I think there is something similar to your idea called
$1. I've never used it but from what I underatand it's a live video app that connects a blind person's phone to a volunteer who can see through the camera. It isn't quite what you are thinking of from what I gathered from your post but it may be a good jumping off point. Your idea of descriptive audio on prerecorded scenes is a really cool idea and I definitely think you should pursue it.