Annotated frequency contours of vocalisations produced by wild Norwegian killer whales

This data descriptor presents a curated set of manually annotated frequency-contour traces for vocalisations of wild Norwegian killer whales (Orcinus orca). The annotations were produced on 46 short audio clips extracted from the WhaleSonic hydrophone-array recordings collected in northern Norway, yielding 68 annotation sets. Audio clips were selected with a deliberate focus on biphonation calls, while also including monophonic calls. Each annotation set comprises class-specific binary masks over a spectrogram, distinguishing vocalisation components such as the fundamental frequency, harmonics and heterodynes. Data are released in complementary formats for different users: GIMP project files (XCF) with source WAV and spectrogram PNG; HDF5 archives for programmatic access; and an Excel export of pixel-level time–frequency contours with summary statistics. The resource is intended to support quantitative comparison of call structure, studies of biphonation components, and training or validation of automated spectrogram-annotation methods. Because selection was expert-driven and contour tracing was performed by a single primary annotator, the dataset should be treated as a high-detail reference corpus rather than a representative sample of the full Norwegian killer whale repertoire.