Picking the right model for random use-case

[roimulia2]

Hey there,
Loving the paper!

Noobie question:
I'm wondering, in case I want to choose a random image (not knowing the content) and matching the right pre-trained model to it, is there a way to automate the process or I'll have to detect what the is image metadata to sync it with the right model?
Is it possible to merge the models, so I can feed it with a single output and expect it to handle inpainting even without a context?