Implementations of identifying character information in media content are described. In one implementation, a frame of media content is marked with a frame identifier including one or more known characters. These known characters can uniquely identify the frame of media content. During transmission, compression, decompression, etc., of the frame, loss can occur. This loss can affect a quality of presentation of one or more of the known characters in the frame identifier. Therefore, when the frame is subsequently examined, the frame identifier can be identified, and best matches of known characters from a character recognition library can be found for characters in the frame identifier.