PawChorale: A Part-Aware Choral Audio–MIDI Dataset (Version 1.0.0)

PawChorale is a quality-controlled, part-aware choral audio–MIDI dataset containing 200 choral works. Each work provides a mastered choral mixture, available isolated vocal parts, aligned per-part MIDI labels, a choral-mixture MIDI representation, MusicXML, and source metadata. The release contains 6.80 hours of master audio, 27.22 hours of isolated parts, and 110,702 vocal MIDI notes. Vocal scoring is variable; works are not required to contain all four SATB parts. Source rights are edition-specific: 172 source editions are marked CPDL, three are marked Creative Commons Attribution Non-Commercial, and 25 are marked Public Domain. See RIGHTS_NOTICE.md and source-rights-manifest.csv in the archive before reuse. Project website: https://pawchorale.github.io/