my i2v experiment has failed.
>The camera pulls back from this painting, revealing the wider gallery scene. In the foreground stands a modern woman, a perfect real-life doppelgänger of the Mona Lisa, captivated by the artwork. After a moment, she shifts her gaze from the painting to the camera, and a subtle, enigmatic smile forms on her lips.

This is my personal benchmark for a usable i2v model for me. I should be able to give it a painting, have it zoom out to show its on a wall in a museum, and have the camera pan and show a person who looks just like the person in the painting.

Whats funny is that for the mona lisa specifically I can probably acheive this in t2v relatively easily