Ask HN: State of accessibility software based on computer vision

most accessibility software nowadays is based on accessibility APIs, which means that: A) it only works (well?) for software which supports it B) it must be installed on the same place as the software being used. (so remote based control may not work. I assume some know to pass the API, but some don't)

this is a issue, when looking for jobs for people who need accessibility help too many of the bigger companies (here ae least) have all sorts of weird setups and software which too often make accessibility software unusable.

I think nowadays this can be solved with a computer vision based solution, (soms type of screen recorder/camera -> AI -> tts flow). I found a small amount of such things, and would like to hear thought on this.

so: what exists already, and how usable is it? what might be worth creating?

for example a seperate (phone based?) tool which will work for any combuter without any setup there?

thanks!