Help creating a prompt that actually works

need help with creating a prompt that actually has the correct spelling. The prompt that I need is: Create a 3d image of the written text “UP TO DOWN TO IT. WE DO IT CAUSE WE USE TO IT” in royal blue and silver on a white background. Please help

1 Like

I assume this is related to DALL-E? If so, what you’re asking for would be hard to achieve.

Image generation models like DALL-E 3 are not the best at text. DALL-E prioritizes visual elements over textual accuracy. This focus on aesthetics can lead to misspellings or jumbled text, especially with longer or more complex strings. It’s best at short phrases or standalone words.

It would be better to keep your text short and prominent - or, use a different tool (Like Photoshop, since you’re essentially just making word-art) to make this.


Heya! Welcome to the dev forum!

I’ve moved your post to Prompting and added the dalle3 tag. (If you follow it, you’ll find a TON of great resources, including an AMA from DALLE team members!)

But yes, @Kesku is correct that it’s not super reliable with text… yet. I’ve seen it improve vastly in the last year, though, so I imagine it will get better.

Have you tried generating the image without the text then using ChatGPT edit image or photoshop to put the text in?

No, I need 3d written text

There’s a specialized Photoshop plugin that does that. They have an AI text tool too, I believe.

I would play with your prompt, though. Include more details…

1 Like

I basically want a 3d image of the written text "UP TO IT, DOWN TO IT in the colors of royal blue and silver with the correct spelling. was having problems generating the image.

Right. I think asking for “3d image” and then adding your “text” is confusing it? Maybe ask for the image (in detail) then add something like, "With 3d text that says, “SHORT MESSAGE”… Usually one or two words are easiest for it to get right, but it is improving.