For many informative videos the service would need to be good at excerpting relevant images (and ideally summarizing them in text for the blind). Maybe by focus

For many informative videos the service would need to be good at excerpting relevant images (and ideally summarizing them in text for the blind). Maybe by focusing on areas that are unchanged during a period of time (such as when graphs or pictures are being talked about).