thanks for that context. Screen readers are gonna be a whole thing, but are a great long term goal.
Technically the hidden image is a stretch, but that's good - it's good to build what's needed and not what's easy. I'm reading into what I can do tidily now.

Color inversion is gonna be easy.
Other fancy solutions for the future could include assessing the 'brightness' of an image and tuning the super bright white ones automagically. Haven't done that before, and not sure how processor intense it is, but also a possibility. (SHH mix, stop getting excited by what you can do and focus on what you can do now)