The concept of filtering out songs based on the interest of a user is the core principle of today’s music streaming (MS) service. Recommendation Systems (RS) are a key component of the MS companies. Different companies use different types of RS. Since the web is now an important medium for almost every kind of business and electronic transaction, it serves up as the driving force for the development of RS technology. There is significant dependency that exists between user and item-based activity which is the basic principle of recommendation. With the rise of digital content distribution, people now have access to music collections on an unprecedented scale. Commercial music libraries easily exceed 15 million songs, which vastly exceeds the listening capability of any single person. With millions of songs to choose from, people sometimes feel overwhelmed.Most common RS are designed using the concept of filtering techniques and deal with the count and similarities between the likenesses of the users. Our approach, in this paper, is to enhance the RS by combining the filtering technique with Deep Learning. It will use the traditional filtering technique and use the album art of the song to recommend new songs. The hybrid RS will scan the album art of the song for unique labels. © BEIESP.