rekordbox ver.7.0.0 / AlphaTheta
Development of music analysis technology using AI for DJ
In a collaborative research project with AlphaTheta Corporation, which plans and develops DJ equipment and software under the Pioneer DJ brand, two new features based on AI technology developed by our company have been released in the latest version of their DJ software. Since 2020, our collaborative research project has included the development of AI-based vocal detection algorithms for DJs.
The new features, "RADAR" and "INTELLIGENT CUE CREATION," are based on a music recommendation system and CUE point (markers set by DJs to indicate where they start playing a track) prediction model developed by Qosmo.
"RADAR" is based on an algorithm that uses both readily available metadata such as key, tempo, genre and vocal contentas well as a deep neural network that extracts high dimensional embeddings to compute an abstract similarity metric between songs. By skillfully combining the two sources of information a list of candidate songs that the user might be interested in is generated. The recommendations can come not only from the user’s personal library, but also from cloud services such as Beatport, thus providing the opportunity to encounter new music.
“INTELLIGENT CUE CREATION” is based on a deep neural network that analyzes the various transition points within a song and develops an understanding of which ones are likely to be cue points. The proposals of the neural network are combined with the preferences of each individual user to deliver a personalized experience designed to help DJs spend less time on organizing their collection, thus enabling them to spend more time on creating their sets.
Interview
Links
Client
AlphaTheta Corporation
Credits
Machine Learning / Technical Direction: Bogdan Teleaga (Qosmo) Technical Direction: Nao Tokui (Qosmo) Project Management: Yumi Takahashi(Qosmo), Ryosuke Nakajima (Qosmo)