Authentication using passwords requires three cognitively difficult actions. First, a good password requires generation of a high level of entropy. Then the person must reliably recall that highly entropic password. Finally, the person must properly map the password to the context. The common implementation and interaction designs of password studies increase these challenges. We present a system that offers entropy to the user by providing randomly-selected visuals, leverages this source of entropy for password creation, and then further utilizes this visual cue to simplify contextualization. While images have long been used for contextualization, this system is distinct in two ways: the use of images for textual password generation and for textual recall. Our results show a significant increase in entropy and length of passwords created using multiple measures of entropy with no decrease in recall of these more entropic passwords.