It certainly would be nice if DALL·E could give us the exact colors we ask for.
To illustrate, I used this prompt to generate the following image. But although I asked for a white background (#FFFFFF) in several different ways, it gave me a #e4e2dd instead.
Prompt:
Ultra-realistic color photo of a laptop on a desk with fingers touching the keyboard. A window and books in the background.
The laptop fills most of the image. Its screen must display a black and white illustration with a pure white (#FFFFFF) background and black (#000000) line work.
The illustration shows a girl sitting on the ground in a dense magical forest, with a laptop on her lap. She has long hair, a short-sleeve dress, and is accompanied by several small curious animals. The style is classic storybook, whimsical, and detailed—like a hand-drawn pen illustration with black lines on solid white.
The background MUST be white (#FFFFFF), and the illustration MUST use only black (#000000) linework. Use only white (#FFFFFF) as the background of the illustration.
DALL-E struggles with accurately controlling colors, the number of objects, directions, positions, and similar details, but it is good at generating a wide variety of art styles.
If you ask a specific color, you can use terms like “solid red”, “minimalist black and white”, “pure white” or “pristine green” to ensure minimal distraction.
And mostly negation (Don’t do that, without, with no…) is not working well, and DALL-E adds into image what did you mention in your prompt even “without any elephant”.
Can you see an elephant here? 
But there is no 100% solution.
It is also unable to count or place objects of the correct color in the right positions.
Additionally, DALL-E often alters the user’s prompt.
Can you see alteration here? This cat’s one eye should be blue, and another one should be amber. But, both are blue here.
Can you see only 5 balls, here?
Can you see only 2 red apple on the table, and 3 green apple on the ground, here?
Prompts - Click on Me
Note:
If you want to use prompts above, you can copy and paste images in ChatGPT and ask “Extract text from the image I uploaded”.
