tidyM-pairs

Speaker verification trial pairs for the TidyM (TidyVoiceM) dataset.

TidyM audio (train & test)

Download the TidyM train and test sets from the Mozilla Data Collective:

https://mozilladatacollective.com/datasets/cmu6xsi8201eenq07hnip96ct

Trial paths in this file are relative to the TidyM audio roots, e.g. train/ or test/test/:

<speaker_folder>/<utt>.wav

Dataset Statistics

Metric	Training Set	Test Set	Total
Speakers	141,623	70,994	212,617
Languages	81	81	81
Utterances	5.4M	218K	5,616,611
Duration (hours)	7,800	350	8,138
Trial pairs	—	~2.8M	~2.8M
Domain	Read Speech	Read Speech	Read Speech

Trial file format

tidyM-pairs.txt — each line is space-separated:

<label> <enroll_path> <test_path>

• <label>: 1 = target (same speaker), 0 = nontarget (different speakers)
• <enroll_path> / <test_path>: <speaker_folder>/<utt>.wav

Example:

1 common_voice_ca_31196622/common_voice_ca_31196622.wav common_voice_ca_31196622/common_voice_ca_31196620.wav
0 common_voice_ur_41170109/common_voice_ur_41133345.wav common_voice_ur_38425380/common_voice_ur_38425380.wav

Citation

If you use TidyM / these trials, please cite:

@inproceedings{farhadipour2026tidyvoice,
  title={Tidyvoice: A curated multilingual dataset for speaker verification derived from common voice},
  author={Farhadipour, Aref and Marquenie, Jan and Madikeri, Srikanth and Chodroff, Eleanor},
  booktitle={ICASSP 2026-2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={19047--19051},
  year={2026},
  organization={IEEE}
}