Didn't know it also does video. But accessibility tools don't just want video, they want APIs to read all text and other GUI structures directly. OCR is just too imprecise and a waste of resources.