Generated with AI:

Model telling using edge-detection (not sure it did, but still), followed by another generation:

With no intermediary prompt.

How computer vision algorithms “see” the world: