Quantifying similarity of melodic features has been a chal- lenge not only in Music Information Retrieval (MIR), but music theory in general. Despite this, music theoretical models are often underutilized in music informatics. Mu- sic theory informed models can provide certain advan- tages over pure machine learning models, such as resource- efficiency and interpretability. This paper proposes the ap- plication of Earth Mover’s Distance onto a novel represen- tation of the interval content of a song based on the cir- cle of fifths, dubbing the resulting distance metric Inter- val Mover’s Distance. Interval Mover’s Distance may be applied to various MIR tasks concerning stylistic similar- ity, such as music recommendation algorithms, and par- ticularly, novel machine learning tasks such as manifold learning. Uniform Manifold Approximation and Projec- tion (UMAP) is then used to demonstrate this metric in manifold learning resulting in a large corpus of music rep- resented on a "music map" by melodic content. This mu- sic map, dubbed "Every Interval at Once", visualizes and sonifies the spectrum of data available, thus blurring the lines between music data visualization and data-driven sound art.In addition, the MIDI dataset collected for this project, spanning 23,059 songs across 75 unique genres, is released for public use.