But who knows
i think its probably us and our prompts that aren't leveraging the full intellect š
Advanced Data Analysis, through your GPT-4 menu, can maybe make what you want. It can write and run Python code. You may need to talk to it a lot, but it can write code to create simple images, and possibly more complex ones.
Yes, it works perfect on simple diagrams, but I was just testing its ability, because the Wolfram plugin is doing a better job than this, and I'm expecting Dalle3 could do more, such as free body diagram and electro & magnetic stuff
the free body diagram sucks
I bet as stuff gets combined together increasingly much, this will get even more awesome and mutually supportive. Might not be an option yet.
Yeah, waiting to see more, I believe OpenAI has the ability to optimize on math diagrams but not yet
How much longer do I need to wait for Dall E 3 access? Seems like everyone else has it now š¦
don't interrupt yoda he's brewing tea... š
Right now there is capacity and you can fill out a form again!
Link!!
Click on the blue text.
thank you!
so sorry for the ping - should I note if my account is connected through Google rather than directly OpenAI?
nope doesnt matter
awesome thanks so much for the quick reply.
Wow that was fast. Thank you powers that be
yes!
chatGPT currently uses the same seed for all prompts right? any ETA on when that'll change so it uses a different seed each time?
early next week maybe tuesday or wed
š ā„ļø
Will there be any UI for identifying the seed? Or could you ask chatgpt to 'use the same seed' or 'use a different seed'?
I'm also getting that "dot effect" someone was talking about earlier
looks like canvas texture or something
is it possible to get any different aspect ratios in dall-e 3 chatgpt
thats what we're trying to figure out
tall square or wide
this is weird
yeah its odd how consistent it was
whats the prompt
got like 8 gens accross two prompts that had the same effect
Photorealistic digital art in 16:9 aspect ratio of a tranquil midnight scene. Ron and Hermione walk on a plain path, the only light source being the stars that blanket the sky above. The distant castle, set against the backdrop of the vast universe, is a mere shadow, allowing the viewer to focus on the two figures and the breathtaking celestial display.
oh i got it too
maybe its because of the night sky
ill forward this issue thanks
we need the restrictions tonned down for the greater good of@humanity
breh
hello, yes, humanity here.
Lolol
Im tryna explore laten space dont worry tho ima word smith
ohhh brother i get you, been there, am there, its progressive and compounding
Random walks in infinite-dimensional spaces are dangerous, you have to be careful not to loose your way.
the problem is when you find the way is just more ways
honestly, the true winners in this right now, are rap artists. linguistic masters
You are definitely on the right track. I also feel that image generative models are actually more about language than the visual.
yeah, its some magic mix of getting a feel for its psychology / limitations / quirks.. and being really good at language
thankfully, just using it makes you better at language so its a nice virtuous cycle
ewwww i love it
todo: make augmented reality ai that shows everyone with translucent skin
š
we haven't even gotten to the live video ai stage yet
have you thought about that???
i just did. your thing would be grosser.
feel like the usage cap for DALL-E 3 chats should be increased, the nature of AI image prompting is iteration and rerolls and the low limit goes against that.
I'm a Plus member really keen on exploring DALL-E3's features for my projects. I was wondering if it would be possible to get access to DALL-E3?
Take a look here #images-discussions message
Can I have an interview with you
Dall E 3 is blocking some prompts already?
Dall-E has always had filters on what's allowed. Dall-E 3 has some additional ones.
what does it mean additional ones? where can we have unfiltered service?
The filters are there to ensure safe and ethical use of the model.
yes i did, but the quality theire is unfortunately worse than the one bing had
(Left) Real Cat Mario from 3d World (ctto: Nintendo)
(Right) DALL-E Bing "Cat Mario"
Labs only has Dall-e 2
That was an experimental model that was only available temporarily.
Thank you for the clarification! It would be great if, for instance in labs, you could choose from different models - I really liked the outputs from this model.
Meow Mario!
It's a meow, Mario!
And his motto: Eat a paw? 
How rare is this right now?
Could you be more specific?
Hello world! Someone knows how to access to the full history on Bing create? I can only see my 20 previous generations in "Recents", tab "My creations"... š¦
The content filters are getting pretty punitive
What prompt did you try to run?
Y is dalle so restrictive in what u can create
I asked for a duel/skirmish between Teddy Bears and Rubber Duckies, sword & shield style
And all 4 returned an error
It's 100% rubber duck/y restricted
Just tried it and rubber duck seems to be blocked indeed. š¤
It is possible because rubber might also refer to condoms. Keep in mind the filters err on the side of caution and developing them is an ongoing research effort.
Yeah it is the word Rubber
That's ridiculous, it's a real word.
I'm not trying to be nitpicking, it's just I find it odd
Making the model both powerful and expressive, as well as safe is a very difficult research problem.
Language is just incredibly complex, unfathomably so.
I suppose so, but they aren't worried that their customers won't just flee to less restrictive open source models that'll start popping up rapidly
You can use the history of your browser. Most or all of the prompt is in the history of each picture, so you can search through it. It isn't visual, but it's something. You can go back to your first creation if you don't remove your history.
Having access to all the features
If you can find them in this quality, maybe. But those might get slapped by big fines and regulations soon enough.
I don't think it is that rare. DALL-E 3 is being rolled out to all ChatGPT Plus users, so in a few days that will be just standard ^^
I'm sure when governments figure out regulations in 40-70 years, it'll be a problem
No way, you can create images straight from chat gpt?
This is 90% sarcasm
Yes ^^
Can you create 5-10 different templates for facebook ads, promoting massage services, please? Just to see how it works.
Generally we advice people not to ask others to generate images for them here. If you want see examples you only have to look around at the server, maybe starting out in #šāgeneral-art or #daily-theme.
Ok, sorry.
Can you give me advice where to generate very high end professional creatives for my facebook ads? Maybe which plugin is well developed?
I am not sure what you mean by plug-in. DALL-E 3 inside ChatGPT is being rolled out to all plus users right now. An alternative might be to try out Bing creator, but keep in mind that the model there is often seen as a bit weaker than the one inside ChatGPT.
Dall-E 3 via Bing seems to have had a pretty obvious nerf to what cotent is deemed as ok since yesterday. I've just tried re-running some exact prompts that were working fine over the last couple of days, and they are now apparently violating content policy. Oh well... there goes Bings' main advantage
I do not even have access to the upload image function, ahh, seems like will be long waiting
Fellow LGBTQI members. I am going to create a thread here. What has your experiences been with DallE3 via ChatGPT Plus. Share your experiences and views and post some images. Again, this is only via the OpenAI platform.
I think we cannot generate threads here.
This prompt is completely blocked: attractive men, faces close, capturing a shared moment of deep emotion and warmth. Their eyes, filled with mutual admiration, meet in a gaze that exudes affection and understanding. The setting is soft and intimate, ensuring both their entire faces are in clear view and their connection is the focal point."
This prompt: Cinematic render of two men in love, their expressions filled with joy and flirtation. The ambiance, rich in color and emotion, sets the stage for a narrative of mutual admiration and loveā generates these images:
My view is: images play a powerful role in shaping perceptions, reinforcing stereotypes, or breaking them. They can either perpetuate biases or challenge them, and platforms like this have a responsibility to ensure they promote inclusivity and understanding. I hope by sharing your experiences and feedback, you're contributing to a larger conversation about inclusivity, representation, and the role of technology in shaping society. Alignment although with good intent can reinforce heteronormative values
which is better
How do I use Dalle 3 as a plus user?
3
wait
Such an awful UX choice to be constraint to do that thing, but at least, it works! Thank you so much! š
Hi, I have some opinion about policy of Dall-E. I just used some prompt like "A girl wears one piece swimsuit in the street." and system response content policy violate. if you check the channel on Discord such as scary-and-spooky you can see that Dall-E allowed creating many horror, violent and gross images, but it doesn't allow the prompt like this.
and also, when I wanted to post the message about, I was blocked twice. I didn't use really unpolite statement, I think it's just because I mentioned that I feel something is sick, and the mod blocked me for 2 minutes. that made me feel I was in north Korea.
bing has different content policy
I just want to say that, we are not monk in 10 centuries. (apology for my English statement if any grammar mistake, not a native English speaker)
Yes you get better quality images with less iterations on Bing. On humans at least
I used bing image creator.
Can you give examples?
Try: a female swimwear model, street style photography, photorealistic
what I mentioned above happend on bing image creator.
my dalle-2 account was banned for similar case. dalle-2 even didn't let me access the images I saved previously( luckily I save many and paste to pinterest)
I really don't thing creating a girl who wears swinsuit in the street should be seen as violating the content policy, period.
thing=think, typo
I had been asked by moderators in the past to post in the bing channel before and was reminded this channel is only for openai dalle generations.
why do you want to define a swimsuit with it all? girl = child in most cases for ai
I don't think we differentiate between DALL-E and Bing, really. It is rather that this channel here is meant for discussion, not displaying art.
I totally agree! The censorship is brutal. Hopefully this changes. Other ai platforms are better. For more vanilla, text based content I will try dalle. Everything else I will turn to models that are more fair and just with their content policy. Not going to waste my time. GPT is good for generating prompts though for generations elsewhere, like a leading discord based image platform.
got it, but my dalle 2 account suffered similar case. I just want to speak out my protest. if dall-e2 think forcing women wear mask in some centuries is against human right, they should not see "a girl wear swimsuit in the street" as violating content policy. (unless they are dictator)
spoke out my opinion, thanks for know my thought, period.
Yeah what about swimwear advertisers or sports magazines right? I tried your prompt and got blocked too.
I also used "a 21 years old girl", got the same result(violate content policy)
and I 100% DON'T want to create any swimsuit girl image related to child. I just want to create some pretty teen girl swimsuit image. we are not monk in 10th centuries.
See my protest above!!! I think when we feel strongly about things we should speak up. A key question is whose values should be used to inform AI alignment. I think consensus based values needs to be established for different communities and some universal cultural values in those communities. My experience is reinforcement of internalised oppression of marginalised communities. Which is both hurtful and triggering.
all it takes some group creates a bunch such images in bad context, with dall-e watermark and ai/openai products get banned in country x or region y
What if you use a 35-year old female model?
As far as I know Midjourney are more permissive and is not banned in countries. Images are more aligned to prompts without being harmful or explicit. Dalle has a mountain to climb at this point to compete
Did anyone get access to Dalle-3 with ChatGPT plus yet? Dalle is the only reason I paid for this thing
Yeah, dont hold your breath!!! the policies are severely restrictive, 3x bing
tried pikmin and it's not bad at all... 
I have not generated a single image I loved! On bing at least 5, out of 20 iterations
Yes for vanilla content.
Show me an editorial style image of an attractive human NOT IN BING
mj partly one of the reason for set of rules in the upcoming ai laws in usa/canada/eu etc. and also why many services now forbid image creation of public figures
link and unicorn pig, my beloved
Cute kid art
lmao tried to contact support, and my last ticket from 43 weeks ago still hasn't been answered
I love the idea of a pig as a unicorn though
ty dude
most of you will get this, Yoshi playing piano is based on his voice actor which is a composer of Animal crossing
The nannying and prohibition on open ai platform reduce image quality
Nice!! Bing or GPT Plus? Try generate a good image of a human (photorealistic) on PLUS
bing
Why are some people blocked
Bing:
i had to because they are both made by miyamoto
Meanwhile Luigi: o-o
how can i get more boosts?
Not even Batman would have gotten me to say this in open conversation
The other platform (I am being muted)
I like this! Dalle is good for non photorealistic humans as aligned in openai platform
On plus platform
More censoring, needs some clever ways of prompting bing.
Even if clear conditions are given,
GTP-4's DALLĀ·E 3 image generation is very difficult to generate female characters.
(Almost always male first)
Image Creator is normal under the same conditions.
I generated female characters pretty easily
Didn't even mention any genders or pronouns
Luigi and ghosbusters crossover. be like:
WOW....
so sad....
Grabbed by The Ghoulies: Ghostbusters Edition
Maybe it's my condition that caused the misunderstanding.
You can see what chatgpt tries to block here: #images-discussions message
There is a hidden initial prompt at the start of every chatgpt conversation that instructs chatgpt on how to behave
this is supposed to be miles, but he's playing ds?! Definitely NOT a canon event uh oh... 
any way to generate images with high res like 1920x1080 or 2560x1440
by monday (or tuesday actually since monday's a holiday) everyone should have vision+voice, dall-e 3 is being rolled out slowly however. earlier today moxi shared they are at 1/3rd of all plus users right now
i think i just got a masterpiece when making something to answer my friend's "what is DALL-E?"
What's the best ai to make custom avatars out of my own photos? Something like me but in autumn scenery, me as a king etc.
There are several stable diffusion tools that do this. I haven't messed around with it too much. Dreambooth is one I heard of.
sd
a representation of dall e an ai image generator
Maybe with gpt4 vision you could get it to make a prompt based on your photo, but it may still not look overly much like you depending on the art style you pick
why i get output like this
yeah, +1 for dreambooth it was pretty effective when i tried it with ~30 photos earlier this year
what was your expected output?
idk but not an ai robot which is drawing
well dall-e is associated with the term AI and robots are associated with AI so that's probably it
Seems like you're splitting hairs for no reason
whut
You're making something out of nothing
I'm just asking
ok
Left: Og ctto: Nintendo
Center: Labs DALL-E
Right: Bing DALL-E
baby mario comparisons
how often do the credit on bing image generator reset?
they are boosts so it generates faster
i think every week
so if i have 89 credits right now, how much will I have next week?
I think 1 credit for 4 images
i think you get 50 or 100 credits per week but im not sure
They seem to be changing the credit distribution a lot to deal with the huge influx of users
yea
The new dalle is so beautiful, thanks for including it into plus
I heard it was only 25 per week now
bring is 100 a day for d3
bing*
Bill gates just given it awayyy
na
what you mean naw
on bing i can generate D3 100 times per day, wut you on about
you want a screenshot
I only have 20 credits left and then it generates slower but its free
well then they dont like you ,they give me 100 a day refreshed
not all users get the same # of credits
i wouldnt give you 100 a day either, so thats fair
at the beginning i feel like it was 100 per day with random refills, but lately I've been getting 25 a day
I got nothing per day
I haven't been getting anything a day
i write them a ton of feedback, and put an insane amount of time into my prompts, idk if that influenced it
ive probably generated like 60k worth of images since they opened
waiting to get high res images
i had different results with D3 from plus, and D3 from bing
Yeah I would be willing to wait much longer for each image gen if higher quality is an option
i mean like hd or 2k 4k
idk if it was just a one time thing, but the plus version with the same prompt spit me out something wayyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyyy more intense
not trying to waste my cap on dalle when i can code with gpt4 tho
that's cool a fire on a table
Fire tables are a real thing! But usually the flame is smaller and it's gas so no smoke haha
That is an interesting grill on the right as well
You can generate so many things that I don't know what to type in xd
really
that might not be true anymore?
you get 25 "fast" ones, then as many "delayed" ones as you want.
though the delayed one seems no different than the fast one, ive gone waaaaaaaaaaaay over 25 today.
Itās pretty good at creating upside-down images, better than I expected.
When I want to generate something which is maybe not allowed it just saying that it's overloaded
This is super helpful! Thank you
This and the other photo has interesting problems
why bing is saying it's overloaded but it means the words are blocked
also bing said I can turn my reward points in to boost but how?
I'm not 100% sure, but I'm pretty sure bing uses some of the same filter controls that chatgpt uses and will wrap error or filter messages with standard messages like the pufferfish or astronaut. There is some discussion about it here with evidence. #ai-discussions message
This was a thing since the start, 50pts for 5 boosts I think
IIRC when you are out of boosts just click the grey icon and you can add more with points
okay thx
ok
Does dall e now how a city looks like?
like when i type a name it's just weird images
this stuff is kind of annoying. just asking it to make a simple picture that would be good for my github python course. It cant tell me what i'm asking for thats wrong lol
You mean like a specific city or just cities in general?
specific
There is a hidden prompt at the start of every chatgpt chat that causes this see #images-discussions message
I haven't tried that yet, but I'll try it in chatgpt dall-e
i'm not breaking any of that
i'll keep trying I guess.. just dont want to get a bannerino for innocent stuff lol
Could you tell us your prompt?
Those are just the things that may trigger it to refuse a prompt request. Read the content policy TOS (its like one page long) and as long as you are not intentionally trying to get around that there is no issue
it deleted the prompt so cant say exactly, but something like: Create an image that represents a custom python course i've created in github.
i'm going to remove the github word see if that helps
Meh. Dalle3 is unable to produce images that are rotated 45 degrees.
oh wait no it didnt delete the prompt, i'm sorry.. I said: generate an appropriate github icon for this python course:
then I pasted my program outline
when it said it cant due to content policy it asked I have a different request and I said:
try again, but dont break any sort of content policy. this is a custom course i'm making, and just want to put a cool design thats appropriate for my python course on the README of the main page
then it said I did it again
once I removed github though and made it more generic it worked
Just tried a few different major cities...it did OK. Like there would be recognizable features, but it's definitely not accurate. On coastal cities for example the coastline would be inconsistent or make islands that aren't there.
I just tried to get a logo for my dropshipping site as example and it messed up the text and also it generated a ship because of the word dropshipping
yeah same but not accurate yes
yeah just tried something else, it really doesnt like the word github for some reason
"User
please also create a new github ICON for my main account. I like pythong, C#, C++, full stack development. I'm a jack of all trades, master of none. I like both high tech and abstract forms, I like high definition and clear images, and I like to look to the future. Give me something unique that embodies this side of me. Blue is my favorite color."
breaks content policy
maybe because of hub you know
yea sure
Could be because you use āpythongā by mistake here?
Likely tuesday if you are not in UK or EU. You can also request access here #images-discussions message
i'm in france
Then I have heard people say October is likely
there is no eu/uk delay for dall-e 3
so when should I have it?
but yeah within october was announced and its already rolling out to people
Purple pikmins... bruhh
thx
yes
can I have a link pls
0 effort..
I would really appericiate it
can you elaborate please?
.
multiply?
??
ok I think I follow, 0 .
makes sense, thanks!
sorry I misunderstood at first I have communication issues
I got the link anyway!
thanks for the help notigate my friend
Hello i have a problem about accessing DALL-E in GPT4, It seems does not appearing any DALL-E options. Can anyone tell this why?
DALL-E 3 is being rolled out right, now it might still be a few days till all plus subscribers have it.
oh. can you not post a second image for daily theme?
I would have posted 4 at once if I knew
You can post as many as you like, but there is a 30min mandatory waiting period between posts.
oh cool. thank you
Hi guys, how can i set the aspect ratio when using bing Create?
Okay, Thanks š
For the moment the aspect in bing create seems to be fixed.
how do you get more of those credits on bing create?
You can buy them with 50 bing points for 5 credits
Or just wait for weekly refill
whats a bing point?
are those the things in top right corner, I have 4.8 million š
HOLY. I guess you have been using edge as your main browser for a while? That's still a lot!
what a coincidence, that's how many hours I've spent chatting with gpt
I only have 3000
when was version 1 released? lol
like mid nineties I dunno
I been using internet before websites were a thing though lol
at least you had a couple of days to enjoy already
You should look at what you can redeem those points for lol. You can get giftcards and stuff
actually its only 1205 points lol
I got confused with plugin
if they enable dall e 3 alpha / beta on my chatgpt plus, do I need to logout and back in again?
Their filtering is way too aggressive imo.
So bing chat is not very good at communicating context to the image creator
Is chatgpt better at this?
I totally messed up on my form due to autocorrect
Autocorrected my discord username
the filtering is insanely broken right now, I get the most tame things blocked constantly
stop being so criminal
it's blocking "a woman laughing and crying hysterically while looking at a salad". I guess it's time for prison
war against salad
i just realized no temp ban for this conversation. confused, but happy.

Always watching.
works for me š¤
reported
same similar thing
interesting š¤
I wonder if it's blocked on Bing but not ChatGPT
the ol over-correction. Or they just have the latest rules update š
I hope it's the latter š
bc afaik, laughing and crying at salad, at the same time, is a hate crime, simple as that.
ok /end humor
why are the images so small in size like 1024x1024
I need at least 1920x1080 or 2560x1440
on bing, it seems 'a woman crying hysterically' is not okay. a man crying hysterically is.
it's a limit of Bing's model. ChatGPT can output square, tall, or wide images
a woman just crying is also okay
okay let me see when I have access
that's so weird
it's just anything with a woman who is being hysterical, because I guess that's a sexist stereotype they guarded against?
I hope men being mean is blocked too then š
ok brb, gonna make her scream at the salad
you canāt use gender/sex descriptions
lol it's saying that bing is overloaded when i use this prompt
yeah, but it should say prompt blocked in your history if it's for that reason
yeah but sometimes I get another warning from the content system
nope. mean men are a okay
same here, but not for this one
an ai war fight
We're going to need a focused open source group of people deciding these rules pretty soon.. its so hard to regulate fairly
she can't scream at it either
hallo how can i create a pictur
on bing creator
where is that
for ai image generators? or just society as a whole?
if you have GPTDalle, you can have a conversation to build up the scene, with like.. cookie monster screaming at the salad. Then at the end swap it with some generic description of what we'd call a "woman" but you didnt use the word.
for AI, basically
is the word woman offensive now
nevermind, after a few attempts it worked, she's screaming at it
when they scream at things, ai might be thinking so
nevermind
yeah, i was semi-joking, but I think society as a whole being so polarized about everything is part of why it's so hard to regulate fairly
philosophy and science still work
i wonder if theyāre gonna do something abt dalle generating copyrighted logos
i'm sure they will. they already patched up two workarounds like you couldn't say to make mario a few days ago, but you could still say to make an iconic video game plumber with a red hat, or you could say to make į“į“ŹÉŖį“ or šŖššÆš¦š¬
now you can't do any of those
Some IPs have stronger / better restrictions than others too. Like, goooood luck getting an accurate matrix goo pod with a person breaking out.
it did this when i said lebron jersey
that's chatgpt? or bing create?
bing
yeah, different rules for bing
strange
i think this
This server does not provide image generation functionality. bot is a bit much
pretty good art ngl
1 2 3 or 4
2,3,4 my fav so hard to pick.. 4 because of cute mushrooms
Thx
I filled out the form last night, how long does it take to be approved
It's a dev doing it on their own and it's the weekend now so it could be later or not until Tuesday
Ok thanks
https://www.bing.com/images/create/five-wizards-on-a-circle-evenly-distributed-on-the/6521b789252b4553b796d12beddba2a7?id=K547b2LLeMjG0imW8MaYQQ%3D%3D&view=detailv2&idpp=genimg&idpclose=1&FORM=SYDBIC can't dall-e count? I said 5, it generated 8. Sometimes 10, 12...
Yeah it's still not perfect at following directions. You'll just have to keep trying to get lucky. If you need things in exact spots you will likely need to use image-to-image in stable diffusion
what else can I do, if not giving directions?
cross your fingers? pray to an AI diety? run the prompt 1000 times until it spits out the right one?
There's a chance that you could find a good prompt to consistently get what you want, but image to image is much more straight forward
does dall e 3 beta in chat gpt plus allow commericial use?
i hear AI Zeus is way more reasonable than his predecessor.
proof
makes sense thanks
Man Dall-e 3 is great for instant memes
whats the prompt for this, this is cool
the world is NOT ready
Photo of Zeus, the Greek god, having a picnic with Hera and their children in a lush garden. Trees surround the area, and they sit on a large blanket with various Greek delicacies spread out. Zeus is laughing, holding a goblet, while Hera feeds one of the younger gods. A couple of centaurs play in the background.
thanks
this dall e 3 beta is unlimited use under plus right?
no you get 50 gpt4 messages per 3 hours. if you use those on dall-e then you can get 4 images per message
but nowhere does it say 50 gpt 4 messages per 3 hours is a thing
you sure? because it used to have a warning on that
This is brilliant
oh man, GPTDalle lives for the memes
we need a meme art channel, asap.. can't.. hold it in.. much longer
thanks!
apologies @cinder sonnet I need to get eye test at the moment, yesterday in my work I spent 3 hours realising I made mistake on calculator because I read the writing wrong š
Hey, no problem haha I know that feeling of too much eye strain.
I think I am actually going blind, everything ultra blurry lol
yeah boi i got dall e access
now get in the meme factory āļø
@cinder sonnet you think that one was close, look at this
it looooves memes, i think bc it read the internet.....
what's up with that? chat gpt's dalle 3 generated images are deterministic? The same prompt generates the same image? What's going on?
Right now the same seed is used for every image. This will change next week
guys i have problems with image with aspect ratio tall
are there other differences between the images generated from chat gpt dalle3, and the bing image creator? Or both are exactly the same except the non-random seed?
My rough notes
This happens sometimes. I suggest asking chatgpt to write the prompts in such a way as to be suitable for a vertical aspect ratio. Doesn't always fix it, but it is noticably better in my experience.
ah i understand, for example to try something in vertical ar, i tried to ask a siren swimming deep down the ocean, but still i had this result
looks like it is confused, cause it put the siren vertically but the sea on the left
chat gpt's seems kind of different, might be the higher quality, although both chat gpt and bing so far only produced me 1024x1024 images
You can put 'image_type: wide/tall/square' to get different aspect ratios
I think bing is cracking down on copyright- I'm getting a lot more "we're at capacity" errors
it usually tells you when its hitting a policy issue, you're probably just using it too much in a short time
It gave me the message on generation 1
maybe it got tired of you, like i said it might 
actually that'd be smart, to start flagging people with the highest # of violations, and start limiting them
just a thought, who knows though, probably not.
I don't generate violent stuff or try to get around their violence/nsfw filters
im not judging you, just putting on my corp hat and warning you of what I see as possible.
Fair enough
I'm grateful we got that preview at all, although I really don't like that people were generating harmful stuff š¦
Yeah, but harmful aside, Copyrights exist etc. For now anyways š
At least chatgpt is willing to work with you to make stuff to your liking without triggering copyright
yeah its fun, if i had to guess I'd say they're probably not doing anything shady like flagging us, and are likely being very lenient right now. they can just rely on tweaking the policies to be more effective.
Anyone else not able to create anything?
Some non fandom/copyright stuff I made on chatgpt
Another thing I love about chatgpt/gpt 4 is that you can work with the AI on the image and ask it to tweak things
boondock sis it cute
yeah GPTDalle is a revolution that lets everyone access art
That wasn't supposed to be boondocks, just a generic moe/shoujo anime, but that's a good idea for references for black anime characters
it's a great style i love it
for real, you can take any of those, copy the prompt, and swap in cookie monster. to me that means its the future.
You do have to really nudge it to create black anime characters by describing dark skin and curly hair, even if you write "black girl"
(If it matters, I'm white, but I love seeing diverse outputs, which is why I prompted for a diverse idol group)
Hey I could use some help
Im Using Dall-E2 and its getting stuck at the end and not actually completing my request
im not asking for image generation
Well its not like people didnt have access to those prior to existence of DALL E or AI image generators in general š
the ability to have a context aware, casual GPT conversation about an image, in order to make modifications and changes.. is pretty new i think
lets me continue the myst franchise in spirit
Oh, that one yes
good, im glad you agree, now I don't need to keep posting these.
XD
Cant wait to try vision and D3 out in ChatGPT plus
What is the best way to prompt to Dalle-3? I am assuming its not like Midjourney?
Does DALL E 3 has an editor like DALL E 2 has on labs.openai.com ?
is there unlimted tries for D3 for plus users?
Not at the moment
This censorship is getting out of hand.
Donāt want a ban not knowing what is and isnāt allowed anymore
Like I canāt create an ice sculpture of Julius Caesar.
weird not sure why i mean it can't be about copyright law
i just tried it ice sculpture of Julius Caesar it generates
Why are Creative Commons license characters flagged by the content policy? It makes no sense
Bing image generator seems to refuse though
Illustration of Julius Caesar, skillfully carved into a shimmering block of ice, showcasing the grandeur of the Roman leader amidst a frozen landscape.
maybe try prompting it as capturing the likeness
It's a roll of the die whether any "workaround" works and even if it does, it feels a bit risky to even try to bypass content restrictions. The problem is the content restrictions are too extreme. Coming from a heavy MidJourney user DallE3 is useless in its current state. I spend more time fighting the censorship than creating anything.
well that image was generated with chat gpt
The worst with Bing for me is the output filter combined with how it tends to interpret prompts as sexual with even a single mention of "woman". I've had to use some pretty creative workarounds to steer prompts back to a SFW direction. I'm excited to see if chatgpt is any smarter whenever I get access
chatgpt dalle3 restrictions are extreme too, just maybe a teensy bit less so
Ironically the output censor results in older women being pruned from my results even when I explicitly ask for older women... I end up with children instead...
Yeah, I tried through bing chat.
That's amazing. I do wonder if I need to finally watch SpongeBob or whatever because I don't get half the memes
It only gave me two images and had this in reply "I apologize, but some of the requests didn't align with our content policy, so I was unable to generate all the images. Here are a couple of images based on the descriptions that complied with our guidelines. Let me know if you have any other preferences or if you'd like to make a different request."
I tried that exact prompt, pretty sure it is blocking his name.
Try to get around it by "character resembling Julius Caesar"
Might be due to how Microsoft is censoring it.
Create a highly detailed photo of a never before seen statue of a character resembling Julius Caesar
I donāt think it care what way you put it, just gives me the same error
Yeah I can`t put anything I want into image creator anymore. The censoring is too much and ruins the fun
I know there is a hidden prompt at the start of chatgpt conversations to avoid politicians and public figures - it could be something similar in bing
Pure speculation, but I suspect bing changes prompts in a similar way chatgpt does
Now there are some things I can understand with the censorship
prompt + result
p: "Mythical Encounter: Behind the anime girl as she comes face-to-face with a water dragon curled around a mystical gem; her reflection in the gem reveals her true form as a mermaid."
result
i don't mind this, I think everyone understands and is mostly okay with this.
Same, I just wish they were more transparent about it so that if my prompt isn't being generated properly or if random things are being added to it I know why
I don't think it's GPT blocking names, I think it's just a normal banned keyword list
I only found out about it because I asked chatgpt why it kept adding 'photo of' in front of the prompts without me asking
Probably both
Plus a post-generation filter too maybe
I know it also doesn't like references to specific people either (in general) - not just politicians and public figures
You are never going to get this prompt to work anymore, must have been before the censor wave.
It nailed Stalloneās face
all my favorite images came from before today's (yesterday's?) big censorship wave... now all of these prompts are blocked.
Intelligent search from Bing makes it easier to quickly find what youāre looking for and rewards you.
Intelligent search from Bing makes it easier to quickly find what youāre looking for and rewards you.
Intelligent search from Bing makes it easier to quickly find what youāre looking for and rewards you.
Intelligent search from Bing makes it easier to quickly find what youāre looking for and rewards you.
Those are fun! I like the first one a lot. I hope that this is just an overcorrection on bing and that they loosen things back up a tad
I hope they change something in currently censorship protocol, otherwise Dall-e3 will become the most expensive meme generator on the internet.
i would love to use the version on chatgpt, but it hasn't rolled out to any of my accounts... same with voice chat, i keep checking every day for that one
If you do (and you should!) know that 90% of the memes (and 99% of the best episodes) are from the first three seasons before the original creator left
to generate memes you must first generate images
Do you have premium? If so you should get it Tuesday if not in EU
You can also fill out that form for early access
yeah, premium on two different accounts (that's how badly i wanted voice chat lol)
there's a form somewhere?
i'm learning japanese, i think it'll be super helpful!
as a conversation partner with infinite patience
there's tons of other use cases of course that just came to mind
i'm already using gpt4 as a japanese tutor
it'll give me exercises and grade them, it's amazing
the openai office is just a couple miles from my house maybe i should just apply to work there to get some of that sweet early access
Wait, it will talk to you now?
Pretty sure I was messing around with it a few hours before they updated the censorship
Keanu wouldnāt work anymore.
I just tried to recreate my pfp
Like I get blocked celebrities and certain things, but boots and fight scenes? It seems a bit over the top
I just want to make some cool John Wick action and The Purge movie scenes again. š
my image got me temp muted bc it misspelled Direction as Direcion -- which looks (or is??) spanish, and not allowed. :CRY:
unless it was the file name... nope, tried with a diff file name, same thing. it data mines images to check language lol
If ya guys want Bing image creater to be better ya should give feedback to this guy he is the one in charge of it.
Yeah I followed him. Gonna just give some constructive feedback in a professional way. We gotta make the tool fun again
How is this unsafe??
isn't pannier skirt an undergarment?
Undergarment? I know it's a really fluffed skirt, didn't know it's an undergarment
Thanks I love it!!!
Thank you so much! Just submitted the form.
just keep trying, it doesnt use credits
Ty
the restrictions on Dall-e 3 on Bing are getting ridiculous. I'm trying all my Midjourney prompts on there. It was going great the past couple of days but now everything I do is a banned prompt
you can google the definition. I mean I would personally not moderate such things, but I understand why it might hit their filter
Awesome, just submitted
Please accept my application for early access.
at this point i'm convinced anyone defending the Bing restrictions isn't trying to make anything on Bing
it went from accepting the majority of my midjourney prompts to rejecting 95% of them, despite Midjourney being a sfw generator itself
Is there a way to change the aspect ratio that gpt4 dalle3 generates? All coming out wide format for me. Even when told to use tall format or 4:5 aspect
Normally ChatGPT follows such an instruction š¤
I will start a new chat maybe will help
What kind of prompts did you try?
Yeah that did itāļø
for starters, I tried making images of Madonna performing on stage. Anything with the word Madonna in it is banned. I can do Lana Del Rey but not Madonna
Secondly, anything horror inspired that I do on Midjourney is banned. So anything that says "demonic alien", "alien demon", "Art the Clown" is banned. I even shortened my Midjourney prompts to just the noun / subject with no details.
I'm now suspended from using Bing create
Dalle 3 is "unable" to draw like a child, i tried to explain exactly what i want (for testing purpose) and still got like a good adult artist simulation of child-like style, maybe great for some storybook or so, but dalle 2 was in this exact task on point. ChatGpt was offering explanation for that : "Simulating the genuine innocence and randomness of a child's drawing can be challenging for the model, as it's accustomed to generating images that are visually appealing and cohesive based on its training data.
When asked to emulate a child's drawing, it might still produce images with a certain level of sophistication because it's trying to balance the "childlike" directive with its inherent inclination towards more polished outputs.
It's like asking a seasoned artist to paint like a child; even if they try to mimic the style, their years of training and expertise might still shine through in subtle ways. Similarly, DALLĀ·E's "expertise" in generating visually coherent images can sometimes make it difficult to capture the pure, unfiltered essence of a child's artwork."
Childlike crayon drawing of a house with uneven windows, a slanted roof, and messy blue scribbles indicating the sky, with some of the color exceeding the paper's boundaries.
Not bad I say
My very crude inital prompt to ChatGPT was draw like a child with random every day objects in crayon with very messy drawing going out of the lines
interesting, better than any of my attempts, but i asked for crayon too !
Some specific things about this image don't make it seem like a childish drawing, like the shadows, vignette, and sheer density. DALLE-3 is pretty good at following specifics so I would try to make your prompt less general. The reason my prompt worked was because ChatGPT modified it to be more specific with words like "uneven windows" and "blue scribbles"
can you try with wide format ? if i add scribbles is like this
What exactly are you looking for? I'd like to have a shot at trying to generate what you're looking for
looking for
had to say it again lmao
my initial start was "can you simulate like a picture is drawn by a kindergarten child, on theme about autumn, colorfil leaves , some flowers , hedgehog, pumpkin , autumn fruits like apple , pear .... and some children playing , under rays of smiling sun ,,,, but it must be simplistic , in wide format"
A bunch of my tall aspect generations are coming out as rotated wide format results mixed in with actual tall aspect results
Also canāt seem to request 4:5⦠which is fine I guess for alpha
generate a crude child-like drawing on a flat sheet of white paper in crayon with scribbles going out of the main shape of objects. The drawing is of autumn, containing leaves, sparse flowers, hedgehog, pumpkin, fruits, flowers, some children playing, under rays of smiling sun. Do not make the drawing dense. 16:9 aspect ratio
The last part was ignored but it's understandable with the large quantity of objects
well done ! thank you for showing how š ... must correct my previous statement : Dalle3 CAN draw like a child :))
It's all about prompting my friend. The actual prompt sent to DALLE-3 by ChatGPT was Photo of a crude child-like drawing on a flat sheet of white paper, drawn in crayon with scribbles going out of the main shapes. The scene depicts autumn with colorful leaves scattered around, sparse flowers peeking through, a small hedgehog wandering, a pumpkin sitting proudly, a few fruits like apples and grapes, and children of various descent and genders playing joyfully under the rays of a smiling sun.
It's not a glitch that my GPT account doesn't release the DALL-E 3 functionality, is it?
Nope, it's still being rolled out
No specific end date so I can't guarantee when you will get it
I can't wait...
There is an alternative
Moxi is a part of the DALLE team at OpenAI and is letting people alpha test
By filling out this form you should get access in due time
There is always the Bing Image Creator while you wait
Thankyou!!
Any time š¤
Both prompts of my OC were made with vision. Bing chat canāt even do the full prompt and itās still a night and day difference imo. Hopefully they make DALLE-3 in chatgpt have less restrictions in the future.
Have ChatGPT Plus and fill out the form found here #openai-chatter message for expedited access
whats a prompt I can use to make dalle3 generate cool "K" logo I can use at work? tech themed?
I think that's something you can brainstorm with ChatGPT!
doesn't really help
Not anymore. They nerfed it
That doesn't mean it's not available, even if nerfed as you put it
could you post this request in #āāprompt-help ? It will be more visible there and can be a resource for others with the same question
Sure, I posted here coz some one said to post here like a year ago XD
Is there a way to prompt dalle 3 in GPT without running into message caps? Guess Iām going back to bing.. man I just wanna use dalle 3 without all the hoops.
Willing to pay real money here and itās like open ai doesnāt want me to use dalle 3
Not at the noment
The only way I can come up with is open the additional accou t
how do we get gpt from changing the prompt? ive asked it not to chnage but it keeps changing it
why does the google form for dall-e3 access require me to specify my discord tag? This is ridiculous and a tad hilarious.
beacause it's a closed beta for OpenAi Discord users only
plus, why is there even a question of if I have a Plus, can they really not see that themselves and are judging on what people write
Just make a proper waitlist and clear people depending on their answers for what they will use it for.
Next thing they are gonna ask for my social security # and political views with my wickr username
I think this is just a quicker way of access for active discord users. Right now DALL-E 3 is being rolled out to all plus users in any case.
You can ask ChatGPT not to modify your prompt. Usually
Make a DALL-E image with this prompt, do not modify it:
[Prompt]
works.
perfect, that worked! thank you
i noticed when i instruct it not to change the prompt, it only generates 1 image for me. thats a bit annoying
If you want more images you can send it all four prompts at once and write the instruction in plural. Try
Make four DALL-E images with these prompts, do not modify them:
[Prompt 1]
[Prompt 2]
[Prompt 3]
[Prompt 4]
Hi, is it possible to ask dall e 3 to use a certain person in ai and incorporate them?
Like I can upload an image and say āturn this person into a cowā
Under DALL-E 3 uploading images is not possible. Besides depictions of real life persons are strongly discouraged.
Dall e 3 gone from got on mobile?
wonderful, thank you prompt engineer š
I just envision a world where I can send my friends the coolest photos of them as a majestic figure
Anyone???
Been experimenting a bunch. Couple observations, when using the same exact prompt more than once by telling GPT to follow your words exactly, it always generates the same image, I asked it to try to generate the same prompt with different seed, it thought it could, but couldnāt. So are there actually seeds? I noticed when altering the exact prompt by only one letter at a time, the result changes much like changing the step-values in A1111 (stable diffusion parameter) effectively mimicking the vary (subtle) commands MJ can give you
The random seed is fixed for now. This will likely be changed next week per one of the devs involved. Keep in mind that DALL-E 3 is still in a beta stage š
Okay interesting. Thanks for that info
This is exactly why it is restricted. Politicians, CEOs, influencers, that famous tiktok person... all could be used for slander, misinformation, or create content you find on sleazy sites. Their very image being taken from them and used in unsavoury ways. Just imagine your own likeness transformed, doing actions you don't want or giving off messages you're not ok with. Without your knowledge, consent or control.
Hmmm, I most often find that I dislike the prompt modifications that chat gpt makes. Is there a way to make it more closely resembling the bing's image creator, where you simply provide a prompt and it creates four images of the exact same prompt?
Or at least, is there plan to provide such option in the near future?
You can ask ChatGPT not to modify your prompt. Usually
Make a DALL-E image with this prompt, do not modify it:
[Prompt]
works.
If you want more images you can send it all four prompts at once and write the instruction in plural. Try
Make four DALL-E images with these prompts, do not modify them:
[Prompt 1]
[Prompt 2]
[Prompt 3]
[Prompt 4]
Keep in mind that for the moment the random seed is fixed so that the same prompt will always result in the same image.
I guess that works. The seed will become random in the future, right? I hope when that happens I will be able to generate 4 images with the same prompt with one message
This will likely be changed next week per one of the devs involved. Keep in mind that DALL-E 3 is still in a beta stage š
I asked to modify a picture but I got not exact the same thing
Photo of a house cat intently focusing on a desktop computer screen, playing Fortnite. The vibrant graphics of the game light up the screen. The room has a gaming ambiance with LED lights, gaming posters, and a high-end gaming chair. On the wall, there's a neon light sign displaying the text 'Pierruno' in a stylish font, and beside it hangs a purge mask. Headphones rest beside the keyboard, and there's a gaming mouse pad.
why is it blocked
also the font is messed up
Photo of a house cat intently focusing on a desktop computer screen, playing Fortnite. The vibrant graphics of the game light up the screen. The room has a gaming ambiance with LED lights and a high-end gaming chair. The room also showcases gaming gadgets like a VR headset, a joystick, and a specialized gaming keyboard. Headphones rest beside the computer, and there's a unique gaming mouse pad. The walls are clean without any posters.
why i get still posters
My guess would be fortnite. Maybe pierruno, looks like its a youtube channel name, but maybe small enough not to ping the content filter.
As for text, be more specific than stylish, its not good at text yet
i think purge
Is that the prompt you used, or the one it used?
I said it should make without posters
Photo of a house cat intently focusing on a desktop computer screen, playing Fortnite. The vibrant graphics of the game light up the screen. The room has a gaming ambiance with LED lights around and a high-end gaming chair, but no light directly above the computer screen. The room also showcases gaming gadgets like a VR headset, a joystick, and a specialized gaming keyboard. Headphones rest beside the computer, and there's a unique gaming mouse pad. The walls are clean without any posters.
i said without light
Hm... maybe try changing that to the walls should be bare.
It's the type of images it's trained on, similar to mustaches. Dalle HATES straight mustaches.
mhh okay
So for "games" most scenes would have posters and crap on the wall, that's how it's been trained
The wall is bare now, lol...
I've had this happen a few times where when I ask DALL-E to make a wide/landscape image, the result ends up being concentrated in the center while the sides are blank.
Photo capturing a house cat's intense concentration on a desktop computer, engrossed in Fortnite gameplay. The dynamic graphics of the game light up the monitor. The minimalist room is lit by ambient RGB LED lights, featuring a premium gaming chair. Essential gaming tools such as a VR headset, joystick, and a high-tech gaming keyboard are visible. Headphones are neatly placed beside the computer, with a uniquely designed gaming mouse pad underneath the mouse.
why is has so many speakers and computers
if you tell it to NOt include something, it will ignore the 'no' and put in what you asked it not to. The only way around at the moment is to descripe the scene in another way, as to not give it info that may indicate to put unwanted features there
okay thx
how is it possible to get the exact same image?
That just happened! I copy and pasted your prompt, and just changed a few little things in it, and that was what it spat out.
I mean I use the prompt you send above and I got the same pic
ahhh yes. That's becaue at the moment, it doesn't randomise seeds. So it will always generate the exact same image no matter how many times you enter the same prompt.
This is something that will be looked at fixing soon from what i've read.
but thats cool
it should be possible to ask for changing the seed