This is the Zenodo repository for the ASVspoof 5 database. ASVspoof 5 is the fifth edition in a series of challenges which promote the study of speech spoofing and deepfake attacks, and the design of detection solutions. Compared to previous challenges, the ASVspoof~5 database is built from crowdsourced data collected from around 2,000 speakers in diverse acoustic conditions. More than 20 attacks, also crowdsourced, are generated and optionally tested using surrogate detection models, while seven adversarial attacks are incorporated for the first time.

• Please check README.txt and LICENSE.txt before downloading the database.
• Database paper (to be submitted): https://arxiv.org/abs/2502.08857
  • Please consider citing the reference listed at the bottom of this page.
• It is highly recommended to follow the rules and instructions in the ASVspoof 5 challenge evaluation plan (phase 2, https://www.asvspoof.org/), if you want to produce results comparable with the literature.
• Latest work using the ASVspoof 5 database can be found in the Automatic Speaker Verification Spoofing Countermeasures Workshop proceeding: https://www.isca-archive.org/asvspoof_2024/index.html
• If you are interested in creating spoofed data for research purpose using the ASVspoof 5 protocol, please send request to info@asvspoof.org

This is a database used for the Third Automatic Speaker Verification Spoofing and Countermeasuers Challenge, for short, ASVspoof 2019 (http://www.asvspoof.org) organized by Junichi Yamagishi, Massimiliano Todisco, Md Sahidullah, Héctor Delgado, Xin Wang, Nicholas Evans, Tomi Kinnunen, Kong Aik Lee, Ville Vestman, and Andreas Nautsch in 2019.

The ASVspoof 2021 dataset is a large-scale public dataset for speaker verification and spoofing countermeasures. The dataset contains various types of audio files, including real and fake audio, with different levels of noise and background music.

The database has been used in the first Automatic Speaker Verification Spoofing and Countermeasures Challenge (ASVspoof 2015). Genuine speech is collected from 106 speakers (45 male, 61 female) and with no significant channel or background noise effects. Spoofed speech is generated from the genuine data using a number of different spoofing algorithms. The full dataset is partitioned into three subsets, the first for training, the second for development and the third for evaluation. More details can be found in the evaluation plan in the summary paper.

This repository contains speech endpoint annotations and filelists for different artefacts we found during our study on the ASVspoof 2017 v2.0 dataset as part of our work in the paper "Dataset biases in speaker verification systems: a case study on the ASVspoof 2017 benchmark" which is to be submitted to the IEEE Transactions on Biometrics, Behavior, and Identity Science (T-BIOM).

This is the Huggingface repository for the ASVspoof 5 database. ASVspoof 5 is the fifth edition in a series of challenges which promote the study of speech spoofing and deepfake attacks, and the design of detection solutions. Compared to previous challenges, the ASVspoof~5 database is built from crowdsourced data collected from around 2,000 speakers in diverse acoustic conditions. More than 20 attacks, also crowdsourced, are generated and optionally tested using surrogate detection models… See the full description on the dataset page: https://huggingface.co/datasets/jungjee/asvspoof5.

ASVspoof 2021 LA subset

sorted original dataset from ASVspoof 2021 LA subset credit: ASVspoof 2021 challenge released under The databases are available under an Open Data Commons Attribution Licence and can be downloaded from the Zenodo repository.

This dataset is a derivitave work of the ASVSpoof 2019 LA condition listening test data found here:
https://datashare.ed.ac.uk/handle/10283/3336
-> LA.zip

"ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech"
Xin Wang, Junichi Yamagishi, Massimiliano Todisco, Héctor Delgado, Andreas Nautsch, Nicholas Evans, Md Sahidullah, Ville Vestman, Tomi Kinnunen, Kong Aik Lee, Lauri Juvela, Paavo Alku, Yu-Huai Peng, Hsin-Te Hwang, Yu Tsao, Hsin-Min Wang, Sébastien Le Maguer, Markus Becker, Fergus Henderson, Rob Clark, Yu Zhang, Quan Wang, Ye Jia, Kai Onuma, Koji Mushika, Takashi Kaneda, Yuan Jiang, Li-Juan Liu, Yi-Chiao Wu, Wen-Chin Huang, Tomoki Toda, Kou Tanaka, Hirokazu Kameoka, Ingmar Steiner, Driss Matrouf, Jean-François Bonastre, Avashna Govender, Srikanth Ronanki, Jing-Xuan Zhang, Zhen-Hua Ling.
Computer Speech and Language Colume 64, 2020.

This form of the data was used for the PRS paper accepted to ASRU 2023:

"Partial Rank Similarity Minimization Method for Quality MOS Prediction of
Unseen Speech Synthesis Systems in Zero-shot and Semi-supervised Setting."
Hemant Yadav, Erica Cooper, Junichi Yamagishi, Sunayana Sitaram, Rajiv Ratn Shah.

Modifications to the original data include converting audio from flac -> wav, sv56 normalization, conversion of labels from an 0-9 rating scale to a 1-5 scale, and creation of training/development/testing splits.

Dataset

This dataset was created by Awsaf

Contents

Conventional copy-paste augmentations generate new training instances by concatenating existing utterances to increase the amount of data for neural network training. However, the direct application of copy-paste augmentation for anti-spoofing is problematic. This paper refines the copy-paste augmentation for speech anti-spoofing, dubbed CpAug, to generate more training data with rich intra-class diversity. The CpAug employs two policies: concatenation to merge utterances with identical labels, and substitution to replace segments in an anchor utterance. Besides, considering the impacts of speakers and spoofing attack types, we craft four blending strategies for the CpAug. Furthermore, we explore how CpAug complements the Rawboost augmentation method. Experimental results reveal that the proposed CpAug significantly improves the performance of speech anti-spoofing. Particularly, CpAug with substitution policy leads to relative improvements of 43% and 38% on the ASVspoof’ 19LA and 21LA, respectively. Notably, the CpAug and Rawboost synergize effectively, achieving an EER of 2.91% on ASVspoof’ 21LA.

Voice conversion (VC) is a technique to transform a speaker identity included in a source speech waveform into a different one while preserving linguistic information of the source speech waveform.

In 2016, we have launched the Voice Conversion Challenge (VCC) 2016 [1][2] at Interspeech 2016. The objective of the 2016 challenge was to better understand different VC techniques built on a freely-available common dataset to look at a common goal, and to share views about unsolved problems and challenges faced by the current VC techniques. The VCC 2016 focused on the most basic VC task, that is, the construction of VC models that automatically transform the voice identity of a source speaker into that of a target speaker using a parallel clean training database where source and target speakers read out the same set of utterances in a professional recording studio. 17 research groups had participated in the 2016 challenge. The challenge was successful and it established new standard evaluation methodology and protocols for bench-marking the performance of VC systems.

In 2018, we have launched the second edition of VCC, the VCC 2018 [3]. In the second edition, we revised three aspects of the challenge. First, we educed the amount of speech data used for the construction of participant's VC systems to half. This is based on feedback from participants in the previous challenge and this is also essential for practical applications. Second, we introduced a more challenging task refereed to a Spoke task in addition to a similar task to the 1st edition, which we call a Hub task. In the Spoke task, participants need to build their VC systems using a non-parallel database in which source and target speakers read out different sets of utterances. We then evaluate both parallel and non-parallel voice conversion systems via the same large-scale crowdsourcing listening test. Third, we also attempted to bridge the gap between the ASV and VC communities. Since new VC systems developed for the VCC 2018 may be strong candidates for enhancing the ASVspoof 2015 database, we also asses spoofing performance of the VC systems based on anti-spoofing scores.

In 2020, we launched the third edition of VCC, the VCC 2020 [4][5]. In this third edition, we constructed and distributed a new database for two tasks, intra-lingual semi-parallel and cross-lingual VC. The dataset for intra-lingual VC consists of a smaller parallel corpus and a larger nonparallel corpus, where both of them are of the same language. The dataset for cross-lingual VC consists of a corpus of the source speakers speaking in the source language and another corpus of the target speakers speaking in the target language. As a more challenging task than the previous ones, we focused on cross-lingual VC, in which the speaker identity is transformed between two speakers uttering different languages, which requires handling completely nonparallel training over different languages.

As for listening test, we subcontracted the crowd-sourced perceptual evaluation with English and Japanese listeners to Lionbridge TechnologiesInc. and Koto Ltd., respectively. Given the extremely large costs required for the perceptual evaluation, we selected 5 utterances (E30001, E30002, E30003,E30004, E30005) only from each speaker of each team. To evaluate the speaker similarity of the cross-lingual task, we used audio in both the English language and in the target speaker’s L2language as reference. For each source-target speaker pair, we selected three English recordings and two L2 language recordings as the natural reference for the converted five utterances. 

This data repository includes the audio files used for the crowd-sourced perceptual evaluation and raw listening test scores.

[1] Tomoki Toda, Ling-Hui Chen, Daisuke Saito, Fernando Villavicencio, Mirjam Wester, Zhizheng Wu, Junichi Yamagishi "The Voice Conversion Challenge 2016" in Proc. of Interspeech, San Francisco.

[2] Mirjam Wester, Zhizheng Wu, Junichi Yamagishi "Analysis of the Voice Conversion Challenge 2016 Evaluation Results" in Proc. of Interspeech 2016.

[3] Jaime Lorenzo-Trueba, Junichi Yamagishi, Tomoki Toda, Daisuke Saito, Fernando Villavicencio, Tomi Kinnunen, Zhenhua Ling, "The Voice Conversion Challenge 2018: Promoting Development of Parallel and Nonparallel Methods", Proc Speaker Odyssey 2018, June 2018.

[4] Yi Zhao, Wen-Chin Huang, Xiaohai Tian, Junichi Yamagishi, Rohan Kumar Das, Tomi Kinnunen, Zhenhua Ling, and Tomoki Toda. "Voice conversion challenge 2020: Intra-lingual semi-parallel and cross-lingual voice conversion" Proc. Joint Workshop for the Blizzard Challenge and Voice Conversion Challenge 2020, 80-98, DOI: 10.21437/VCC_BC.2020-14.

[5] Rohan Kumar Das, Tomi Kinnunen, Wen-Chin Huang, Zhenhua Ling, Junichi Yamagishi, Yi Zhao, Xiaohai Tian, and Tomoki Toda. "Predictions of subjective ratings and spoofing assessments of voice conversion challenge 2020 submissions." Proc. Joint Workshop for the Blizzard Challenge and Voice Conversion Challenge 2020, 99-120, DOI: 10.21437/VCC_BC.2020-15.