A method for processing camera images in real-time is disclosed. The user starts a process that generates a feed from a mobile device of camera images of a document. An image from the feed is obtained and set as the current image. One or more of a plurality of aspects regarding the image are computed and tested against a threshold. If the aspect fails to satisfy the threshold, then the user is prompted to change the move the mobile device. Once the aspects satisfy their respective thresholds, the camera feed is stopped, the last image is saved, and character recognition of the last image is performed and displayed to the user.