A tool has been developed that recognizes words or individual letters from a video stream in real time. The possibilities and prospects for its application in modern society are considered. The results of experiments to test the performance of this tool using the example of English words and Latin letters are presented.