Online singing platforms have attracted many users to sing and publish songs online. Accurate singing-song recommendation systems are essential to enhance user experience and stickiness for such platforms. User’s vocal ability is one of the major factors to consider in the recommendation model. However, how to predict a user’s vocal ability is challenging. Meanwhile, online singing platforms provide social networking services. How to utilize the social feedback information to improve recommendation performance is worth studying. To address these issues, this paper proposes new methods to predict a user’s vocal ability on a song and to model social feedback factors. Further, a novel model for singing-song recommendations is developed. Experiments conducted on a real-world dataset demonstrate the effectiveness of our proposed model and the necessity of modeling vocal ability and social feedback factors to improve recommendation performance.