JKU-ABPU "Zauberflöte" Dataset: An Open Access Opera Dataset for Music Research

The JKU-ABPU "Zauberflöte" Dataset is a new multimodal open-source opera dataset based on a live performance of the opera "Die Zauberflöte" by Wolfgang Amadeus Mozart that includes audio and video recordings of the complete live show. The opera was staged and performed by the Johannes Kepler University (JKU) orchestra and singing students from the Anton Bruckner University of Music, Linz, in 2019. We also include manual bar and lyrics annotations affixed to the given performance, as well as to an additional performance conducted by Herbert von Karajan in 1980 at the Berlin Philharmonie and recorded on a CD recording commercialized under the label Deutsche Grammophon (Catalog Number: 410 967-2; MusicBrainz ReleaseID: 3bdd8575-0930-49e0-9257-86fefed16adc).

The data comprises a first folder (01_Recordings) that includes the audio (01_audio) and video (02_video) recordings of the 2019 performance conducted by Christian Radner. It also specifies the structure of the live performance as well as the cutting into parts for Radner (2019) and Karajan (1980) performances.

The data also comprises a second folder (02_Annotations) that contains bar and lyrics annotations manually affixed to both Radner and Karajan performances. It also includes the complete lyrics table provided by the Vienna State Opera, Vienna, Austria.

This dataset is particularly meant to be a resource for the music alignment and tracking research community and could also be used for another topic of research (actors’ analysis, conductor analysis, etc.). For examples of research works using pairs of annotated recordings, we point the reader to the following publications, where the bar and/or lyrics annotations were used for developing and evaluating various algorithms for on-line opera tracking:

• Brazier, C., Widmer, G. (2020). Towards Reliable Real-Time Opera Tracking: Combining Alignment with Audio Event Detectors to Increase Robustness. In Proc. of the Sound and Music Computing Conference (SMC), pages 371-377, Turin, Italy.
• Brazier, C., Widmer, G. (2021). On-Line Audio-to-Lyrics Alignment Based on a Reference Performance. In Proc. of the International Society for Music Information Retrieval Conference (ISMIR), Online.
• Brazier, C. (2023). Autonomous and Robust Live Tracking of Complete Opera Performances. PhD Thesis. Johannes Kepler University, Linz, Austria.

We are indebted to the JKU University Orchestra and the Anton Bruckner University of Music (and in particular the members of the opera singing class) for permitting us to publish this dataset, including recordings of their performance. A complete list of all musicians and contributors can be found in: https://www.bruckneruni.at/fileadmin/user_upload/12_Presse/Presse_SS_2019/2019_05_17_Zauberfloete_Programm_ABPU.pdf