Can DALL-E generate my desired image? Or is the target too complex?

Hello,

I’m trying to generate a picture that seems “simple” to generate. My initial prompt was:

Generate a realistic photo, where a professional IT manager is sawing a branch on which one female IT student and one male IT student are sitting. The students are working with their laptops.

The first result was promising:

But no matter how I updated the prompt, I couldn’t get what I want.
I tried to “guide” the generation without updating the prompt on my own:

Enhance the picture so that people have just two legs, and both students have laptops:

No, the male student is missing and both students must be sitting on the branch:

:frowning:

Then I tried to write detailed prompt based on what chatGPT provided to Dall-e (and added the frightening atmosphere):

A frightening photo depicting a professional IT worker on the ground and one male and one female IT students on the branch of a tree. The professional IT worker is standing on the ground under a branch he is sawing with big saw he holds in both hands. On this branch, the female IT student and the male IT student are clearly focused and working with a laptop, and there are no books. The tree is lush and green, and the scene is set outdoors with a frightening sky in the background.

A photo depicting three people : one male and one female IT students sitting at the end of the branche of a tree, and a professional IT worker, in professional IT outfit, on the ground below the branch on which the students are sitting.
The professional IT worker is sawing the branch with a big saw he holds in both hands. The female IT student and the male IT student are clearly focused and working with a laptop, and there are no books. The branch is lush and green, and the scene is set outdoors with a frightening sky in the background.


:frowning:

Since the first attempt was the best I tried to repeat it in new chats:

Generate a realistic photo, where a professional IT manager is sawing a branch on which one female IT student and one male IT student are sitting. The students are working with their laptops.

:frowning:

The best result I got was:

It’s close but not exactly what I want.

Finally my questions:

  1. Any clue on what I’m doing wrong?
  2. Any good resource to improve my prompting to generate picture?
  3. In general, at the moment, the image generation process to get what I want is so long, that I don’t use it much. Is this only me?

Thanks a lot for your help :pray: :pray: :pray: !!

Generate a realistic photo, where we see a view up in a tree. A professional IT manager stands at the base of a tree branch and is starting to saw through the branch. Seated on the end of the branch with legs dangling are two male and female IT students, and both students are working attentively on their laptop computers. The photograph illustrates metaphorically the manager putting the students in danger.

I don’t think the AI understands cause and effect…

1 Like

You got it faster than me!

I don’t get your comment regarding causality.

I don’t see causality in my prompts. Do you?

The sawing to send students plummeting needs understanding of where to cut?

Perhaps a left-to-right description of each item we see, without relying on intelligence, will affect the items appearing in the correct manner.

1 Like

Okay, I see. I may not be descriptive enough.
I will try .
Thanks a lot :slight_smile: !