This study aims to improve movie recommendation systems by integrating visual and text similarity based on cosine similarity techniques. This research incorporates the use of multimodal datasets for the recommendation system. The experiment explores the utilization of Convolutional Neural Networks (CNN) for extracting visual features and suggests a hybrid recommendation approach that combines both visual similarity and textual similarity. The proposed system not only takes into account the user’s preferences based on genre and rating but also utilizes visual features to suggest movies that visually match the user’s preferences. The proposed approach has been implemented using a large movie dataset, and the results show improved accuracy and user satisfaction. This approach is of immense benefit to the advancement of movie recommendation systems by incorporating visual content analysis alongside traditional collaborative filtering techniques. The system analyzes several Movies from the IMDB dataset, taking into consideration movie features such as the Movie ID, Title, Images, Summary, and other properties.