DALLE3 Rolling out to ChatGPT Plus members

Looks like DALLE3 is gradually rolling out to ChatGPT Plus members… Should be hitting others soon… Quality is out of this world…

It’s not available via Labs or API yet, but that’s coming later this year after this rolls out, I’m sure…

Any prompts you want me to test?


First thing I’m noticing is prompting is way different… you could do pseudo-code before and get what you want with lots of control with arcane language, but now it’s a lot easier… here’s some assets in DALLE2.experimental then first shot at DALLE3 and a second shot at DALLE3…

Also, no copyrighted names… ie “Unreal Engine 5” was blocked… you don’t need “trick” keywords like that anymore really, though… just natural language…


Looks to be better at text…compared to DALLE2.exp…


This is crazy. GPT-4 improved prompting makes it even better.

1 Like

Welcome to the forum.

Yeah, the prompting is a lot different.

It’s still stumbling on text for comics, but it sometimes gets it… Could be my prompting too… When it eventually hits the API, I’ll need to change my software prompts, I think…

Epic fantasy comic panel rendered as a photorealistic illustration, styled as a woodcut-sketch with the essence of a Conan Graphic Novel. The focus is on a dramatic character, appearing heroically triumphant after an adventure. The scene is enriched by ultrafine crosshatch halftone details and cross-hatching. The background features a striking black and white radial fractal geometric pattern, while the foreground pops with luminous, moody watercolors in an ink-wash approach.

dalle3 vs #DALLE2exp

Another comparison… hands seem a lot better…

1 Like

Now compare to images API :coffin:

I mean, that’s still DALLE2 for the moment, so it’s gonna be worse. :wink:

The DALLE3 API will be here after they get ChatGPT and Labs sorted, I’m sure…In the meantime, I’ll be changing my prompts in my SaaS for the new model when it drops…

Ugh. I’m so envious.

What are the image dimensions? Those pillars look amazing.

Really hope they continue with in & out painting

1 Like

I’ve not played too much yet, but I’ve got it to do a “wide” single image… standard is four… Looks like they’re coming out 1024x1024 standard…

It’ll probably be in Labs next then finally API before the end of the year maybe?

1 Like

A bit better with text…


How do I know when this is available on my Plus subscription?

1 Like

It’ll show up here… I had to refresh…

1 Like

What was the resolution for the wide single image? Was it still the 1024x1024?

1792x1024 for the wide landscape orientation. It also does a tall portrait orientation that is 1024x1792.

1 Like

Awesome, thank you for sharing. Excited to have access!

Looks like it moved here for beta…

1 Like

Mahatma Gandhi ji image of wearing black goggles and jeans shirt blazer. hairstyle modern

Same here. Did a refresh and it was in the list for me as well. Did not have alpha access.

All I have is “Browse with bing” and I only enabled it by going into settings → new features. The only thing listed there is “Browse with bing”. I just upgraded to chat gpt plus :frowning:

If I am a ChatGPT Plus member, will I still have to buy credits to use it or it would just work like GPT-4?