Can Dall-E3 not add text already?

API
1

[rant]I know GPT-4o image generation is soon available for API (hopefully affordable), but why does Dall-E3 still add text to generated images? Just today it essentially pasted in the picture prompt, of course misspelled. I hope GPT-4o will not have this issue at all.[/rant]

Are there any tricks of the trade to secure text is not added? The only advice I’ve seen is to not ask it to not add text.

2

Hi, welcome back!

DALL-E cannot create text perfectly because official FAQ page says:

https://help.openai.com/en/articles/6781228-how-can-i-generate-text-in-my-image

How can I generate text in my image?

You might be tempted to instruct DALL·E to generate text in your image, by giving it instructions like “a blue sky with white clouds and the word hello in skywriting”.

However, this is not a reliable or effective way to create text.

DALL·E is not currently designed to produce text, but to generate realistic and artistic images based on your keywords or phrases.

While you can request text in your image descriptions, the results might be distorted, unclear, or not as expected, as it does not have a specific understanding of writing, labels or any other common text.

3

The problem is I don’t want text (or ask for it) at all and it seemingly randomly generates it anyway.

1 Like
4

Yeah, this is another problem with DALL·E.

If you describe something with certain words, objects, or details, it might still include text in image.

Even if you say “don’t add an elephant,” it doesn’t really understand negative prompts and might still put one in.

But new 4o Image Generating (ImageGen) is good.

This is DALL-E:

Can you see any elephant? :smiley:

polepole-image1
polepole-image1679×599 118 KB

This is ImageGen:

polepole-image2
polepole-image2695×422 46.8 KB

5

I see two :).

In my case it’s a 3-stage approach:

  • The user writes what a site should be used for
  • AI generates a site description etc based on my instructions and the user prompt
  • AI then generates a Dall-E3 prompt based on my instructions (photo-realistic etc) and the generated site description
  • That information is then fed to the Dall-E3

It’s a rather effective approach except for the text bit.

I asked AI for how I should write for Dall-E3 to not add text, and that still generated text :).

1 Like
6

DALL-E modifies user’s’ prompt. To prevent it, add a phrase on the top:

I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:

It is from OpenAI official page:

https://platform.openai.com/docs/guides/image-generation#dall-e-3-prompting

polepole-image
polepole-image1895×907 293 KB

7

Seems to stick better to the provided description, but I will evaluate it further.

Thanks.