# HEARSAY test audio Author: Alex Picon Four labeled clips selected without examining detector scores, from the ASVspoof 2019 Logical Access evaluation partition. Files are byte-for-byte copies of the local benchmark FLAC inputs, renamed for convenience. No audio was changed. Labels and original IDs/hashes are in labels.json; 0 = real, 1 = synthetic/spoof. - real-1.flac: real human speech, 3.37 seconds. Original ID: la19/test/LA_E_9102181; attack: bonafide. - real-2.flac: real human speech, 3.41 seconds. Original ID: la19/test/LA_E_4936594; attack: bonafide. - synthetic-1.flac: synthetic/spoofed speech, 2.94 seconds. Original ID: la19/test/LA_E_8970871; attack: A17. - synthetic-2.flac: synthetic/spoofed speech, 2.29 seconds. Original ID: la19/test/LA_E_9760280; attack: A19. ## Source and attribution Audio: ASVspoof 2019 database, provided by the ASVspoof organizing committee: Junichi Yamagishi, Massimiliano Todisco, Md Sahidullah, Héctor Delgado, Xin Wang, Nicholas Evans, Tomi Kinnunen, Kong Aik Lee, Ville Vestman and Andreas Nautsch. Alex Picon assembled this sample pack; the source audio belongs to its credited contributors. Source: https://zenodo.org/records/6906306 DOI: https://doi.org/10.7488/ds/2555 License: Open Data Commons Attribution License (ODC-By) 1.0, https://opendatacommons.org/licenses/by/1-0/ This sample pack contains information from ASVspoof 2019, which is made available here under the Open Data Commons Attribution License (ODC-By) v1.0. ## Use Extract the ZIP, enter your dashboard access code, then upload one FLAC file at a time and run the selected detectors. A higher raw score means more synthetic for each detector; numeric scales differ between detectors. These are demonstration files, not an independent accuracy test. Our model was trained using material from this partition; other detectors may also have seen related data. Labels describe dataset ground truth, not guaranteed detector answers. Never rank the systems from four examples.