Great stuff. I think such AI image conversions definitely can be used to create VGA-like versions of Sierras parser EGA games. Either by using SCUMMVMX (little work) or creating genuine SCI01 VGA parser games. Some challenges, though, like the generated images moving the camera, modifying objects, use different colors in different images, etc. But still possible, I would think.