Google has filed a patent for the recognition and use of text contained in images and videos.

Finally, the understood text is run through a different program that’s trained on context for in-image text blurbs.