Pith. sign in

REVIEW 3 major objections 6 minor 48 references

HISPASpoof: A New Dataset For Spanish Speech Forensics

T0 review · 3 major / 6 minor · reviewed 2026-08-04 · deepseek-v4-flash

Pith's one-line read HISPASpoof introduces the first large-scale Spanish dataset for synthetic speech detection and attribution, and training on it substantially improves Spanish deepfake detection.

desk verdict Useful new Spanish deepfake dataset, but the detection gains may partly be a channel artifact; worth engaging as a benchmark, not as proof of detector skill. read the letter →

arxiv 2509.09155 v1 pith:D3JHWX7V submitted 2025-09-11 cs.LG cs.AI

classification cs.LGcs.AI
keywords syntheticspeechdetectionforensicsSpanishdeepfaketext-to-speechvoicecloningattributionzero-shotTTSbenchmarkdataset
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper claims that Spanish synthetic speech is poorly handled by detectors trained on English or on small multilingual corpora, and that the gap is a data problem rather than an unsolvable technical one. To close it, the authors build HISPASpoof, a benchmark of real and synthetic Spanish speech spanning six accents and six modern zero-shot voice-cloning systems, with both detection and attribution subsets. They show that every tested detector improves sharply when trained on HISPASpoof, with error rates dropping from over 40% to under 5% on the Spanish test set. The dataset also enables attribution, identifying which synthesizer generated a clip, near-perfectly when the generator is known and with usable accuracy for unseen generators. If the benchmark holds up, it gives Spanish-speaking communities a foundation for detecting voice deepfakes and for standardizing future forensics research.

What carries the argument

The load-bearing object is the dataset itself, engineered for controlled generalization tests. HISPASpoof pairs real utterances from public corpora with synthetic versions of the exact same transcripts, removing content bias, and separates speakers and generators into seen/unseen groups so that test performance reflects genuine generalization to new voices and new synthesizers. The evaluation machinery is the battery of five detector architectures spanning feature-based (LFCC-GMM), image-based (MFCC-ResNet, Spec-ResNet, PaSST), and waveform-based (Wav2Vec2-AASIST) approaches, trained under four language conditions and scored by equal error rate for detection and accuracy/F1 for attribution.

What would settle it

Train a detector on HISPASpoof and evaluate it on Spanish real speech that has been passed through the same codec, resampling, and channel simulation as the synthetic samples, while keeping speakers and generators unseen. If equal error rate jumps toward chance levels, the original gains came from channel artifacts. Alternatively, train on HISPASpoof and test on a matched recording of the same speakers under identical microphone conditions.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that synthetic speech detection is language-sensitive, and a large-scale language-specific dataset can overcome that sensitivity. The authors introduce HISPASpoof with 6,241 real Spanish speech signals and 37,446 synthetic signals for detection (43,687 total), plus an attribution subset with 492,000 generated samples. Real speech is drawn from six Spanish accents (Peninsular, Argentinian, Colombian, Mexican, Chilean, Peruvian); synthetic speech is produced by six zero-shot TTS systems (ElevenLabs, F5-Spanish, FishSpeech, XTTS-v1.1, XTTS-v2, YourTTS). Using disjoint unseen speakers and two unseen generators, they evaluate five representative

Load-bearing premise

The benchmark's validity assumes that real speech from public corpora and the TTS-generated speech differ primarily in synthesis artifacts rather than in recording conditions, channels, or speaker-matching quality; if detectors exploit those incidental differences, the reported improvements would not measure deepfake-detection ability.

Editorial extensions

If this is right

  • Detectors trained only on English should not be trusted for Spanish speech; HISPASpoof-trained models cut EER from over 40% to under 5%.
  • Language-specific data is a practical lever: even classical GMM and small ResNet detectors become strong Spanish detectors with enough matched training data.
  • Attribution is feasible: known generators are identified near-perfectly, and unseen generators can be flagged as unknown with about 78% accuracy by the best model.
  • The benchmark's seen/unseen split makes it a reusable testbed for measuring generalization to new voices and new synthesis methods.
  • Cross-lingual generalization is asymmetric: training on Spanish and testing on English hurts less than the reverse, suggesting shared acoustic knowledge partially transfers.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: because real speech comes from different public corpora with different recording conditions and sampling rates, part of the reported gains may come from detectors learning channel or noise cues rather than synthesis artifacts; a channel-matched test would quantify this.
  • Editorial inference: the confusion between XTTS-v1 and XTTS-v2 in open-set attribution suggests architecture-level fingerprints; probing a continuum of XTTS checkpoints could reveal how attribution confidence degrades with model similarity.
  • Editorial inference: the accent-balanced design invites accent-sensitivity studies, e.g., training on Peninsular and testing on Mexican Spanish, to identify which synthesis artifacts are accent-invariant.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper introduces HISPASpoof, a large-scale Spanish-language dataset for synthetic speech detection and attribution. It contains real speech from six Spanish accents (taken from VoxPopuli, CIEMPIESS, and crowdsourced Latin American corpora) and synthetic speech from six zero-shot TTS systems, with a total of about 535k signals. The authors evaluate five detection methods (LFCC-GMM, MFCC-ResNet, Spec-ResNet, PaSST, Wav2Vec2-AASIST) under four training conditions: English ASVspoof2019, multilingual ODSS, the Spanish subset of ODSS, and HISPASpoof. They report EERs on test sets with unseen speakers and unseen generators, and also report closed-set and open-set attribution results. The central claims are that English-trained detectors generalize poorly to Spanish and that training on HISPASpoof substantially improves Spanish synthetic speech detection.

Significance. If the claims hold, HISPASpoof is a valuable public resource: it is the first large-scale Spanish dataset supporting both detection and attribution, it covers multiple accents and modern zero-shot TTS systems, and the paper provides baseline results for several representative detectors. The dataset release and the use of seen/unseen speaker-generator splits are strengths. However, the validity of the benchmark depends critically on whether the detection improvements measure synthesis artifacts rather than incidental recording-condition differences between the real and synthetic speech, and on the statistical reliability of the reported results given the small unseen-speaker/generator test set.

major comments (3)
  1. [§III-C, Tables I–II, Table VIII] The real and synthetic speech are not matched for recording conditions. Real speech comes from VoxPopuli (16 kHz), CIEMPIESS (16 kHz radio), and crowdsourced Latin American corpora (48 kHz), while synthetic speech is generated cleanly at 16–44.1 kHz. No resampling, filtering, channel normalization, or noise augmentation is described. A detector could therefore succeed by exploiting bandwidth, noise floor, or channel statistics rather than synthesis artifacts. This concern is reinforced by the unusually low EERs on the unseen test set in Table VIII (LFCC-GMM 1.57%, Spec-ResNet 0.72%), especially for a cross-generator task. To support the central claim, the authors should add channel-matched controls, e.g., resampling all audio to a common rate, adding noise/channel augmentation, and reporting performance on low-band versus full-band features or on a real-speech-only/channel-shift control.
  2. [§V, §VII, Tables V–X] All detection and attribution results are reported as single-run EER or accuracy values with no confidence intervals, error bars, or significance tests. The HISPASpoof unseen test set is built from only six unseen speakers and two held-out generators (Section III-C), so the reported values such as 0.72% EER are point estimates with potentially large uncertainty. The paper should provide error bars across multiple training runs or bootstrap over speakers/generators, and should report per-speaker and per-generator breakdowns for the key comparisons, at least for the HISPASpoof-trained models in Table VIII.
  3. [§VI, Experiment 2 (Open-Set Attribution)] The open-set threshold δ is selected on a 10% held-out portion of the test set, but the manuscript does not state whether the final metrics in Table X are computed on the remaining 90% or on the entire test set. If the entire test set is used, the threshold selection leaks test information and the open-set attribution numbers are optimistically biased. The authors should define a dedicated validation split for threshold selection and evaluate only on a fully disjoint test split, or clearly report both the threshold-tuning portion and the final evaluation portion.
minor comments (6)
  1. [§III-C] Typo: "contains both synthetic and and real speech" should read "contains both synthetic and real speech."
  2. [§I] Typo: "V oice Cloning" should be "Voice Cloning."
  3. [References] Reference [47] is a duplicate of reference [19] (same title, same authors, same venue). Please merge or differentiate.
  4. [§III-C, Table II] Inconsistent naming of FishSpeech vs Fish-Speech across the text and tables; please unify.
  5. [§III-C] The relation between the detection subset (which uses exact transcripts of real signals) and the attribution subset (which uses ChatGPT-generated text) should be clarified, since content mismatch can affect both detection and attribution generalization.
  6. [Fig. 1] The confusion matrix figure is referenced but the axes and normalization are not fully described in the text; please ensure the figure has clear labels and captions.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HISPASpoof is an empirical benchmark paper with external comparisons and disclosed calibration.

full rationale

The paper's central claim is empirical: constructing a Spanish dataset and showing that training on it improves Spanish synthetic-speech detection relative to training on English or multilingual datasets. There is no derivation chain in which an output quantity is defined in terms of an input quantity or in which a fitted parameter is renamed as a prediction. The real and synthetic speech are assembled from independent public corpora and six named TTS systems; detection and attribution results are measured on held-out speakers and generators and compared against the external ASVspoof2019 and ODSS benchmarks. The only tunable constant is the open-set attribution threshold δ, selected on a disclosed 10% held-out portion of the test set; this is a transparent calibration step, not a parameter fitted to the training set and then presented as a prediction, and it does not affect the detection claims. Self-citations appear only as background/method references or as the authors' own code repository; none is load-bearing for the dataset's validity or for a uniqueness theorem. Thus, no step reduces the paper's results to its own inputs.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The central contributions are empirical, so the ledger's main entries are domain assumptions about representativeness and confounding, plus one fitted attribution threshold. The paper introduces no theoretical entities.

free parameters (1)
  • Open-set attribution threshold delta = not reported
    Selected on a 10% held-out portion of the test set by maximizing balanced true positive rates (Section VI, Experiment 2); this is a free parameter fitted to the test distribution.
assumptions (4)
  • domain assumption Synthetic speech artifacts dominate recording-condition differences between real and synthetic clips
    Real clips come from public corpora (radio, crowdsourced) while TTS output is clean; if false, detection gains reflect channel mismatch, not forgery cues. Invoked implicitly in Section III-C.
  • domain assumption Six zero-shot TTS systems and their default settings represent modern Spanish deepfake generators
    Section III-C and Table II: the dataset's scope is defined by these six systems; there is no evidence they cover the broader space of synthesis tools.
  • domain assumption Four speakers per accent, with six unseen speakers in the test set, support generalization claims about Spanish accents
    Section III-C states 'The dataset includes 4 speakers per accent'; with 24 total and 6 unseen speakers, accent-level conclusions rest on very small speaker samples.
  • domain assumption Exact-transcript synthesis controls content bias
    Section III-C: 'Synthetic speech was generated using the exact transcript of each real signal to avoid bias with respect to speech content', but this does not control acoustic conditions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HISPASpoof: A New Dataset For Spanish Speech Forensics." pith.science (2026). https://pith.science/paper/D3JHWX7V

@misc{pith2026250909155,
  author       = {Pith},
  title        = {Pith review of: HISPASpoof: A New Dataset For Spanish Speech Forensics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D3JHWX7V}},
  note         = {Machine review of arXiv:2509.09155}
}
read the original abstract

Zero-shot Voice Cloning (VC) and Text-to-Speech (TTS) methods have advanced rapidly, enabling the generation of highly realistic synthetic speech and raising serious concerns about their misuse. While numerous detectors have been developed for English and Chinese, Spanish-spoken by over 600 million people worldwide-remains underrepresented in speech forensics. To address this gap, we introduce HISPASpoof, the first large-scale Spanish dataset designed for synthetic speech detection and attribution. It includes real speech from public corpora across six accents and synthetic speech generated with six zero-shot TTS systems. We evaluate five representative methods, showing that detectors trained on English fail to generalize to Spanish, while training on HISPASpoof substantially improves detection. We also evaluate synthetic speech attribution performance on HISPASpoof, i.e., identifying the generation method of synthetic speech. HISPASpoof thus provides a critical benchmark for advancing reliable and inclusive speech forensics in Spanish.

Figures

Figures reproduced from arXiv: 2509.09155 by the authors.

Figure 1
Figure 1. shows normalized confusion matrices of all methods. [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 1
Figure 1. Confusion matrices for Attribution Experiment 2: the open-set scenario. [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 2 linked inside Pith

  1. [2]

    J. P. Gallego-Arjona,F5-spanish huggingface repository, https : / / huggingface.co/jpgallegoar/F5-Spanish

  2. [3]

    Liao et al.,Fish-speech: Leveraging large language models for advanced multilingual text-to-speech synthesis, 2024

    S. Liao et al.,Fish-speech: Leveraging large language models for advanced multilingual text-to-speech synthesis, 2024. [Online]. Available: https://doi. org/10.48550/arXiv.2411.01156

  3. [4]

    co / coqui / XTTS-v1

    Coqui.ai,Xtts-v1 huggingface repository, https : / / huggingface . co / coqui / XTTS-v1

  4. [5]

    co / coqui / XTTS-v2

    Coqui.ai,Xtts-v2 huggingface repository, https : / / huggingface . co / coqui / XTTS-v2

  5. [6]

    YourTTS: Towards Zero-Shot Multi-Speaker TTS and Zero-Shot V oice Conversion for Everyone,

    E. Casanova et al., “YourTTS: Towards Zero-Shot Multi-Speaker TTS and Zero-Shot V oice Conversion for Everyone,”Proceedings of the In- ternational Conference on Machine Learning, pp. 2709–2720, Jul. 2022, Baltimore

  6. [7]

    Flitter and S

    E. Flitter and S. Cowley,Voice deepfakes are coming for your bank balance. the new york times, Aug. 2023. [Online]. Available: https://www.nytimes. com/2023/08/30/%20business/voice-deepfakes-bank-scams.html

  7. [8]

    Nguyen,A couple in Canada were reportedly scammed out of $21,000 after getting a call from an AI-generated voice pretending to be their son

    B. Nguyen,A couple in Canada were reportedly scammed out of $21,000 after getting a call from an AI-generated voice pretending to be their son. The New York Times, Mar. 2023. [Online]. Available: https : / / www. businessinsider.com/couple-canada-reportedlylost-21000-in-ai-generated- voice-scam-2023-3

  8. [9]

    Mollman,Scammers are using voice-cloning A.I

    B. Mollman,Scammers are using voice-cloning A.I. tools to sound like victims’ relatives in desperate need of financial help. It’s working. The New York Times, Mar. 2023. [Online]. Available: https://fortune.com/2023/03/ %2005/scammers-ai-voice-cloning-tricking-victims-sound-likerelatives- needing-money/

Show all 48 references
  1. [10]

    Verma,They thought loved ones were calling for help

    P. Verma,They thought loved ones were calling for help. It was an AI scam. The Washington Post, Mar. 2023. [Online]. Available: https://www. washingtonpost.com/%20technology/2023/03/05/ai-voice-scam/

  2. [11]

    Neural V oice Cloning with a Few Samples,

    S. Arik et al., “Neural V oice Cloning with a Few Samples,”Advances in Neural Information Processing Systems, vol. 31, Dec. 2018, Montr ´eal, Canada

  3. [12]

    LIMMITS’24: Multi-Speaker, Multi-Lingual Indic TTS with V oice Cloning,

    A. Singh et al., “LIMMITS’24: Multi-Speaker, Multi-Lingual Indic TTS with V oice Cloning,”Proceedings of the International Conference on Acoustics, Speech, and Signal Processing Workshops, pp. 61–62, Apr. 2024, Seoul, Korea. [13]Deep learning for siri’s voice: On-device deep m...

  4. [14]

    Schindel,The Rise of Synthetic Audio in Documentary Films, https : / / immerse

    D. Schindel,The Rise of Synthetic Audio in Documentary Films, https : / / immerse . news / the - rise - of - synthetic - audio - in - documentary - films - 860e943f3503, Apr. 2022

  5. [15]

    Wakefield,Deepfake presidents used in Russia-Ukraine war, https : / / www.bbc.com/news/technology-60780142, Mar

    J. Wakefield,Deepfake presidents used in Russia-Ukraine war, https : / / www.bbc.com/news/technology-60780142, Mar. 2022

  6. [16]

    Mack,This PSA About Fake News From Barack Obama Is Not What It Appears, https://www.buzzfeednews.com/article/davidmack/obama-fake- news-jordan-peele-psa-video-buzzfeed, Apr

    D. Mack,This PSA About Fake News From Barack Obama Is Not What It Appears, https://www.buzzfeednews.com/article/davidmack/obama-fake- news-jordan-peele-psa-video-buzzfeed, Apr. 2018

  7. [17]

    Smith,Goldman Sachs, Ozy Media and a $40 Million Conference Call Gone Wrong, https://www.nytimes.com/2021/09/26/business/media/ozy- media-goldman-sachs.html, Sep

    B. Smith,Goldman Sachs, Ozy Media and a $40 Million Conference Call Gone Wrong, https://www.nytimes.com/2021/09/26/business/media/ozy- media-goldman-sachs.html, Sep. 2021

  8. [18]

    PS3DT: Synthetic Speech Detection Using Patched Spectrogram Transformer,

    A. K. S. Yadav et al., “PS3DT: Synthetic Speech Detection Using Patched Spectrogram Transformer,”Proceedings of the IEEE International Confer- ence on Machine Learning and Applications, pp. 496–503, 2023, Florida, USA

  9. [20]

    Efficient Train- ing of Audio Transformers with Patchout,

    K. Koutini, J. Schl ¨uter, H. Eghbal-zadeh, and G. Widmer, “Efficient Train- ing of Audio Transformers with Patchout,”Proceedings of Interspeech, pp. 2753–2757, Sep. 2022, Incheon, Korea

  10. [21]

    Towards end-to-end synthetic speech detection,

    G. Hua, A. B. J. Teoh, and H. Zhang, “Towards end-to-end synthetic speech detection,”IEEE Signal Processing Letters, pp. 1265–1269, Jun. 2021

  11. [22]

    ASVspoof 2019: The 3rd Automatic Speaker Verifi- cation Spoofing and Countermeasures Challenge database,

    J. Yamagishi et al., “ASVspoof 2019: The 3rd Automatic Speaker Verifi- cation Spoofing and Countermeasures Challenge database,”University of Edinburgh, The Centre for Speech Technology Research, Mar. 2019

  12. [23]

    ASVspoof 2021: Accelerating Progress in Spoofed and Deepfake Speech Detection,

    J. Yamagishi et al., “ASVspoof 2021: Accelerating Progress in Spoofed and Deepfake Speech Detection,”Proceedings of the Automatic Speaker Verification and Spoofing Countermeasures Challenge, pp. 47–54, Sep. 2021

  13. [24]

    FoR: A Dataset for Synthetic Speech Detec- tion,

    R. Reimao and V . Tzerpos, “FoR: A Dataset for Synthetic Speech Detec- tion,”Proceedings of the International Conference on Speech Technology and Human-Computer Dialogue, pp. 1–10, Oct. 2019, Timisoara, Romania

  14. [25]

    TIMIT-TTS: A Text-to-Speech Dataset for Multimodal Synthetic Media Detection,

    D. Salvi et al., “TIMIT-TTS: A Text-to-Speech Dataset for Multimodal Synthetic Media Detection,”IEEE Access, vol. 11, pp. 50 851–50 866, May 2023

  15. [26]

    Does Audio Deepfake Detection Generalize?

    N. M. M ¨uller et al., “Does Audio Deepfake Detection Generalize?” arXiv:2203.16263, Aug. 2024

  16. [27]

    CFAD: A Chinese dataset for fake audio detection,

    H. Ma et al., “CFAD: A Chinese dataset for fake audio detection,”Speech Communication, vol. 164, p. 103 122, 2024

  17. [28]

    ASVspoof 2019: Future Horizons in Spoofed and Fake Audio Detection,

    M. Todisco et al., “ASVspoof 2019: Future Horizons in Spoofed and Fake Audio Detection,”Proceedings of the ISCA Interspeech, pp. 1008–1012, Sep. 2019, Graz, Austria

  18. [29]

    Deep Residual Neural Networks for Audio Spoofing Detection,

    M. Alzantot, Z. Wang, and M. B. Srivastava, “Deep Residual Neural Networks for Audio Spoofing Detection,”Proceedings of Interspeech, pp. 1078–1082, Sep. 2019, Graz, Austria

  19. [30]

    FGSSAT : Unsupervised Fine-Grain Attribution of Unknown Speech Synthesizers Using Transformer Networks,

    K. Bhagtani et al., “FGSSAT : Unsupervised Fine-Grain Attribution of Unknown Speech Synthesizers Using Transformer Networks,”Proceedings of the IEEE Asilomar Conference on Signals, Systems, and Computers, pp. 1135–1140, 2023, Pacific Grove, CA

  20. [31]

    Synthetic Speech Attribution using Self Supervised Audio Spectrogram Transformer,

    A. K. S. Yadav, E. Bartusiak, K. Bhagtani, and E. J. Delp, “Synthetic Speech Attribution using Self Supervised Audio Spectrogram Transformer,” Proceedings of the IS&T Media Watermarking, Security, and Forensics Conference, Electronic Imaging Symposium, Jan. 2023, San Francisco, CA

  21. [32]

    Instituto Cervantes,El espa ˜nol en el mundo 2024, https://cvc.cervantes.es/ lengua/anuario/anuario 24/default.htm, 2024

  22. [33]

    An Open Dataset of Synthetic Speech,

    A. Yaroshchuk et al., “An Open Dataset of Synthetic Speech,”Proceedings of the IEEE International Workshop on Information Forensics and Security, pp. 1–6, Dec. 2023, N¨urnberg, Germany

  23. [34]

    Crowdsourcing Latin American Spanish for low- resource text-to-speech,

    A. Guevara-Rukoz et al., “Crowdsourcing Latin American Spanish for low- resource text-to-speech,”Proceedings of the Twelfth Language Resources and Evaluation Conference, pp. 6504–6513, 2020, Marseille, France

  24. [35]

    CIEMPIESS: A New Open-Sourced Mexican Spanish Radio Corpus,

    C. D. Hernandez Mena and A. Herrera Camacho, “CIEMPIESS: A New Open-Sourced Mexican Spanish Radio Corpus,”Proceedings of the Ninth International Conference on Language Resources and Evaluation, 2014, Reykjavik, Iceland

  25. [36]

    V oxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation,

    C. Wang et al., “V oxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation,” Proceedings of the Annual Meeting of the Association for Computational Linguistics, pp. 993–1003, 2021

  26. [37]

    An Overview of Recent Work in Media Forensics: Methods and Threats,

    K. Bhagtani et al., “An Overview of Recent Work in Media Forensics: Methods and Threats,”arXiv:2204.12067, May 2022

  27. [38]

    FairSSD: Understanding Bias in Synthetic Speech Detectors,

    A. K. S. Yadav et al., “FairSSD: Understanding Bias in Synthetic Speech Detectors,”Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pp. 4418–4428, 2024

  28. [39]

    Replay and Synthetic Speech Detection with Res2Net Archi- tecture,

    X. Li et al., “Replay and Synthetic Speech Detection with Res2Net Archi- tecture,”Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 6354–6358, Jun. 2021, Toronto, Canada

  29. [40]

    Detection of Copy-Move Forgery in Audio Signal with Mel Frequency and Delta-Mel Frequency Kepstrum Coeffi- cients,

    F. Akdeniz and Y . Becerikli, “Detection of Copy-Move Forgery in Audio Signal with Mel Frequency and Delta-Mel Frequency Kepstrum Coeffi- cients,”Proceedings of the Innovations in Intelligent Systems and Appli- cations Conference, pp. 1–6, Oct. 2021, Elazig, Turkey

  30. [41]

    ASSD: Synthetic Speech Detection in the AAC Compressed Domain,

    A. K. Singh Yadav et al., “ASSD: Synthetic Speech Detection in the AAC Compressed Domain,”Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, pp. 1–5, Jun. 2023, Rhodes Island, Greece

  31. [42]

    Transformer Ensemble for Synthesized Speech Detection,

    E. R. Bartusiak, K. Bhagtani, A. K. S. Yadav, and E. J. Delp, “Transformer Ensemble for Synthesized Speech Detection,”Proceedings of the Asilomar Conference on Signals, Systems, and Computers, pp. 1100–1105, Oct. 2023, Pacific Grove, California, USA

  32. [43]

    DSV AE: Disentangled Representation Learning for Synthetic Speech Detection,

    A. K. Singh Yadav et al., “DSV AE: Disentangled Representation Learning for Synthetic Speech Detection,”Proceedings of the IEEE International Conference on Machine Learning and Applications, pp. 472–479, 2023, Florida, USA

  33. [44]

    Mdrt: Multi-domain synthetic speech localiza- tion,

    A. K. Singh Yadav et al., “Mdrt: Multi-domain synthetic speech localiza- tion,”Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing, pp. 11 171–11 175, 2024, Seoul, South Korea

  34. [45]

    Synthesized speech attribution using the patchout spectrogram attribution transformer,

    K. Bhagtani et al., “Synthesized speech attribution using the patchout spectrogram attribution transformer,”Proceedings of the ACM Workshop on Information Hiding and Multimedia Security, pp. 157–162, Jun. 2023, Chicago, IL, USA

  35. [46]

    Improving fairness in synthetic speech detectors,

    A. K. Singh Yadav, K. Bhagtani, P. Bestagini, and E. J. Delp, “Improving fairness in synthetic speech detectors,”Proceedings of IEEE Asilomar Con- ference on Signals, Systems, and Computers, pp. 362–366, 2024, Asilomar, CA

  36. [47]

    Automatic Speaker Verification Spoofing and Deepfake Detection Using Wav2vec 2.0 and Data Augmentation,

    H. Tak et al., “Automatic Speaker Verification Spoofing and Deepfake Detection Using Wav2vec 2.0 and Data Augmentation,”Proceedings of the Speaker and Language Recognition Workshop, Odyssey, pp. 112–119, Jul. 2022, Beijing, China

  37. [48]

    Casanova,Yourtts github repository, https : / / github

    E. Casanova,Yourtts github repository, https : / / github . com / Edresson / YourTTS/

  38. [49]

    Improving Language Understanding by Generative Pre-training,

    A. Radford, K. Narasimhan, T. Salimans, and I. Sutskever, “Improving Language Understanding by Generative Pre-training,”OpenAI, Feb. 2019

  39. [50]

    ChatGPT: A comprehensive review on background, applications, key challenges, bias, ethics, limitations and future scope,

    P. P. Ray, “ChatGPT: A comprehensive review on background, applications, key challenges, bias, ethics, limitations and future scope,”Internet of Things and Cyber-Physical Systems, vol. 3, Apr. 2023

  40. [51]

    Risques,Hispaspoof: A new dataset for spanish synthetic speech foren- sics, https://gitlab.com/viper-purdue/s3d-spanish-syn-speech-det.git

    M. Risques,Hispaspoof: A new dataset for spanish synthetic speech foren- sics, https://gitlab.com/viper-purdue/s3d-spanish-syn-speech-det.git

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.