Web-based Dynamic Visualization of Rhythmic Latent Spaces

We present a web-based visualizer designed for the dynamic representation of rhythmic latent spaces. These spaces are learned from user-provided MIDI clips with rhythms. Instead of computing a specific set of hand-coded features to describe the rhythms and using them for visualizing the space, we rely directly on the rhythmic patterns and their pulses to trigger visual cues in the canvas of the browser. To the best of our knowledge, this is the first time that a dynamic visualization has been implemented to observe a latent space learned from rhythmic patterns.