Oftentimes, learning a new language is a difficult task, and especially if it requires an unfamiliar method of communication. ASL(American Sign Language) requires the movement of hands or the face to form sentence structures that convey information and everyday expressions. To allow a greater number of people to communicate with ASL, AI image recognition through CNN (Convolutional Neural Network) can be used as a means of alphabet translation without having to learn the language. By testing different setups of multiple layers and dropout, max pooling, or other optimization methods, the efficiency and accuracy of different training models were compared for the specific task.