Abstract Hearing-impaired people communicate with each other through sign language (SL). Deaf and hard-of-hearing people use sign language (SL) to communicate their emotions and ideas. There are numerous sign languages, and each sign language has a variety of regionally based dialects. The lack of communication between hearing people and people with hearing impairments is a problem affecting the entire globe. This thesis proposes an application that recognizes Arabic Sign Language (ArSL) using the convolutional neural network (CNN) model in real time. ArSL2018, which is made up of 32 categories and more than 5,000 images, is used to train the CNN model. The performance of the model achieved a weighted-average, F1-score of 97%, the macro-average F1-score was 97%, and accuracy reached 97%. Next, real-time recognition was implemented using OpenCV. It started by capturing a real time frame. After that, the pretrained CNN model is fed the capture image to predict the letter. The application was implemented successfully, and real-time recognition was achieved.