Gave Claude advanced image generation / batching / editing skills
I have a project, created long ago, that is able to pick the best model for a certain image prompt - and economically the best one, of course. It has an API, and I just gave the Swagger and openapi.json URLs of that webservice to Claude, and told it example prompts of what all should be possible (batching, resizing, transparency, etc.). The last feature is it instructed to do with Python in a temp directory and only copy the final results to the repository.
End result: I can tell Claude to:
"Please create 20 cat avatars with transparency images of 32x32px, 64x64px and 128x128px in PNG format with alpha."
(I know, the prompt sucks - very badly described cats 😅)
And see the cute results that it made for the cheapest price with the best model for the task.
Comments
Look how cool, I just created a whole village to be used in a game or so! Transparent PNG's with relative sizes and stuff. How cool is that? 😄 I'm very fond of the emoji's that DevPlace uses, else I would've replaced them. Also, emoji's load so fast.
@blindxfish so, if you ever need to create custom graphics for a game, and like a lot, this would be a well option. The houses above were again generated with a very mediocre prompt:
@niggel look, pussies! 😼
haha, that automatically added cat-emoji of a *pussy. So funny 😂*