Hi all,

I messed with this idea this morning and wanted to share my small progress. Maybe it can help others with ideas. The idea is to use AI to create new gumps backgrounds, buttons, etc.

I started with background since I believe they are the most useful and hard to do. Here are the steps I followed :
  1. I used BING image generator to create a "Square game UI" with some requirements.
  2. Chose my image, this is the one I got:
    _0ba47315-862c-4c46-a2d2-b966411d68f9.jpg
  3. Split the image in GIMP
  4. Import in the gump files and test, here is my result, which is fully dynamic:
    1703857807505.png
Now obviously this is choppy and needs refinement, but I found the result to be promising for a quick 1 hour fiddling. I believe there should be a way to better instruct the AI to generate an image that is easier to chop and repeat seamlessly.

I will use this thread to document my process / progress.
 
Mind boggling that we can even do this. I've been wanting to start toying around with OpenAI/ChatGPT but im intimidated with all the other learning that goes into servuo/uo customs. Probobly doesnt help that I toy with something for a few weeks/a month then dissappear for years only to come back not remembering anything I had previously learned.
 
Indeed, with no Photoshop / drawing skills I feel like someone can get very acceptable results. Here's a try I just did. I instructed Bing to generate images without "lighting" and with a matte finish. I think this helped in having images that can be easily repeated without clashing with the tone.

I still had to tweak in GIMP, but I'm not skilled at all and could get the result below. First the AI generated image and second the result once tweaked and put into MUL, displayed in Gumpstudio.

1703956098471.jpeg
1703956124485.png
 
Hi all,

I messed with this idea this morning and wanted to share my small progress. Maybe it can help others with ideas. The idea is to use AI to create new gumps backgrounds, buttons, etc.

I started with background since I believe they are the most useful and hard to do. Here are the steps I followed :
  1. I used BING image generator to create a "Square game UI" with some requirements.
  2. Chose my image, this is the one I got:
    View attachment 22811
  3. Split the image in GIMP
  4. Import in the gump files and test, here is my result, which is fully dynamic:
    View attachment 22813
Now obviously this is choppy and needs refinement, but I found the result to be promising for a quick 1 hour fiddling. I believe there should be a way to better instruct the AI to generate an image that is easier to chop and repeat seamlessly.

I will use this thread to document my process / progress.
The thing is though you are not giving much for the AI to Process, When working with AI Details matter a lot in the prompts, But you are also dealing with game art which is more complex than your simple AI art. so For Example, The Prompt I am using for this is - A game's user interface (UI) for a Role Playing game. The game seems to be set in a medieval fantasy world, Detailed Texture of a 2000's Role playing game graphics, and another Prompt like this - A game's user interface (UI) for a Role Playing game. The game seems to be set in a medieval fantasy world, Detailed Texture of a 2000's Role playing game graphics with ( Stone Textures that are Monotone colors) There are plenty of ways to come about all this, but With how AI currently is you would need to learn how to edit the textures and make them the way you are truly wanting With More Details you add though the more it can pull to create what you are wanting, Another Prompt Example Idea - A game's user interface (UI) for a Role Playing game. The game seems to be set in a medieval fantasy world, Detailed Texture of a 2000's Role playing game graphics with Stone Textures that are Monotone colors) Blank space in the center, Fit to Aspect Ratio Full Image, (Nordic Carvings Inspired)
 

Attachments

  • image_2024-01-06_004754919.png
    image_2024-01-06_004754919.png
    304.3 KB · Views: 32
  • image_2024-01-06_004734927.png
    image_2024-01-06_004734927.png
    417.5 KB · Views: 33
  • image_2024-01-06_004123370.png
    image_2024-01-06_004123370.png
    484.8 KB · Views: 34
  • image_2024-01-06_004013925.png
    image_2024-01-06_004013925.png
    430.9 KB · Views: 30
  • image_2024-01-06_003802143.png
    image_2024-01-06_003802143.png
    329.8 KB · Views: 33
  • image_2024-01-06_003425410.png
    image_2024-01-06_003425410.png
    478.8 KB · Views: 37
  • image_2024-01-06_003149156.png
    image_2024-01-06_003149156.png
    473.5 KB · Views: 36
  • image_2024-01-06_003139421.png
    image_2024-01-06_003139421.png
    342.7 KB · Views: 33
  • image_2024-01-06_004837561.png
    image_2024-01-06_004837561.png
    511.5 KB · Views: 31
Hi Erachiro, nice gumps! I agree, you can also create full UI and use those directly. I just feel like everytime you have to make a change to your you are stuck with a fixed image and you have to fiddle with AI.

My idea is to create new gumps textures to
build traditional "à la pièce" gumps in game.
 
Hi Erachiro, nice gumps! I agree, you can also create full UI and use those directly. I just feel like everytime you have to make a change to your you are stuck with a fixed image and you have to fiddle with AI.

My idea is to create new gumps textures to
build traditional "à la pièce" gumps in game.
I understand what you saying but most AI images been tweaked by someone who often edits photos etc.
 
Last edited:
Back