jae

that's super fair! and I appreciate your response :) For these, I would probably aim to describe the general vibes of the input and then contrast it with the output -- because that's most of the information I'm seeing from them. As an example, for the second one, I'd write something like:

Another side-by-side comparison. The input image (left) has a gaunt looking guy holding up a lit cigarette between his fingers, wearing a dark purplish shirt that's unbuttoned at the top. The output image (right) depicts an anime girl - her long hair has streaks of pink and purple and cyan, and she's wearing a vest-like thing with flowers.

Probably the first image would have a general description of the format as well, with later ones piggy-backing on it so you're not re-typing the same info (and a screenreader isn't re-reading it!).

Not all alt text needs to be this descriptive, of course! If you can't spend as long on it, try pick a few key things that the image conveys within the context. This webpage has a lot of good info: https://accessibility.huit.harvard.edu/describe-content-images