TidyVoiceX2_ASV

[Update — 2 April 2026.] A new version of this dataset is available with the same folder structure as TidyVoiceX_ASV. The official trial pair lists were released on 2 April 2026.

TidyVoiceX2_ASV Dataset

Overview

TidyVoiceX2_ASV is a multilingual speech corpus curated for cross-lingual speaker verification research. The dataset is organized speaker-wise, with per-speaker language subfolders and WAV utterances. This structure supports controlled experiments on speaker verification under language mismatch conditions.

This dataset is the official evaluation set used in the TidyVoice2026 Challenge, an official challenge at Interspeech 2026, focused on advancing cross-lingual speaker verification systems.

Challenge website: https://tidyvoice2026.github.io/

It is also used as the evaluation benchmark data partition for the TidyLang2026 Challenge, an official challenge at Odyssey 2026, presented at: https://tidylang2026.github.io/.

This dataset is disjoint from the separately released TidyVoiceX_ASV dataset: the speaker set and utterance recordings in TidyVoiceX2_ASV are different from those in TidyVoiceX_ASV.
TidyVoiceX_ASV: https://datacollective.mozillafoundation.org/datasets/cmihtsewu023so207xot1iqqw

Challenge Provenance and Evaluation Benchmarks

The evaluation benchmarks and official trial pair lists for speaker and language recognition on this dataset are publicly released by the TidyVoice/TidyLang challenges.

• Official Automatic Speaker Verification (ASV) trial pair list: https://drive.google.com/file/d/1OnfzE_YFMGJKU1LOGsD1NKNaMJOvOvmM/view?usp=share_link

• Official LID and low-resource language recognition trial lists: https://drive.google.com/file/d/1XjMURDIe026IapEorg1sVQ5eZ3Jjz9Dv/view?usp=share_link

Dataset Statistics

Total duration was computed by summing per-file duration from WAV headers (16 kHz PCM) using scripts/compute_wav_duration.py on 2026-04-02.

Language Coverage

Total languages: 62

Key Features

• Multilingual coverage with 62 language codes

• Speaker-wise directory structure for verification workflows

• Cross-lingual speaker verification focus

• Pseudonymized speaker IDs

• WAV audio format

Citation

If you use this dataset in your research, please cite:

@inproceedings{farhadipour2026tidyvoice,
  title={TidyVoice Challenge: Cross-Lingual Speaker Verification},
  author={Farhadipour, Aref and Marquenie, Jan and Madikeri, Srikanth and Vukovic, Teodora and Dellwo, Volker and Reid, Kathy and Tyers, Francis M. and Siegert, Ingo and Chodroff, Eleanor},
  booktitle={Interspeech 2026},
  year={2026}
}

@article{farhadipour2026tidyvoice_arxiv,
  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},
  journal={arXiv preprint arXiv:2601.16358},
  year={2026}
}

License

This dataset is derived from Mozilla Common Voice. Please refer to the Mozilla Common Voice terms for usage conditions: https://commonvoice.mozilla.org/en/terms

Usage Restrictions

Per the stewarded usage rules for this dataset, speaker identification is not allowed.

Contact

• Email: aref.farhadipour@uzh.ch, areffarhadi@gmail.com