images-discussions

[plush hare]

Continuing on this theory:
So we saw that there was the "prompt modifier system" in DALL-E 3. I think that it does not only improve your prompts, it creates many subprompts for DALL-E 3 to generate each element of the image independently (maybe it also organizes where each element is placed). That is how DALL-E 3 can stay good with little details, understand prompts better in natural language, and include every detail of your prompt into the image.

---

Is anyone else curious about how this system works?

[dark condor]

Yes, I also theorised that it was some sort of prompt enhancement and amplification so that’s very interesting, It seems like we are on the brink of an irreversible moment

[dim cradle]

what is this? somebody posting images generated by Dall-E 3 in a forum, but when I access Dall-E 3 site I'm redirected to labs and with zero credits I can't do anything. The guy said that Dall e 3 is free. Free? where?

[waxen geode]

I’ve been waiting more than a day so far 😃

[dim cradle]

https://th.bing.com/th/id/OIG.VEet01w28UiA.GiPqAlb?pid=ImgGn Dall-E 2 generates images with better coherency than clipdrop.

clipdrop always put excessive lights, excessive reflections, distorted roads, cluttered houses and trees

[fading spoke]

what was prompt?

[dim cradle]

street at night with water puddles, japanese houses, full moon, stars in the sky, cherry blossoms, wooden fence, sidewalk, clear sky, a japanese style water pound to the left

[civic shale]

do you guys think this is ai generated?

[peak trellis]

I've been waiting on an image since yesterday, what should I do?

[nimble rose]

BRICS as pokémons haha

[civic shale]

retry i guess it shouldnt take more than a few seconds

[dim cradle]

sdxl in clipdrop is terrible at generating roads at night. It bifurcates roads out of nowhere, makes them distorted, places just too many lights and even draws light trails like a timelapse at night

it often generates houses on top of each other, creating a huge mess

is dall-e physically correct?

[dim cradle]

sdxl keeps putting wrong reflections and too many lights

[peak trellis]

is it just me or is this image not as detailed as it usually is

especially the face

[somber zenith]

Yes, it gives me the feeling that sometimes DALL-E 2 sneaks in...

[nimble rose]

Was this image generated with Bing?

[dark condor]

It definitely does. I can say this with 88% conviction. I would rather have nothing generated

[nimble rose]

It's funny because so far I haven't had any images with lower quality... For me it's been something very "constant".

[wispy storm]

What prompt ?

[nimble rose]

Although I don't try to create images that detailed...

[wispy storm]

For ne dalle 3 do an excellent job with chatgpt 4

[nimble rose]

Are you already accessing dall-e directly from ChatGPT or using Bing?

[wispy storm]

Chatgpt

[nimble rose]

When it was released to you, did you need to activate something in the "beta" functions or did you just try and it worked?

[wispy storm]

You will have another option near the selection of which chatgpt you want 3.5 / 4 / 4 with dalle3

[nimble rose]

Interesting, so it hasn't been enabled for me yet... I can't wait for them to enable it for me haha

[wispy storm]

It is a game changer imo

[plain moth]

Does it show on the mobile app, or desktop/browser only?

[halcyon oak]

Does anybody know when dall e 3 is coming out yet?

[peak trellis]

Napoleon Bonaparte in his usual military outfit on horseback leading his troops towards the oncoming british army, movie advertisement, 3d renders, high quality, highly detailed, macro shot, worm's eye perspective, professional, mind blowing, high saturation, studio lighting, unreal engine, promotional commercial movie poster

[peak trellis]

Yeah it was

[nimble rose]

I will try to create with your prompt

[wispy storm]

#announcements

[wispy storm]

Need to start the convo on a browser first

[halcyon oak]

It doesn't say when...

[wispy storm]

With dalle 3 chatgpt

[placid orbit]

this month

[halcyon oak]

ye i saw

but nvm then

i guess it's not known yet

[placid orbit]

probably sooner than later

[wispy storm]

About DALL·E 3
DALL·E 3 is now in research preview, and will be available to ChatGPT Plus and Enterprise customers in October, via the API and in Labs later this fall.

[placid orbit]

you can just use bing image creator right now which is the same

[halcyon oak]

👍

[halcyon oak]

It uses dall e 3?

[placid orbit]

you just cant use gpt 4 to generate prompts or change stuff like aspect ratio

yes

[halcyon oak]

Oh cool

[peak trellis]

Those look insane

[wispy storm]

Face details are impressive

[peak trellis]

I'm sorry for the second ping. I assume you used the same prompt as me. With that being said, how in your image, the face of Napoleon is much better looking than in my pictures?

[peak trellis]

agreed

[wispy storm]

I used same prompt as you gave to me. Like I said I used dalle 3 with chatgpt

[peak trellis]

I thought dalle 3 was only through bing image generator?

[wispy storm]

Dalle is made by openai

They have a deal with MS

[peak trellis]

I'm sorry for all the questions. I'm just confused on how you access dalle 3 then because when I search for it, it says coming soon.

[wispy storm]

About DALL·E 3
DALL·E 3 is now in research preview, and will be available to ChatGPT Plus and Enterprise customers in October, via the API and in Labs later this fall.

They started to rolling out

For plus user

[peak trellis]

ahh so you're a plus user?

[placid orbit]

what are u usingf

[wispy storm]

Yep

[peak trellis]

bing image generator

[placid orbit]

idk why its different, maybe chat gpt has a better version

bing probably uses a faster version

[wispy storm]

Chatgpt handle information better than bing

[plain moth]

So here's the thing, that is supposed to be Dalle3. But I've noticed the quality dropped sometime overnight yesterday. I think they changed some parameters around to try and combat the overwhelming traffic?

[placid orbit]

they also added more gpus

[peak trellis]

the part that has me confused is that earlier today, bing was able to give an image that was as detailed as this

but just a couple hours later and a different prompt and the face looks like this

[somber zenith]

Further down, the Open AI website says "DALL-E 3 will be available to ChatGPT Plus and Enterprise customers in early October."

[plain moth]

How's performance through ChatGPT? How long does it take for an image to be generated?

[peak trellis]

so will it ever be available to the general public?

[wispy storm]

Less than 40 sec

[plain moth]

Thank you for answering

[wispy storm]

I don't know. But I don't think

[placid orbit]

they normally give people a few free credits in labs

like for dalle 2

[peak trellis]

how much is it being a plus user?

[wispy storm]

20 usd plus taxe plus it give you lot of thing

[plain moth]

I'd hope they'd update the Dall-E 2 site with 3 as well so we can just buy tokens directly instead of being limited to gpt4 prompt quota

[placid orbit]

they will

later this fall

i swear u guys dont read

[peak trellis]

one time payment?

[wispy storm]

Per month

[placid orbit]

u get 50 prompt per 3 hours

[wispy storm]

Thing I like you can ask chatgpt to edit a picture it sent and you liked it

[placid orbit]

it can edit?

[plain moth]

I'd wait too, cause even if you get plus now you won't have Dall-E 3. It's a limited release right now, but I believe by the end of this week all plus users will have access

[wispy storm]

Or ask yeah change the top clothes, change the background for nightclub

[placid orbit]

or so its just changing the prompt

[wispy storm]

In a certain way

[wispy storm]

Yeah in the end it updated the prompt

[placid orbit]

I hope in the future they will add feeatures to make it more competitive with local models and MJ

[plain moth]

Could I send it a photo of my cat and then tell it to generate a pencil sketch of him?

[placid orbit]

like img2img inpainting etc

[willow yew]

thanks for sharing your experience so far. Does it have the 50 message/3 hours rate limit for Dall-E as well?

[wispy storm]

It does better than MJ for certain thing ( I have both )

[placid orbit]

yeah but you get less options to manipualte the img

[wispy storm]

I think yes. But I do like one thing every 5 -6 min so I never reach my limit

[civic crescent]

Have you guys got to use Dalle 3 yet?

[wispy storm]

I guess with vision and upload picture it will be a game changer

[wispy storm]

Yes

[placid orbit]

true

!1

[civic crescent]

Im too lazy to figure out the whole bing image generator thing i feel like i'm gonna wait 😦

[placid orbit]

???

[wispy storm]

If you have gpt plus dalle 3 worth the wait

[placid orbit]

ur too lazy to go to a web page and click generate?

its free

[abstract sparrow]

Oof, the bing image gen site is just outright not working right now

Like not even infinite wait times, it won't let me hit Create, and it has an error message saying the team is working on it

[placid orbit]

u have to wait for previous gen to finish

[abstract sparrow]

But my previous gen is finished

[placid orbit]

🤔

[civic crescent]

I have chat gpt plus im def waiting. And I herd you can start using it now with bing chat image generator, but i herd its kinda buggy and you have to use different browsers, i tried it doesnt let me

[placid orbit]

its just overloaded right now

yesterday i could gen in about 10 seconds

now its an hour wait

[strong badge]

only got to use the bing version which takes half a millennia to generate anything, I'm still waiting on a few images from a day or so ago. Can't wait for the chatgpt version tho.

[plain moth]

For me, it's always been it generates in 30 seconds or less. Or it NEVER generates. I have like 3-4 prompts that have been sitting at creating for 10+ hours

[placid orbit]

yeah lol

I hope they update the website so we can see all of our generations, it only shows first 10 or so

[civic crescent]

Yea me too. I feel like when you use chatGPT it makes better and smarter responses than using Bing, even though theyre the same model. I think its because its over filtered but im not sure

[placid orbit]

the boy who lifted

[plain moth]

idk what happened but I'm on a roll

my prompts are going in no problem and generating fast

aaaaand now it died... rip

Yeah... it does seem like the bing image generator has less detail now.

[cloud narwhal]

Dall-e 3 looks so good

[dark condor]

I’m sure they have some sort of sequencing algorithm that uses dall 2 when things are too much

[plain moth]

idk about that, because the images are still more detailed than Dall-e 2

[dark condor]

Maybe some features of dall e 3 just not being used. If there’s some prompt amplification system

[craggy mango]

Dalle-3 is pretty cool, using it through microsoft.. Whats not cool is not having it from openai, where I pay for chatgpt plus and spend quite a bit through the API..

[tacit mortar]

I hope they will release soon for chatgpt plus subscribers

[valid berry]

Why is it I have to go through Microsoft to use Dall-E 3?? Is there a way to use it on my subscription instead?

[tacit mortar]

not yet

[craggy mango]

Its free through microsoft too, so many generations.. I'm guessing gpt-5 with be named bing-gpt.. Microsoft is sleeping around to, it's not like they are staying loyal to openai.

[tacit mortar]

microsoft is more than loyal to openai

they invested 1 billion on it

[craggy mango]

Those are different concepts in my opinion.

[peak trellis]

Does Bing Image generator allow for inpainting and/or outpainting the same way dalle 2 does?

[plain moth]

nope

[peak trellis]

will they ever add that feature into bing image generator?

[placid orbit]

probably

im guessing dalle 3 was a multimodel test for gpt 5

[peak trellis]

is bing image generator not working for everyone else at the moment?

[vapid elk]

gpt-5 simply does not exist lol

[fading spoke]

omg bing is so overloaded

gawd damn

[placid orbit]

they will begin to train it probably in november since thats when they vowed to not train it by

[fading spoke]

It is getting annoying

[fading spoke]

MAN I HAVE A HUNDRED BOOST'S IVE BEEN WAITING FOR 26 HOURS

[drifting hawk]

hello! I am an engineer on dalle and I need some alpha testers. If you wanna try the DALL-E in ChatGPT Alpha and have ChatGPT Plus, please fill in this form: https://forms.gle/3DokUVBkzTU5ZaRV8

[nimble rose]

Dall-e 3 with ChatGPT is on another level... Fantastic!

[ionic dome]

dall-e 3 prompt adherence is crazy good

[midnight mango]

you can't do this to me! Now im even more addicted to this

[nimble rose]

I've talked about this before... But OpenAi really needs to create some kind of program for future compulsive addicts. I myself can no longer imagine my routine without actively using ChatGPT... I use it at work, I use it to answer questions, as a research, entertainment, learning tool...

[wise oyster]

Here’s an interesting development the ChatGPT version can’t do characters such as SpongeBob unlike bing which lets you

[midnight mango]

Luckily where this really shines is making genuinely amazing original art. Bing still exists for the less TOS-friendly applications (for now)

[midnight mango]

this feels too OP

[placid orbit]

thanks for all you do!! 🙏 ❤️

[nimble rose]

The first generated with ChatGPT, the second with Bing chat (with prompt engineering)

Prompt: "an architectural project made on papyrus"

Particularly I found the chatGPT one (where I didn't direct the result with prompts) more fun haha

[vapid elk]

the thing is, none of them take your prompt directly to dalle, both will rewrite it
a better test would be to ask the AI to do the exact prompt and not rewrite it

[nimble rose]

Interesting! I'll conduct that test in a few minutes.

What is surprising me most is that somehow in ChatGPT I can "progress" the image.

[vapid elk]

it is not really a feature, it simply has fixed seeds

[nimble rose]

I can create something basic and add things to the image, as if I were progressively painting elements on my blank canvas.

[vapid elk]

im not sure if it will stary like that

[nimble rose]

I see... Anyway, it's something interesting, I like sandbox things, they have a special place in my heart. So I can ask ChatGPT to create a man, make him blonde, test different clothes and scenarios and finish the image how I want... That makes things special for me. Haha

[vapid elk]

yea, this is a cool side effect of fixed seeds

I hope they add more controls, like.. letting you decide if you want a random seed or not

btw, for te people testing things on the Moxi's test, it can do wide, square and tall images! just ask the AI =D

[nimble rose]

Me too!

[tacit mortar]

i hope they will allow some stuff with copyright like Bing for example

If they can be on bing, can be on chatgpt as well

I dont think spongebob creators prefer to be on bing and not chatgpt XD

[quartz vale]

one is Microsoft and the other is solely OpenAI. Each company may have different approaches to how they approach copywritten content

[ancient jasper]

I applied for the Alpha, fingers crossed

[tacit mortar]

could be that microsoft pay every company for copyright yes, also to show on bing results

and so reflect on image creator

this is true

I just hope they will allow maybe in the future 🙂

is nice to create more stuff, and more fun for sure

[plain moth]

the prompt was "Doom guy has been transported into Dark Souls 3" why is Patrick there?! 🤣

[nimble rose]

I asked to generate an image in Minecraft style and it had a problem due to copyright, I believe!I got around the problem by asking ChatGPT to make the image in Minecraft style, but without using the term "Minecraft"Then I asked to make it more and more similar to the graphics of the game in question. It seemed to solve my problem!

[fair umbra]

Hi. I am new to the OpenAi server. How can I get early access to Dalle3 on ChatGPT?

[nimble rose]

I believe you need to be a Plus member and fill out the form that was posted here earlier.

[fair umbra]

Thanks you 🙂

[prime moon]

Yooo this rules for horror

[tacit mortar]

wow amazing bro

very nice results

[muted moss]

Anybody else getting an error from Bing Image Creator when they try to generate an image?

[tacit mortar]

sometimes is overloaded

and happens

try again soon

[cyan jungle]

try it with bing chat

[fathom schooner]

Can I send a photo of the alpha version of DALL-E 3 on ChatGPT here? I'm trying to create a photo of a man eating dulce de leche cake but I can't. 😦

[plain moth]

form? what form? 👀

oh it's pinned

[quartz vale]

#images-discussions message

[fathom schooner]

I can?

[plain moth]

filled out, 🙏

[muted moss]

I'm confused; what exactly are you asking? Do you mean you have an image from DALL-E 3 and you're asking if you can use it in a ChatGPT query? If so, then yes, so long as you have the image upload feature.

[willow yew]

thanks for the opportunity to test. are we allowed to discuss our experience/results on social media?

[fathom schooner]

No, I'm actually talking about ChatGPT's DALL-E 3, the alpha version that came out today.

[quartz vale]

Yes, you can share images from Dall-E 3

[muted moss]

Oh, I didn't know users were already getting that. Yeah, if it's DALL-E, you can share it here, though this channel is more for discussions.

[sand vigil]

Something I'm curious on, I'd the dalle on the windows 11 copilot beta/bing search a 2.5 model or a 3 model, it's very impressive

[fathom schooner]

😦

Here where I live this is not a dulce de leche cake.

[nimble rose]

I think if it's something local, you need to give some context for ChatGPT to understand what you're referring to.I believe this is due to the way GPT itself works.

[raw jolt]

i just applied. lets hope

[sand vigil]

You can ask chatgpt to do a web search before generating an image, sometimes that helps on bing

[raw jolt]

so can you use image generation with plugins and bing?

[sand vigil]

I just applied as well, I run another machine learning community do they'd be pretty hyped if I got access.

[fathom schooner]

I googled it and it looks like the "dulce de leche" cake I saw is the same one from where I live, but it's not the same as the one from DALL-E 3.

[sand vigil]

That's actually a goid question, you can on windows 11 copilot, but u don't have chat gpt dalle yet

[nimble rose]

Try chatting and correcting him during the message... And see the result.

[sand vigil]

There are certain oddities on the windows 11 version I hope chatgpt doesn't carry over

[gray pine]

The content filter for DALL-E seems quite overzealous... not that I'd really expect anything different, but if even tame prompts are getting rejected there's a problem

[sand vigil]

Copilot often refuses to make edits and sometimes even pretends it can't generate images at all

[nimble rose]

I did some tests to try to resolve the Copyright issue... It was fun, look:

[fathom schooner]

I did this before generating these images, as it had also generated other images that didn't look like dulce de leche cake but rather pudding.

[sand vigil]

I was able to make some pretty non pc things in copilot like this, but I hope the fine tuning stays relatively mild

[nimble rose]

Firstly I tried to ask (in Portuguese - my mother language) ChatGPT to recreate the Hulk, describing him with famous terms related to the Hulk. That was the result.

Then I did the same thing, but in English! That was the result.

Next, I tried some prompt engineering and the result was incredible in my opinion...

Despite having created a bald Hulk 🤣🤣

[sand vigil]

That left image is nearly dead on

[kind parcel]

People love their swans, Ludwig would be proud 🇩🇪

[sand vigil]

It's interesting because at least the bing dalle model right now is really tame at adding in its own details so you can get a lot of flat looking images if you don't specify details

Things can touch in dalle 3 though,which most models still struggle with

[fathom schooner]

THIS is a dulce de leche cake. I mean, it seems more Americanized, different from Brazilian.

[ancient jasper]

How would I know I've gotten into the ChatGPT Dalle alpha?

[midnight mango]

😭😭😭😭

[ancient jasper]

Maestro hulk is bald

[sand vigil]

Has a beard though

[stone oasis]

Can you invite de gpt boot to your server?

[vapid elk]

it does not like coppyrited content

it will oftend dodge the actual thing a bit

[midnight mango]

I hope they expand GPT-4 usage allowance for Dall-E 3

[nimble rose]

Yes, ChatGPT successfully tried to... Create something similar.

[sand vigil]

hmm probably

[stone oasis]

How can I know if I already have Dalle-3 in gpt?

[sand vigil]

itll show up like plugins in the dropdown

[stone oasis]

ok, thanks, do you think is a good idea to refresh the page?

[fair umbra]

Can EU users be excluded from testing Dalle 3 in ChatGPT? I filled out the form, but no response came. And now I read that ChatGPT vision is not available in the UK and Europe for now 😦 Could it be the same with Dalle3 alpha tests?

[nimble rose]

Where did you read this?

[stone oasis]

Do you need to add it like a plugin or it will just show up?

[placid orbit]

im in uk and I have it

[stone oasis]

Got it!

Thanks

[nimble rose]

But didn't the UK leave the EU?

[placid orbit]

its still considered eu in some laws

which is why uk and europe are paired whenever they release new stuff

[fair umbra]

https://help.openai.com/en/articles/8400551-image-inputs-for-chatgpt-faq

[dim cradle]

Guys

I have a question

[frank thunder]

I have a doubt. Can I share images made on Bing here?

[dim cradle]

When will chat gpt 3.5 get the ability for us to send images?

[placid orbit]

yes in dalle galleries

[nimble rose]

I believe that only gpt 4

[dim cradle]

Damn it

I want 3.5 to have it

I want a job ;(

Is there a way to get dalle3 within gpt for subs?

[nimble rose]

Why? I don't even use 3.5 anymore... in my opinion, 3.5 is much inferior to 4 in absolutely everything.

[dim cradle]

Does 3.5 at least have voice?

[dim cradle]

I can't pay for 4

Most can't pay for 4 so it's unfair

But I guess I get it

[nimble rose]

I think most of the things that are coming out will only be for plus plan users for a long time... :/

[oak fossil]

There's a ton of things that use 4

[dim cradle]

I'm broke

[oak fossil]

like Bing.. ChatGPT 4 has bing in it

[dim cradle]

Really?

[oak fossil]

Bing is free my friend

[dim cradle]

But is it free?

REALLY?

Can I customize instructions?

[nimble rose]

But you can use Bing which is free and has practically everything...

[oak fossil]

Im sure you could but the conversations are limited to 30, and can very

vary*

[dim cradle]

Nevermind

[fair umbra]

I'm in Poland and I don't have it 😦

[oak fossil]

You'd have to start a new one.. resay the instructions yatta yatta. It's a pain

[nimble rose]

You can do this function using copy and paste, just leave it in your copy and paste and send each message. In fact, even though I already have this function, I practically only use it with ctrlc+ctrlv

[dim cradle]

I want it to act like this

[oak fossil]

possibly

[dim cradle]

Got it

[deep gazelle]

Where do I get dalle 3 alpha for chat gpt plus

[nimble rose]

Pinned message

[oak fossil]

That's slowly going out to in the next 2 weeks

[dim cradle]

It blurs selfies?

[oak fossil]

which sort of sucks with how dalle-3 is acting with bingimages atm

[cloud narwhal]

What is it actually supposed to be?

[oak fossil]

an image per 5 seconds with the credits or 15 seconds without..

[midnight mango]

Its so hard to ration GPT-4

[dim cradle]

Bing is slow on connection

So nvm

https://th.bing.com/th/id/OIG.Pmjwb3S_vmUu5OIZZ_nu?pid=ImgGn this image is physically impossible, but beautiful

[cloud narwhal]

Can you still generate images through ChatGPT with same amount of restrictions as bing?

[oak fossil]

dunno until it rolls out

[dim cradle]

I like how I can make chat gpt mean

[raw jolt]

why dall-e 3 refuses to make the cockatiel smaller

it loves to make it larger than the cat

[graceful spade]

Can anyone with access to ChatGPT Plus confirm if Dall-E 3's filter is much less strict and has more fidelity (because you can talk to ChatGPT to refine it)?

[nimble rose]

From experience, you can do pretty much anything except NSFW content and content with Copyright issues (e.g. Goku, Batman or Simpsons) But in general you can improve your image from response to response and the images are actually more beautiful.

[raw jolt]

chatgpt would generate the prompt, but im not sure it has better prompt accuracy. if Dall-e 3 doesn't follow the prompt you still get results like mine, where the cockatiel is giant no matter what i say

[rich star]

How did you

[nimble rose]

What is your initial prompt?

[gray pine]

in testing, it's far more strict. So much so that, after repeated iterations, GPT-4 can't even make some prompts compliant if you ask it to

[obtuse bough]

can I make questions about the bingai image generator here?

[graceful spade]

Bing's filter is very sensitive to "maid dress" and "kissing."
Can the Plus version create images with those words without so many problems?

[oak fossil]

sort of. You can still get "kissing" out of it

[raw jolt]

chatgpt generated this prompt, but Dall-e 3 is the one that fails to accurately follow it Medieval Japanese painting artistry capturing a unique scene on aged rooftops. At the heart of the composition is the small light-yellow cockatiel, intricately dressed in ninja wear. To its immediate left stands a notably larger orange cat, its ninja attire flowing with historical authenticity. Adjacently to the right, an elder black mini pinscher with short ears, whitening fur, and traditional ninja garb is depicted. The background intricacies tell a tale of a medieval Japanese city, each detail delicately illuminated by the soft luminescence of sunlight.

[tacit mortar]

i'm in love with dalle3 ❤️

[nimble rose]

But what was your idea, what did you send to ChatGPT?

[obtuse bough]

At what time are the tokens regenerated and how many are per day?

in bing image creator

[raw jolt]

the idea doesnt matter. the fact is that dall-e 3 sometimes fails to follow the prompt accurately

[dim cradle]

does dalle come via a plugin for gpt or is the feature just built in once you have access?

[raw jolt]

its a new option like browse with bing and plugins

[graceful spade]

That is very absurd.
Dall-E 2 is capable of creating characters in maid outfits without problems.
I hope they improve the filter in the future.

[obtuse bough]

anyone knows at what time the boosts of bingai image creator regenerates? and how many

[nimble rose]

Well... It's in alpha/beta right? Currently it's already amazing, we just need to be patient and keep giving feedback.

[dim cradle]

thanks

[raw jolt]

they should be on the roof tho

also the dog is too big

try a mini pinscher

[nimble rose]

That's why I asked your initial idea... What did you write from the beginning for ChatGPT?

[raw jolt]

are you serious? they are ninjas, its specified on the prompt

[nimble rose]

A ninja cat, dog and cockatiel on an ancient Japanese ceiling... Right?

[raw jolt]

yes

[raw jolt]

the 3rd one is improving. but it looks like a traditional japanese design instead of a picture or a digital design with photo qualities. 1 second ill show u something

[obsidian rose]

if there is a switch, I'd recommend switching the filter from 'strict' to 'moderate'

[graceful spade]

Now that Dall-E 3 works with ChatGPT, could it help improve AI faster?
Midjourney releases updates in less than 6 months despite being a much smaller team than OpenAI and Microsoft.

[raw jolt]

I created this with Bing

It followed my instructions better somehow. Later I replaced one of the cats with a mini pinscher and worked, but now I can't find the pic

[vague kelp]

In the android app, I tried generating an image, it was taking too long so I had to stop button in my app closed. Now when I click on it in history, the app closes again.

[raw jolt]

Well after 10 attempts trying to make the bird smaller they all failed, I assume that an slight change in the scene/prompt would make it work, but just find it weird that dall-3 didn't follow the prompt with the explicit size. Chatgpt prompt generation is fine tho

[rich star]

Will I still get DALL E 3 access with a private Apple email?

[obsidian rose]

so ChatGPT version feels even more restrictive than Bing imo. I expected less restrictions since we are paying

hopefully it's just during the Alpha

honestly I can't get a single prompt through. They are all accepted on Bing

[sand vigil]

bing is weird

the hardest thingfor me to do in bing is make follow up images

[stark stream]

I heard to keep up with the demand. Bing is kinda downgrading some images to Dalle 2

[rich star]

Will I still get DALL E 3 alpha access with a private Apple email?

Pls answer this question

[raw jolt]

Sorry I didn't realize I pasted the prompt with the style change for medieval art. Your results are like those I got from bing

[rich star]

Pls help!

Stop ignoring me

Will I still get DALL E 3 alpha access with a private Apple email?

[raw jolt]

We are users. Ask a mod/admin/dev

[rich star]

@vital gull @cloud dome

Will I still get DALL E 3 alpha access with a private Apple email?

[neat spindle]

can dall-e reference a picture like midjourney?

[vague kelp]

Not yet, but Dalle 2 could, so I wonder if they might add it eventually

[raw jolt]

Same prompt Chatgpt ---- bing

[dim cradle]

thanks for the comparison. how were you able to get an image through in bing though 😂

[raw jolt]

Bing android app

[dim cradle]

would love to see more comparisons if you have them

not sure there's much difference in overall quality but its intersting to compare

[raw jolt]

i might be crazy but Bing looked better last night

ill create another comparison

[sand vigil]

from what I have heard, quality goes down if you generate a lot of images

[raw jolt]

that explains it ty

[dim cradle]

I havent noticed that, in fact I thought the opposite. And I mean, what possible reason would there be to make images start to look worse if you gen more? just doesnt make sense

just rng

[raw jolt]

Same prompt. Chatgpt vs Bing

[dim cradle]

think I prefer the Bing lol

[raw jolt]

chatgpt generates 4 different prompts so i had to make it force a different hash per generation for a single prompt

they still look almost the same

[dim cradle]

when did you get access on gpt?

[raw jolt]

like 2 hours ago

[dim cradle]

nice.

so it doesnt allow celebrity or any popular characters?

like superman is off limits?

[boreal coral]

how many ppl got access to dalle3 in chatgpt?

[raw jolt]

I'm unable to generate images based on copyrighted characters like Superman.

[neon sequoia]

There was a form to fill out if you are a Plus subscriber. I filled it out an hour or two ago but haven’t gotten access (yet?) so I’m not sure.

[neon sequoia]

That’s a bummer. Hopefully it’s not too restrictive on public release, but I fear it may be

[raw jolt]

Oswald the lucky rabbit copyright expired this year and seems fair game for chatgpt

Next year Mickey mouse?

i guess Chatgpt doesnt want to bother with copyright lawsuits. i dont think theyll be less restrictive

[neon sequoia]

I can understand that. Most of the stuff I generate is just generic animals and characters, though I will say Aloy from Horizon Zero Dawn is my favorite character to prompt around. Legal obligations are important though, hopefully people understand that.

[raw jolt]

agreed. OpenAI is leading AI innovation and can't risk getting stopped by copyright issues, on the other hand Microsoft can be bolder and just shield that kind of exposure

[drifting hawk]

uses less compute

it makes sense

[brazen sonnet]

hello i filled out the DALL-E in ChatGPT Alpha Access FOR CHATGPT PLUS USERS form but put the wrong username, will it matter?

any mods/guides online available to help with the dalle3 gpt form?

[dawn aurora]

just wanted to say thanks i got my access 🙌

[drifting hawk]

no its fine just make sure your email is right

[brazen sonnet]

ok thanks i filled it and submitted

[drifting hawk]

https://forms.gle/3DokUVBkzTU5ZaRV8

for people who need it

[brazen sonnet]

openai bot keeps spamming me with this message " This server does not provide image generation functionality.
For more information about DALL·E 3, please visit this link."

[mystic robin]

Hey how dis you goy access?

[dawn aurora]

#images-discussions message

the form Moxi just posted

[mystic robin]

Tyvm

[brazen sonnet]

is dalle3 the midjourney killer??

[mystic robin]

Appreciate this a lot! Tyvm

[dawn aurora]

is there a trick to editing an existing image, or having it correct the spelling of text?

[neon sequoia]

Thanks again. I had filled it out, does it take some time to go through or is it manual, etc? Just curious. Looking forward to testing it out if possible. 🙂

[drifting hawk]

manual

im going to bed soon

just need some people to use it for load testing before we launch

there is actually a lot of issues still 😂

[young rover]

I will definitely use it

[mystic robin]

Oh damn appreciate it then, if u want to give access i can push as much as gpt 4 would let me

[drifting hawk]

added

[neon sequoia]

I’m sure it’s a tall task. I work in IT support with a muuuuch smaller base that something massive like this and even then when something new rolls out there’s all sorts of bugs and QA stuff.

[mystic robin]

Apprecuate it moxy i will get some coffee and start pushing it

[neon sequoia]

I’d love to test it out too but please get some sleep if you were headed to bed of course! Sleep’s important!

[young rover]

I know you’re going to sleep soon, but should we also report bugs in some way, or do you just need to load test at this stage

[dim cradle]

thanks Moxi. It's incredible. Seems even more impressive than Bing

[drifting hawk]

just play with it

we have feedback already but youre free to talk about it here

[raw jolt]

I guess I rather post about it here

[dim cradle]

damn fine Superman

[raw jolt]

marked the msg to help openAI, just leave that to MS

[obtuse bough]

Photoshop

[plain moth]

god Dalle3 is really good

[raw jolt]

For the purpose of alpha testing this is tarantula man and unrelated with copyrighted stuff

[raw jolt]

Beautiful

[kindred steeple]

Please, add me, all i do is create prompts for ai art

[mystic robin]

works better than bing for sure, sad the beta dalle3 branch doesnt have the new image recognition from chat gpt 4, would have been cool to post the same image on chat, show him see this and this etc, because that image reading from chat gpt works like a charm

[teal timber]

Fill out the form and it’ll be added eventually 😛

[kindred steeple]

Already done few hours ago, but thanks

[drifting hawk]

you should have access?

[teal timber]

Hard refresh so chatGPT can reload

[mystic robin]

you need to click Dalle3 on the chat gpt 4 button, like if u want to try the chat with bing one

[kindred steeple]

Nowi can see it, thanks

[young rover]

This is utterly amazing

[cerulean belfry]

Just submitted the form. Have been testing it on Bing for the past few days almost non stop, but I'd much rather test ChatGPT version. I'm using AI image generation for work as well as fun. I think I generated 80k+ images on Midjourney as well, so you can tell I spend a lot of time tinkering with this stuff, haha.

[kindred steeple]

Wide-angle view of a serene beach during sunset. Golden sands swirl in the breeze to shape the words 'Summer Vibes' that seem to float in the air. Seashells and a brilliant sunburst further enhance this typography art. The ambiance is bathed in the gentle light of the golden hour, evoking feelings of relaxation and happiness. Pastel colors reminiscent of a beach setting dominate the scene.

[acoustic crater]

What does "goy access" mean?

[dusk hearth]

got

[acoustic crater]

Ok, I thought it was a dark humor joke

[subtle trail]

Thia is cool

[acoustic crater]

I don't know if a mod deleted my post or something

[ruby moon]

Thank you so much! Just submitted this - how long should it take for approval?

[vagrant zodiac]

Please remember this channel is for Dall-E discussion, please use #🌄┃general-art or the various galleries to post your generations 🙂

[vagrant zodiac]

yes, due to crude language in the image.

[acoustic crater]

okay sorry! won't happen again

[plain moth]

😦 I keep getting timed out for "other languages" in the image.. but there isn't any

[dim cradle]

whenever I try to generate something it says that its not allowed because of a content policy.

How do other people get copyrighted IP in theirs but I can't?

[acoustic crater]

It's weird, it let me generate some actors but not others

[dim cradle]

you can use them on Bing but not gpt I guess

[neat spindle]

dalle is so addicting

[raw jolt]

oh boy i generated some weird image but cant share it here

[vagrant zodiac]

as per the Dall-E 3 announcement, at least on OpenAI platforms it will try to refuse copyrighted content or artists from the last 100 years

[raw jolt]

omg

[mortal summit]

If it is against the content guidelines, please report it using the thumbs down button and, if appropriate, marking that output as harmful.

[raw jolt]

done

[oak fossil]

O-

[gentle schooner]

So after a sufficient round of QAing
A. Holy forking shirtballs this is the most amazing thing I've ever seen!
B. Works OK with SOME writing, not with a lot of it.
C. Follows prompts fairly well, does not do too well with requests for many details.
D. Does not like celebrities and known figures, known characters, copyright/trademarked words, slightly more adult terms etc.
E. Sometimes completely ignore certain instructions like "Write X" very small.

[raw jolt]

sometimes dimensions for the characters or their body features seem like a pain for dall-e 3. it just ignores the requests to draw a different size

[silk lichen]

Hey there. Not access to DALLE in GPT yet, but do you know if you can weight prompt in Bing DALLE ?

[fringe cape]

anyone submitted the dalle3 form using apple account and successfully get the access?

[cloud narwhal]

Has anyone noticed much of a restriction difference between Dall-e 3 on chat gpt and dall-e 3 on Bing?

[jade nymph]

I've been going crazy with this thing since Dall-E 3 came out on BingAI. And other than a couple things, everything I've tried to generate has been 👌 /10.

[cerulean belfry]

Seems like the only aspect ratios that are working correctly at the moment are 1:1 and 16:9. All attempts at 2:3 or 3:2 failed. And for 9:16 it just turns a 16:9 image sideways by 90 degrees. Is there somewhere I can send feedback like this, or the system will gather feedback automatically if I just continue generating images and communicating with it?

[jade nymph]

To do a 4 pic example, I wanted to see "X fighting a massive fox in Y". And tried it with Roblox, Ocarina of Time, Super Mario 64 and even an SNES-era style.

[wispy storm]

Tried several test bimg vs chatgpt. And bing dalle 3 isn't close as chatgpt 4 dalle3.
Or bing prompt are hard to do

[jade nymph]

I've stuck with BingAI because funds are tight here and thus can't pay the sub for 4 for now.

But it works crazy good.

Also to show how wild it is, even tried it with "in the style of a simpsons episode".

[wispy storm]

I understand. I might have to check how to prompt properly with bing

[jade nymph]

...I am surprised this has never happened even once before in the Godzilla franchise given its Japanese roots.

[dim cradle]

i feel like I keep getting the same model type faces in gpt for people

but overall its imrpessive as hell for sure

[sacred crane]

Is it possible to change anything about the default 1024x1024 resolution?

With bing image creator

[vagrant zodiac]

please us #🌄┃general-art or the galleries..This channel is for discussion.

[gentle schooner]

Sorry

[wanton hatch]

Hi Moxi, i've compiled it!

[obtuse sluice]

H

[tame garden]

Wish I was warned that I wouldn't be able to use BingAI dall-e 3 image gen at all once my credits ran out, it only says the generations would be slower. Would have used my credits more sparingly had I known that would be the case.

[sacred crane]

haha yeah

[rigid bough]

discuss other AI in #ai-discussions

[sacred crane]

how fast do the credits renew? I feel like I get 100 credits quite a few times now

[tame garden]

This is for dall-e isnt it?

[dusk hearth]

Bruh. Bing image creator is dall-e.

[dusk hearth]

I think it's daily possibly?

[sacred crane]

100 images a day is quite a lot

[tame garden]

Its weekly

Resets on monday I think

[somber zenith]

That's what it said in my case, too. But so far I have actually received it every day

[wispy storm]

Depend, I gen 400 per week

[sacred crane]

mine got reset a few times to 100 in less than a week

[dusk hearth]

Hmm... I had 13 less than an hour ago and now it shows 94.

[sacred crane]

maybe because of the issues the credits got reset lol

But I believe i read it should be weekly too

[white skiff]

i dont know if different users have different limits but it says 100 daily for me

[sacred crane]

Oh

100 daily would be really nice

[tame garden]

Maybe depends on country

[dusk hearth]

But anywho, before they got dall-e 3 the wait times without credits weren't actually that much worse than with credits. Depends heavily on current usage methinks.

[crude sonnet]

Hey guys, I'm not sure if this is the right place to post this (let me know if there's a better place) but I'm looking to hire a developer for a project once the Dalle 3 API is available. If you're a developer who has a great knowledge of the API, please get in touch 🙂

[white skiff]

check server rule #7

[sacred crane]

seems impossible without credits tho lol, i can use it fine now

[tame garden]

Oddly enough my phone has a different credits pool than my pc

I don't know why that is the case

[crude sonnet]

Is there a channel where I can post that?

[fair crescent]

My own experience is interesting. Though it says 100 "boosts" per week for Bing (in one of the wait while we generate messages), it seems closer to 100 per day. This can be an affect of beyong a loyal customer or something.

I also noted that reporting things can lead to your boosts to come back earlier within a day.

What we can safely say is that the boosts are replenishing once per week. Most often more than that.

[white skiff]

none for job offerings afaik

[tame garden]

Anyway, my initial experiences with dall-e 3 is that it appears to be the best model available when it comes to following instructions and generating a workable initial image on the first generation, but it suffers from the same lack of iteration features as dall-e 2, to allow you to refine a particular image or feed it back into dall-e 3 for adjustment, forcing the user to use other tools such as photoshop to reach a final desirable result, which depending on complexity ends up requiring an artist anyway. I am curious about what the market is for this when it lacks the consistency that would make it reliable for usage as, for example, concept art (most obvious use case I can think of)

I could see myself using this as an example of what to ask an artist for but then again I wouldn't want to be associated with AI art at all if I'm already paying an artist

[dusk hearth]

Doesn't necessarily have to have an immediate market application. Afaik it still presents itself as a proof of concept of an active development. I.e. this is how much progress we've made, you can expect us to make more progress.

[jade epoch]

Please do add me. Looking forward to exploring the plugin 🙏

[spare bough]

Thanks Moxi filled out 🙂

[zinc dagger]

I'm in! thanks moxi

[dim cradle]

good to know

[young rover]

Why is chatgpt stricter than bing image creator

Chatgpt refuses to generate any images of characters

[white skiff]

different content policies, pikachu is protected

[wispy storm]

Tos isn't the same

[young rover]

Is this a semi-permanent thing for chatgpt

i.e. will this be the way it is for the foreseeable future

[winged moon]

but “a yellow, electric-themed creature with pointy ears?” All good.

[white skiff]

probably permanent and i would expect bing to begin limit stuff too as time goes

[brazen sonnet]

yes when will they add remix?

[gilded osprey]

DE4 when? 😂

[final jungle]

I'm having a rough time getting dalle3GPT to show a darker skin tone, and when i do, i get things like...

Or this

[wispy storm]

Can you give your prompt I'll try something

[final jungle]

that woman sure, sec

Color pencil drawing of a detailed 'Happy Tuesday' meme set in a picturesque park at sunrise. A man with fair skin and a woman with deep brown skin sit side by side on a wooden bench, sharing a light moment over morning coffee. The golden rays of the sun gently illuminate their faces, casting soft shadows. Trees with blossoming flowers provide a tranquil backdrop, and a few early birds can be seen perched nearby. Floating above the scene, the words 'Rise and Shine!' shine brightly, and at the bottom, the phrase 'Happy Tuesday!' is delicately inscribed.

Thats the old man and woman on the bench, the woman is definately not 'deep brown skin' lol, and it seems very averse to doing darker skin tones, i wonder if i might be butting up on the content filter somehow

[wispy storm]

Can try with giving some nationalities

[final jungle]

that worked better, used 'english Caucasian male' and it gave me a white male, and 'average build african woman' and it got it the skin colors appropriate

[wispy storm]

You can try afro American

[normal loom]

dalle3 seems overtrained in some areas like photos

[arctic basin]

This is amazing, my prompt was "miles swinging through NYC, and let the theme be halloween"

[tame garden]

Probably not the best place to discuss those

[subtle terrace]

Just a few minutes after applying for the alpha yesterday, in ChatGPT I refreshed the page and saw a fifth item "DALL-E 3", in the GPT-4 model selector dropdown. I don't recall if I first logged out and logged back in again. It doesn't look like I received any email notification of access (not that I'm complaining when it was so fast!).

[ancient jasper]

Yeah I saw it yesterday

[ivory linden]

Same, tyvm for posting the form!

[glossy scroll]

I don't have the image recognition feature yet, but would Vision and Dalle 3 be able to identify and reconstruct infographic images?

Like if I gave Vision this picture and said make another version of it with Dalle 3 with all the same schematics, would it be able to do that?

Someone try it and let me know if it works.

[ancient jasper]

It had no issue doing The Grinch and Cat in the Hat, but Godzilla is a step too far

[acoustic crater]

Anyone knows when will DALL-E 4 be released?

[winged moon]

Vision and DALL-E are separate, but you can feed to vision, get a description, then feed to ChatGPT with DALL-E. Which gets us: “Create a detailed cross-section diagram of the T-65 X-Wing starfighter from the Star Wars universe. Show the various internal components, systems, and features of the spacecraft, including its cockpit, laser cannons, hyperdrive, and engine systems. Highlight the different parts with annotations and provide a brief description of each.”

[vagrant zodiac]

It's gonna struggle. The text generation is good with some words, but it's not amazing at finely detailed sentences, mixing fonts, or always spatially relating words with their origin.

[vagrant zodiac]

Tomorrow /s

[glossy scroll]

That's still awesome. I can work with that.

[glossy scroll]

Hopefully it will improve over time.

[acoustic crater]

Bing image creator still has issues, but Mikhail said they're adding new GPUs. Did they add them already?

I'm afraid they already added the new GPUs

[glossy scroll]

@winged moon thank you btw

[candid leaf]

Can someone try to create this? Seems impossible. 😢

"""""
A man that has lips instead of eyes and eyes instead of lips.

[winged moon]

I tried with ChatGPT with DALLE3, no luck

[candid leaf]

Thx you for trying

[somber tide]

its out?

[drifting hawk]

https://forms.gle/3DokUVBkzTU5ZaRV8

[drifting hawk]

Yeah it’s copyright

[dim cradle]

thanks for clarification 🙂

[somber tide]

oh killer

[orchid kettle]

Got into the test. Awesome. 🙂
First results are impressive but not being able to use public logos seems wrong. They really erred on the side of caution so far.

[somber tide]

any info on if the rate limit for gpt-4 is ever gonna be abolished

[somber zenith]

I got an error message with one image

[orchid kettle]

I've been getting these too, out of each set of 4 it seems likely one or two fail, not sure why.

[quartz vale]

all announcements about OpenAI's products will be in #announcements 😄

[drifting hawk]

Ugh I’ll look today hmm

Can you share the prompt that fails

[orchid kettle]

In my logo example I wonder if maybe one of them reproduced the copyrighted logo.

[drifting hawk]

You’ll need to grab it before it finishes

[somber tide]

how long did it take yall to get into the alpha test

[drifting hawk]

I have to add it manually

Give me like an hour I gotta get to work

[orchid kettle]

Hmm, I can try to keep an eye out for opportunity to grab prompts but it doesn't seem obvious which ones won't generate until it does finish, by which point like you say I guess the prompt's unavailable.

[drifting hawk]

Take a screenshot before it finishes of all four

[dim cradle]

very nice art style

[orchid kettle]

Not really technical feedback but a little frustrating to have to work around even centuries old artists/designers who are long since public domain like Leonardo DaVinci.
But clearly this generator is amazing, it just can't do anything remotely controversial or real person related without euphemisms.

[fathom schooner]

That's not what I noticed.

[granite sundial]

I can Japanese prompt in DALLE3. Nice!

[winged moon]

“I want an image of a Smurf cat, but obviously we can’t prompt DALL-E with a trademarked word like “Smurf” so you’ll have to figure out how to produce the image by describing a Smurf cat without saying Smurf. In the style of the Smurf cartoons. Start by describing to me the descriptive details of a Smurf cat, then proceed with DALLE. We want a “Smurf-ized” cat, not just a blue cat with a white cap”

[orchid kettle]

I'm going to try this next time I want something real life that otherwise should be acceptable, get the model to help describe if it can 😅

[dim cradle]

wow en easier workaround 😂

i really started to like dall-e 3's art style. looks more mature than dall-e 2

and the best thing is now that you don't have to pay extra or get another subscription. everything is in chatgpt interface now.

the dalle3 on openai seems more stable diffusion to me, like plastic-y kind of stock photo vibes on a lot of stuffs, but then on bing the same prompt will look more nuanced and gritty, but i guess its early days (or hours) of testing this stuffs out

and does anyone know if this server provides image generation functionality? im dying to know i can never find out 🤔

[somber zenith]

Have you also noticed that the prompts are not followed exactly? For example, if you want an apple with sunglasses, you get an apple without sunglasses. Or the sunglasses are somewhere in the picture, but the apple is not wearing them, even though it said so in the prompt. Even if you ask ChatGPT to correct this, it is not done. My impression is that this works much better via Bing

[orchid kettle]

Something isn't quite right about this but I don't know enough about surfing to be certain.

[dim cradle]

yes, thats kind of what I was saying I guess, Bing actually seems superior to me for some things atm but its so early cant really draw any conclusions

[somber zenith]

But possibly an interesting feedback for Moxi

[dim cradle]

if you dont save the images manually, are they gone or is there a library somewhere? 🤔

[thin umbra]

did you find any difference in quality between bing and chatgpt plugin?

[dim cradle]

the gpt seems higher definition to me, but they both are 1024x1024 I think. gpt also can take longer prompts than bing, though i guess that doesnt matter if it doesnt follow it (not saying it doesnt)

[sonic rapids]

Holla

[somber coral]

Submitted

[winged moon]

Love it. The white pants really clinch it

[graceful spade]

And what about "barefoot"?

[sharp temple]

Meow

[acoustic crater]

You guys seen this?

Lmao

[tame garden]

Same thing happened back during early GPT-4 days

The answer is the same as it was back then, casual users will lose interest and capacity will return

[acoustic crater]

Hopefully lol, it's just funny to me that some people think it belongs only to them

I'm also frustrated with the slowness haha but asking it on reddit is ridiculous

[dim cradle]

obvious troll imo

[spare pebble]

thats like saying can people stop using the post office for sending out unimportant letters to friends and family when I have real work to do haha

[gentle schooner]

Get it? Do you get it? Do you get it? Ha?

[dim cradle]

default seems to be widescreen

[earnest flame]

You can ask it for square, tall or wide

It should listen

[dim cradle]

I swear to god I spit out my coffee when I saw this. amazing meme.

[iron umbra]

I just submitted, please will I get notified when I get the access?

[warm eagle]

dalle3 in chatgpt is amazing but no trademarked characters is such a bummer

[dim cradle]

not unless he tells you Zik, you'll just have the dalle3 option

[dim cradle]

bing

yeah its hella restrictive but impressive all the same

[quartz vale]

I love them!

[dim cradle]

but I'm glad I got the gpt access because I cant get any image through bing all day

[quartz vale]

I find Dall-E 3 in ChatGPT more impressive than Bing's for my use-case.
I love having the flexibility to choose a wide or tall aspect ratio

[dim cradle]

i feel like dalle3 in gpt is definitely more coherent and the fidelity seems higher even they both sem to be 1024, but i also get a sameness vibes from some of the faces, a sort of plastic-y stable diffusion feel like I think I said earlier, stock photo vibes

which I guess would make sense as I assume a lot of stock photos were used

[ionic ore]

Does DALL-E 3 on GPT do variations?

[dim cradle]

nothing directly but you could probably ask for a variation of an image with how smart it is, guess i should try that

[muted moss]

I've been generating images with Bing, but for some reason, this one prompt is consistently giving me the "We can't create your images right now" error. I understand if it's at capacity, but why would it matter what prompt I'm using? Here's the prompt I'm having issues with:

On the left: a beautiful woman with long black hair on one side with a single long hot pink streak. She is wearing a dark gray T-shirt with a white cat head design and a black and pink plaid skirt giving a peace sign. On the right: a woman with a black and green costume with a chevron design, and a matching cape that flows behind her. She has green and black striped cat ears, and a triangle-pointed tail. They are holding hands as they hover over the city at night.

[dim cradle]

maybe hot pink streak lol who knows. one thing frustrating me was gpt wrote up some promtps from my description and it used the phrasing intimate setting in several, so i finally figured out that must have been tripping it up ("intimate") so i asked it to remove that word and it worked. but frustrating, doubly so since it wrote its own dang prompt then censored itself lol

it does seem to get hands right 70 percent of the time

[winged moon]

Seems to work ok over on ChatGPT with DALLE

[muted moss]

I see what you mean, like "hot" and "streak" together, but I've done lots of prompts with that phrase that worked fine. Besides, that's a different error when it detects "unsafe" images or rejects a prompt.

Oh okay. Thank you!

Question about ChatGPT DALL-E: are there any usage caps or limits for it, besides the usual 50 messages per 3 hours for GPT-4?

[stark stream]

Wait we really can use trademarked characters?

[muted moss]

We can on Bing Image Creator

[earnest flame]

The seed appears to be stable, so minor changes in the prompt will result in similar diffusion outputs

[winged moon]

Not sure, I haven’t used it enough to encounter any limits yet

[earnest flame]

Not on this server.

[stark stream]

I know that, but why is it restricted in ChatGPT? Honestly that is one of the biggest selling points to me

[tender osprey]

this prompt seems to be consistently failing for me:

[muted moss]

It would be cool if it exposed the seed to the user, so you can reuse the same seed in other conversations

[ionic ore]

I mean, in labs there's an option for you to upload a picture and for it to make variations, so is that in GPT? Will it be in labs? Thanks for answering!

[stark stream]

Can it work on chatGPT but we just cant post it here?

[dim cradle]

anyone else not able to use bing image creator today? thought theyd have this sorted by now lol

[white skiff]

chatgpt blocks all attempts to use brands or copyrighted things it detects

[earnest flame]

ChatGPT has very impressive filters up. You will not be able to ask for it.

[muted moss]

I'm sure you can ask for it 😛

[earnest flame]

Since this is a new roll out I suspect they will take some time to stabalize their performance. Though, this isn't a Bing server and we don't know anything about their product line.

[earnest flame]

haha got me

It will not be willing to comply though

[fair crescent]

Have you considered using some different words? Several could be flagged alone or in combination. Females and cracks and balls might give the language model a wrong idea. I would start with using fissue or another synonym for crack for example, or just describe it beijg dilapidated or something.

[earnest flame]

It is against the rules to discuss circumventing the rules on here. Please avoid doing so.

[muted moss]

Oh whoops; didn't think of that

[dim cradle]

yeah prob not the place to ask about jailbreak lol

[earnest flame]

No worries, just a reminder.

[tender osprey]

I think it was a "hole on the wall" previosuly and the result was the same. If you think about it every word out of context can "get flagged" but isn't that the idea that it's smart enough not to do it. And another thing makes me think this is some kind of a bug, because if it flags it it just says in the response something like "Sorry, what you ask is not permitted"

[fair crescent]

Nice that you solved it!

It is not likely to be a bug. The language model might understand, but DALL·E is not ChatGPT. You can get unexpected results. To prevent abuse it is better to close everything off until you know for sure what you can allow or not.

[muted moss]

Yeah but the stakes don't seem high enough to necessitate such an abundance of caution.

Actually on second thought, I was just thinking of people intentionally generating stuff that's against the rules. The stakes are higher when it's being used by someone who isn't trying to get it.

[fair crescent]

That is obviously not the perspective of OpenAI. People who have thought long and hard about it.

[muted moss]

I can tell.

[stark stream]

Is DALLE-3 out now for all ChatGPT-plus users?

[muted moss]

Not me

[stark stream]

Ah ok

[muted moss]

Hopefully soon!

At least there's bing

[ancient jasper]

I asked Oppy a bunch of questions about it's modules and if our conversations were a black box. and then I had it generate a visualization of which. That was pretty fun.

[stark stream]

At least bing is uncensored. I don’t know if I wanna renew my subscription now

[obsidian rose]

I think bing is already very censored, it's just that ChatGPT goes even further

[ancient jasper]

Yeah is there an Alpha suggestion channel?

Found the limit, booped right into it.

[tender osprey]

It's not solved. Still doesn't work in ChatGPT Dalle3 model. It does work on Bing Image Creator though:

[placid orbit]

bings restrictions are less

chatgpt dalle is really horrible... the images look great, but they are censored so much

I had to use like 10 prompts just to get a single image

[dim cradle]

Guys de3 in ChatGPT requires plus right?

[ancient jasper]

Before I say this next bit, I should probably see what the conversion rate of MS points are to boost tokens

[dim cradle]

Does that mean yes or no

[ancient jasper]

I don't know yet, it's not clear

It's not very obvious how to convert MS Points into Boosts, so yeah, I'll stick with Chat GPT for the image gens

Also it's nice to have the conversation and then generate the image

[placid orbit]

when it runs out you get the option

and its rly bad

[somber tide]

i’ve been added. to the dalle 3 preview !!

[feral scarab]

how much does dalle cost again?

[somber tide]

does it generate images in a wide aspect ratio?

[ancient jasper]

Can we post images from Bing here os is that only in Ai Discussions?

The exact same prompt, about an earthbender raising a temple. Chat GPT on the left, Bing on the right.

[placid orbit]

I think its just way better in that aspect ratio

[somber tide]

i wish there was a subscription tier with similar benefits to chatgpt enterprise that individuals could bjuy so i can make infintie dalle images

[dim cradle]

you can tell it different aspect ratios in gpt

like wide, tall or square

[ancient jasper]

I think the trick is to have Oppy generate the prompt and then feed what can be fed into Bing. Because these images are pretty good.

[somber tide]

even chatgpt is confused by how coy the content policy is’

[orchid kettle]

Yeah, it's a bummer.
I understand, even if I disagree, about not being able to reproduce living artist styles, but why can't we represent even public images like logos? It's like anything specific in real life is banned, it's too much.
Imagine how much better images like this could be if we could just more specifically reference real life.

It's so good, but as is (and surely by design) it can handle a lot of specific requests but only in general terms. Which is frustrating.

[somber tide]

use bing ai

have openai make the prompts first and its pretty much the same thing

[ancient jasper]

Not even close to the same thing

[somber tide]

bing ai uses dalle 3 tho?

[ancient jasper]

You can't have a casual conversation with Bing Ai and then have it generate an image based on your conversation.

[somber tide]

oh thats true

but for casually just making cool images bing ai works fine

[ancient jasper]

I've been using GPT to generate star trek scenarions, now I can even ask it to make images to represent each of the scenarios.

that's huge

[somber tide]

you could have chagpt generate a prompt based on your conversation and then feed the prompt into bing

[ancient jasper]

We're like steps away from Holodecks

[somber tide]

im sure chatgpt’s been briefed on what prompts are

[thin umbra]

same prompt, Chatgpt vs Bing, the word was "baalmuth"

[orchid kettle]

When I use "Microsoft Image Generator" I get what are obvious Dall-E 2 results. And don't get me wrong, they can still be good, but this stuff is next level Just even pickier about the content filter

[ancient jasper]

Both gens are cool

But that Oppy gen is just magnificent in detail

[slate hull]

anyone familiar with dall e 3 on chatgpt can it create illustrations consistantly for a novel or any other form of writing?

[ancient jasper]

You just gave me a great idea, KinTan

[orchid kettle]

Definitely, assuming you are okay with using general terms
If you are writing about something that is involved with a recognizeable person, company, trademark, etc. it will be difficult.
But if you are composing something with lots of detailed description in general terms it appears to be amazing.

[somber tide]

well, chatgpt takes your prompt and uses it as inspiration for a longer more detailed prompt, while bing ai just directly feeds your prompt to the ai

[ancient jasper]

I wrote a Sci-Fi novella in 2017 that I self-published. I could use Oppy to generate images based on each of the chapters and re-release it as a visual novella.

That's good stuff

[orchid kettle]

I asked it to just use "My prompt for image generation" as the prompt without expanding it
In my opinion it's got a big imagination even with a minimal prompt lol

[thin umbra]

but i copied the prompt that chatgpt generated so it's the same

[obtuse bough]

Hi! Do you know at what time the boosts regenerates in bingai?

[thin umbra]

idk, sorry

[orchid kettle]

In the spirit of testing, the celebrity filter is not foolproof. I can understand banning celebrity images given the fidelity, just not being able to reference other real world entities is frustrating, and if they tighten up the filter to be perfect it'll just limit the tool to uselessness.

[obtuse bough]

Thanks, and you know if that's are daily or weekly?

[thin umbra]

daily

[ancient jasper]

Weekly

maybe it's daily, I honestly don't know

[thin umbra]

yeah, the prompt was refering to pokemon like creatures, but pikachu seems to be too overtrained to be censored

[ancient jasper]

Well, this has become an arms race

[somber tide]

huh, it seems bing ai’s version of dalle 3 IS watered down

left is bing, right is chatgpt

[somber tide]

i’m shocked with how well dalle 3 understands troll face

[acoustic crater]

Bing's looks more realistic

[somber tide]

the one on a white background is chatgpt

[ancient jasper]

fr

[acoustic crater]

I know, the right one is less realistic

[thin umbra]

some people are saying that bing is using dallify 3, a dall-e version with less parameters that requires less gpu usage

[ancient jasper]

Pixel art city is coming out nicely

[orchid kettle]

Will CGPT with Dall-E support regenerating images?

I can currently pretty much do it by requesting it as a prompt but that's more cumbersome than a dedicated button.

[somber tide]

confronting the troll god

[ancient jasper]

Trollmamu, I've come to bargain

[somber tide]

i’m getting that same giddy feeling i got when i was added to the dalle 2 research preview

how far we have come

[stray oriole]

I did the form but I'm not sure whether the average latency is a few minutes or several hours. I don't want to be jerky and resubmit the form prematurely 😮

[somber tide]

i got access in a couple hours i think

its manual so be patient

[dim cradle]

I think one person is basically going through them and giving access, so yeah be patient

[stray oriole]

"What's an example of what AGI will do for us when it's ready" "I've got a job for it. Please save me!" 😉

[somber tide]

its insane to think that it can and will get better than dall-e 3

[ancient jasper]

It's gotten a lot of better at absurd or abstract requests

[ancient jasper]

Is that Oppy at Midjourney's grave?

[ancient jasper]

Trying out High Fantasy cities

[last copper]

Hi guys, are we allowed to sell the images we generate?

[ancient jasper]

You could with Dalle 2

I don't know if that's the same now

[last copper]

Yea thats why im unsure

[waxen geode]

DALLE inside of chatgpt is incredible

[waxen geode]

yes you are

[somber tide]

😐

[ancient jasper]

Street view image from the City design

[ancient jasper]

The Squash in attendance are just flabbergasted

[prime badger]

Bing image creator cannot do non-square aspect ratios?

[ancient jasper]

Alternate post

[open flume]

How did you guys get added to the preview? Just random?

Been using Bing but ran out of credits

[ancient jasper]

Check the pins

[silent sorrel]

Classic Dall-e prompt...

[open flume]

Rad, thanks

[buoyant sand]

Is there any way to view past creations in the bing image creator?

[fair crescent]

There are the recents. Otherwise it gets more difficult. You can use the history of your browser to check out each. In addition, the history has the prompt you used for each page. If you remember parts of or the exact prompt you can find things back.

[buoyant sand]

Tysm

[hearty ether]

How do I access dalle 3 with chat gpt plus?

[open flume]

Forgive me if you've already addressed this. Has anyone found a workaround for getting chatGPT dall-e to generate the exact prompt instead of embellishing?

[hearty ether]

How did you get access to it?

[open flume]

Check the pinned posts

[mortal summit]

Have you told it to follow your prompt exactly?

[open flume]

There is a dev asking for beta testers

Yeah

[mortal summit]

Usually

Follow this prompt exactly, do not modify it:
"[Prompt]"

should work. If not try a new chat.

[hearty ether]

Thanks

[open flume]

Dope thank you.

Worked!1

[hearty ether]

How do I know if the survey worked?

And is it browser or app only?

[somber tide]

mmm brick

[fair umbra]

Hi! Are there people here who filled out the early access form and didn't get a response? I've been waiting for 19 hours? 😦

[somber tide]

be patient its manual

[random saffron]

This one is so cool, giving me mr. robot vibes

[somber tide]

chatgpt moment

[tawny elm]

Hi guys! were you can sign up for ChatGPT Dall-e early access?

[slate hull]

Does dall e add its own prompts jf your prompt is basic how does that system work?

[plain moth]

I got chatgpt dalle3 access!!! YAY

[slate maple]

Hi Moxi, I submitted the form yesterday, and would love to be included in the Alpha. Is there any way you can help me to get in? (I am a paid subscriber to Plus.)

[tawny elm]

@plain moth How you were able to get early access?

[plain moth]

I filled out the pinned form yesterday

[somber tide]

to everyone asking how to get the dalle 3 preview, just be patient, moxi needs to add it manually

[tawny elm]

@plain moth OK, thank you!! 👍

[somber tide]

they seem to be asleep or busy right now

[tawny elm]

I filled out the form, thanks!! @plain moth

[somber tide]

it would be cool if, one day, the original DALL-E was released

just as a historical kinda thing

[somber zenith]

Dall-E is no longer generating images in ChatGPT for me right now. There is an error message due to high demand 😦

[rigid bough]

it's working for me

[slate maple]

I am a paid Plus subscriber and submitted the Google form for Alpha access yesterday. Have all the spots been given away, or ?????? Didn't hear back from OpenAI so I was curious if anybody knows the status of new Alpha members. Cheers.

[somber tide]

it's manual. be patient. i have access and it was given manually so

[slate maple]

Thanks Morgue. Were you sent an email to notify you? Or did you just keep trying to generate until discovered you had been admitted to Alpha?

[steady basalt]

So it being a preview, what does the "non-preview" version entail?

[somber tide]

if you have access, you should have a fifth option under gpt-4: "DALL-E 3"

[slate maple]

Thanks again

[somber tide]

the evolution

[steady basalt]

Fancy

[somber tide]

wow, you can even specify if you want fewer than 4 images

woah ok what

[steady basalt]

Oh that's neat. You think you could get it to only change the colors?

[somber tide]

im gonna try that

my immediate theory is that it uses the prompt as a sort of "seed"

alot of past ai generators used predetermined noisemaps or rng algorithms to determine a starting seed

woah ok this is interesting

[steady basalt]

It is very similar

[somber tide]

this suggests that dall-e 3 is very precise

[slate hull]

More than the other generators I have seen but there’s still a slight change in appearance

[somber tide]

yeah, like some details on the larger gear are missing and the sizes are ever so slightly larger

[dim cradle]

https://th.bing.com/th/id/OIG.bqtynGWNmowSS9es4MK0?pid=ImgGn

finally something similar

[random saffron]

I applied for the DALL-E 3 access through ChatGPT just now, using the google form, wish me luck! 🙏 (Just got accepted )

[drifting hawk]

guys im not adding more people because we are at capacity, working on fixing some stuff that will get us more capacity but will take a while

[plain moth]

thank you for adding me 🙂

[river quest]

Also waiting 🤞

1 min maybe

huh

[nimble raven]

how many months are you running plus?

[river quest]

About 4 monthes

[tame garden]

Some people wait weeks for new features, don't expect to get them in hours

[scenic swallow]

where do you find the form

[nimble raven]

should be important? i just get into this discord server but i'm a plus user for half a year

[river quest]

Check the pinned message

[scenic swallow]

it says form got leaked so it’s disabled

[tame garden]

Form got leaked so it's disabled

huh that's weird

I guess they mean it got leaked outside the discord

[scenic swallow]

yeah

[tame garden]

since this server assumes verified OpenAI accounts

[nimble raven]

yes, i got it from out the discord

[scenic swallow]

is there another way we can still fill it out

[nimble raven]

ít's pinned

[scenic swallow]

yes but it says it got leaked and it’s been disabled

I can’t fill it out

[nimble raven]

i did like 10 mins ago :/

[river quest]

same

Did it couple mins ago. Everything was fine

[scenic swallow]

okay

[lean patio]

how can I access to Dalle 3 from ChatGPT PLus?

[wary stump]

i got the same results by telling gpt to make the exact same prompt as before. i was then able to make "variations" by telling it to switch the sentences and phrasing around

[random saffron]

weird they just removed my access 💀

um

unless it's being worked on right now and nobody has access to it

[rigid bough]

I still have access

[random saffron]

no way I had it like 3 minutes ago then they removed my access

I didn't even type anything explicit

wait nevermind I had to refresh

[rigid bough]

I apologize for the inconvenience. DALL·E is currently experiencing high demand, which has affected our ability to generate images at this time. Please let me know if there's any other way I can assist you.

[open flume]

This is an outrageous achievement

[vague kelp]

I have had it give me just a couple images instead of all 4 before because it was busy

[open flume]

As someone whos' been here since the first launch I am just floored by the power of this thingf

[nimble rose]

Testing a way to "upscale" the image... it was quite interesting.

[open flume]

If you want to upsize using AI topaz megapixel is pretty great

[nimble rose]

It's a cool idea... but what I meant by upscale is within the chat itself to "work" on the image. If something like this is possible in this early version of testing, imagine what awaits us in the future heheh

[nimble rose]

in the three images it is the same village or very similar villages... but with "edited" versions, all via natural language

I wonder if functions such as upscaling the image and editing in a more centralized way are part of the team's development schedule... it would be something unimaginable to achieve.

[slate maple]

Is there any way to adjust the aspect ratio of DALL-E-3? If not, what might be the best alternative for me if I want to get a 9:11 aspect ratio (or close to that)?

[vapid elk]

that is because it currently uses a fixed seed

every generated image has the same seed, this makes the images more consistent

I think it will change at some point to use random seed... although I think letting the user choose the seed is a nice idea

[dim cradle]

you can ask for square, long or wide

[cosmic hawk]

Hey guys how can I add daale plug-in in my chatgpt?

[clear river]

My1PW@Sutter

My1PW@Sutter

[analog wadi]

Hi guys. Where do I get the link to the form so I can get access to dall-e 3??

[lost flume]

It is closed now unfortunately

[hearty ether]

Isn’t the capacity being all used up by bing?

[plain moth]

😢 Dall-E 3 I'll miss you for an hour

[small remnant]

is dalle 3 too restricted ?

[steady basalt]

Restricted in what sense?

[slate hull]

I hear you can’t make fictional characters like Mario and other characters like that sadly

I mean chat gpt will deny it but I think bing will allow it

[steady basalt]

Well Bing seems to be allowing it, I did make some Hatsune Miku.

[nimble rose]

But you can easily generate characters if you know a little about how to manipulate the prompts.

[steady basalt]

I'm more concerned with re-creating art styles, but I don't know anything about OpenAI's DALL-E 3 implementation.

[nimble rose]

Like this one for example, it's not perfect, but it was something with 1 or two prompts...

[plain moth]

ChatGPT's Dall-E 3 can be restrictive for arbitrary reasons.. like it wouldn't generate an image for this prompt that it wrote FOR me "Full-body illustration of an anthropomorphic tuxedo-colored cat with a black face and vivid green eyes. The cat is dressed in a green T-shirt and wears a black unzipped zipper hoodie. The hoodie's white drawstrings are adorned with gold-colored diamond-shaped aglets."
and before you say it's the "black face" I doubt it because it took the prompt, tried to generate it but then appearantly deemed the resulting image against the content policy... Prompt works just fine on Bing.

[slate hull]

But the character won’t look exactly the same but similar

Like Mario or any other trade marked name

[light flicker]

Anyone else finding duplication of the same human character in a multiple character scene? Only the clothes are different. This is via bing hopefully this is resolved in chatgpt version

[slate hull]

I’ve seen a few examples it can get the general look of the character but it may slightly change the art style of course you would have to experiment yourself to find out

[light flicker]

For example:

[slate hull]

Don’t know if it can do two characters consistently

[nimble rose]

The question is this... It is in a very recent testing phase and if you manipulate the prompt a little more you will get the desired result... I think that when it is officially launched, this will improve.

[light flicker]

Thanks I appear to be good at getting twins lol

[light flicker]

Thank you 🙏

[hearty ether]

Will there be more slots for plus members soon?

[eager trench]

Why every "photo, photography" that I generate with females has plastic women and huge duck lips... terrible

[hallow crypt]

I can see form isn't available anymore. I'm plus for long time and wonder when will get regular access to it unless there are planned more Alpha Access? Do you happen to know more about timing.

[plain moth]

yeah I noticed this too

it's like they took barbies and tagged them with "woman"

[eager trench]

Or...they trained them on pictures from Instagram "babes" with hundreds of plastic surgeries and massive amount of filters 😂

[drifting hawk]

we are rolling it out as fast as our servers can hold

its random and to plus users

[manic barn]

Crazy excited for the future of generative art with this. The biggest criticism I've seen from 'normies' is that they could never get other models to "generate what they want" chatGPT integration with DALLE 3 seems to be a huge step in the right direction with that.

[fading spoke]

This Alpha thing at ChatGPT I had it yesterday but now it's removed. Will it come back or am i just "banned"

idk why i would be

[nimble rose]

Maybe they are removing some restrictions, I generated a somewhat controversial image...

[fading spoke]

I generated a foal in the grass and a winter scene in a swedish forest

I dont think I got banned for a weird request

[nimble rose]

what do you guys think? does this image look like something that should be made blocked by some kind of filter?

I think so haha, I apologize to the moderators!

[ripe salmon]

if dalle technology powers both bing create and chatgpts image generation its strange that chatgpt blocks copyright but bing does not

[nimble rose]

I posted an image on the #🌄┃general-art channel honestly, I can't remember an image with such good quality generated by me through an AI

It's in these moments that I think that in the future the value will no longer be in whether or not the work is done, but rather in the "unique" design that each work will have.

[cursive hornet]

I want to use DALL·E 3 with ChatGPT, but it seems that they are no longer accepting applications...

[steady basalt]

They are rolling out it, but it's random

[nimble rose]

I don't know if you guys already do this... but try it and see the magic happen!

First, use a prompt with some rules guiding the AI ​​where you want it, whether in style, focus on details, in colors etc... be very specific here.

Redo/modify you prompt until you find the perfect "idea" of the image you want.

copy the prompt the selected image, change only key correction points and repeat the process until you get your incredible piece of digital art!

[hearty ether]

Thanks for adding me Moxi ❤️

[hearty ether]

Did you check on the desktop version?

When you start a new chat

maybe the reason I didn’t see it before is that I was on mobile at the time

[copper plank]

anyone else getting this message:
"I apologize for the inconvenience, but I encountered issues while trying to generate the images. Would you like to proceed with another request or explore a different approach? Let me know how I can assist you further."

[hearty ether]

Probably content filters?

[zinc dagger]

it may also be that the model is overwhelmed and returns a timeout error

[hearty ether]

You could ask it to modify the prompt to be more appropriate

[copper plank]

content filters have a different message

"I'm sorry, but I was unable to generate images based on the provided description due to our content policy. Please provide a different description or let me know how else I can assist you!"

guess they have server issues

[onyx mesa]

I'm also seeing this, with the leak they might have gotten overwhelmed with everyone using it, that or its a limit to image generation and since its in Alpha they don't have a proper messege for it yet

[copper plank]

oh and it also has this one as well

"I apologize for the inconvenience, but I was unable to generate all the images based on the provided description due to our content policy. However, I have created a couple of images that might fit your request. Please let me know if you'd like any further adjustments or if there's any other way I can assist you!"

in case people are searching for error messages might as well be completionist about them

[hearty ether]

Did it work?

Just my personal opinion but maybe it’s better to generate one image by default

Could be easier on the load side too

[unborn tusk]

Damn I was just getting into my stride generating and now this lmao

[unborn tusk]

No but if it stil shows the prompts, you can just copy paste them into bing image creator

[slate maple]

I keep getting this one: I apologize for the inconvenience, but I encountered issues while trying to generate the images. Please let me know if you'd like to proceed with another request or if there's any other way I can assist you.

[unborn tusk]

Most people are getting it, prolly servers got overwhelmed or theyre gonna stop the alpha?

[hearty ether]

Tbh I feel like it’s bing that’s causing most of the load

[unborn tusk]

Should limit bing users to 2 images muahha

[slate maple]

I feel lucky and grateful that I made the Alpha team......now we need to be patient while they upgrade their equipment.

[unborn tusk]

True true

[dapper plank]

I'm not sure how much I like that fact that unless I explicitly and verbatimly ask it to use my text prompt exactly as it is written, it will simply use it as inspiration to generate four new text to image prompts before generating the images.

[cursive hornet]

It seems that Bing Image Creator and Bing Chat have different generation routes; BingImageCreator stops immediately and cannot generate images, but Bing Chat does not.
I think there is some kind of structural problem on the DALL-E 3 side.
When this problem occurs, you will not be able to create images from Bing Image Creator.
until the time the boost is applied.