ChatGPT, anyone that is good with image generation?

Configuration
1

I’m trying to get ChatGPT to generate an image of today’s meal and present this on a e-ink display.
I however have issues with the border it keeps adding to the images.
I tried with a prompt that was:

params:
  domain: openai_conversation
  service: generate_image
  service_data:
    config_entry: 01KA5WZN3Y1QDT96BV86MQQAXA
    size: 1024x1024
    quality: standard
    style: natural
    prompt: >-
      generate one image of a plate of

      Carbonara in an icon style or with simple graphics. Make sure the food
      fills the image and there is not much dead space around the food
  target: {}
running_script: false

It gave me: img-OwxwqMLTCxPvf4qO3gXrlwtH.png (1024×1024)

The last prompt was:

    prompt: >-
      generate one image of a plate of

      Carbonara in an icon style or with simple graphics, make the it a close up
      image of the food

And I got this: img-GHldgCVgYmCgjhBAgnwwQQBp.png (1024×1024)

I’m not at home at the moment so I can’t see the result on the display but I would like to get less of the border around the food since the screen size and resolution is limited.
My plan is to use 150 x 150 for the image. But I can’t have this much dead space.
Other images has worked fine with 150x150, so there is hope.

Anyone that knows how to speak to them to make them give me a good image?

EDIT:

ohh…

    prompt: >-
      generate one close up image with a plate of Carbonara in an icon style or
      with simple graphics, make the it a close up image of the food

img-3Nh9Vrh9PRoMVGnzMRWOUPOw.png (1024×1024)

I believe I found a winner!

2

While undoubtedly fun, each time you do this a data center is burning energy that is roughly equivalent to turning on a microwave for 2.5 seconds. Imagine what happens to our climate if a substantial part of the 1.000.000 HA users start doing this.

Having your own pictures of often had meals is a whole lot more efficient.

3 Likes
3

Perhaps that is true, but because of that reason then I don’t understand why it’s not possible to specify the image resolution.
Why am I forced to get a 1024x1024 as the lowest resolution?

But either way, my plan is to get the images locally, but finding e-paper friendly images of food is not easy, it more or less has to be AI generated.

4

Use a photoshop filter?

Glad you saved energy by picking e-ink over LCD. :upside_down_face: