Project Naptha can recognize the text found inside images and expose it for normal uses, such as copying and pasting, or routing to a screen reader. Pretty amazing. This could transform how scanned PDFs are handled for accessibility, and lower the agita over artsy text in general, often found in logos and content-rich images.
Naptha can also remove text from images, making the images easier to see or understand. Just speculating here, but if the technique can be applied to videos, one could turn open captioned videos into closed captioned ones by capturing the text (in each frame), erasing it, but saving the text into a synchronized text file.
Hat tip to Steve Faulkner at The Paciello Group.