Pith. sign in

REVIEW 4 major objections 4 minor 34 references

Low-Resource Text-to-Speech Synthesis Using Noise-Augmented Training of ForwardTacotron

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Five minutes of speech beats zero-shot TTS baselines on voice similarity.

desk verdict A useful low-resource TTS recipe, but the abstract oversells the five-minute naturalness claim; worth reviewing with revisions. read the letter →

arxiv 2501.05976 v1 pith:IMHMO2UX submitted 2025-01-10 eess.AS

Kishor Kayyar Lakshminarayana , Frank Zalkow , Christian Dittmar , Nicola Pia , Emanuel A.P. Habets This is my paper · ORCID
classification eess.AS
keywords text-to-speechlow-resourceTTSForwardTacotronnoiseaugmentationspeakersimilaritybinnedsamplingzero-shot
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

This paper tries to show that high-quality text-to-speech for a new speaker does not require tens or hundreds of speaker recordings. It claims that training a multi-speaker ForwardTacotron acoustic model on just four high-quality speakers plus five minutes of target-speaker data, after splitting the target audio into short segments and adding white Gaussian noise, produces speech that listeners rate closer to the target voice than the zero-shot HierSpeech++ and the low-resource AdapterMix, with comparable naturalness. If true, the result lowers the data and compute barrier for adding voices to TTS systems and suggests that the speaker set used for pretraining can be small and high-quality rather than large and uneven. The paper also claims that increasing the target-speaker data to twenty minutes closes the naturalness gap to HierSpeech++ while keeping the speaker-similarity advantage.

What carries the argument

The load-bearing object is the modified multi-speaker ForwardTacotron acoustic model, which maps text to mel-spectrograms and is trained on four high-resource speakers plus one low-resource speaker. The mechanism has three parts: short-sample segmentation (either short utterances or ASR-based splitting with WhisperX), noise augmentation that creates several constant-20 dB white-Gaussian-noise copies of each scarce sample with a dedicated 'noisy' condition embedding, and two sampling techniques, weighted sampling and binned sampling, that make the underrepresented target speaker appear often enough in gradients. The clean target-speaker condition embedding is used at inference time, so the model learns to treat noise as a training-time regularizer rather than as part of the voice.

What would settle it

Run the proposed 5-minute training on long-form recordings without splitting and measure naturalness and similarity against the split version; the paper itself reports that long unsplit training sentences substantially degrade quality, so a result that does not degrade would contradict the claimed mechanism.

Watch

Extended reading notes

Core claim

The central discovery is that a simple, low-complexity extension of ForwardTacotron can synthesize a target speaker's voice from very little data, provided the scarce samples are cut into short segments, duplicated with noise added at a fixed SNR, and presented through sampling strategies that prevent the four high-resource speakers from dominating training. With five minutes of low-resource data the model attains higher speaker-similarity ratings than HierSpeech++ and AdapterMix, and with twenty minutes its naturalness matches HierSpeech++ while keeping the similarity advantage. The authors attribute the effect to separating clean and noisy condition embeddings for the target speaker, which lets the model learn speaker identity from both versions without conflating noise with identity.

Load-bearing premise

The method assumes that the target speaker's audio can be divided into short segments at speech pauses, either because short recordings already exist or because an ASR tool like WhisperX can find those pauses reliably.

Editorial extensions

If this is right

  • A TTS voice for a new speaker can be added from about five minutes of short utterances, with only four high-quality anchor voices, instead of requiring a large many-speaker training corpus.
  • With twenty minutes of target data the synthesized speech becomes as natural as a zero-shot system while staying closer to the target voice in subjective similarity.
  • The approach uses fewer parameters (43M) than HierSpeech++ (108.5M) or AdapterMix (52M), so it is a comparatively lightweight route to personalized TTS.
  • The ablation results imply both noise augmentation and binned sampling are needed together for consistent gains, since each alone helps different datasets differently.
  • Because the method requires retraining, it is a training-time adaptation approach rather than a zero-shot inference method.

Reading between the lines

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

  • If the English results carry over, the same four-anchor recipe could be applied to any language where short recorded utterances or reliable ASR segmentation exist; the paper states this as a possibility but does not test it.
  • The constant-SNR noise copies may be doing double duty: increasing sample count and acting as a regularizer that keeps the model from memorizing the few clean target utterances. This could be tested by ablating noise level and number of copies.
  • A practical extension would be an incremental voice-addition service: keep the four anchor speakers fixed and adapt only the condition embedding and sampling weights for each new speaker.
  • The dependence on speech-pause splitting suggests the method would need an ASR front end for each new language, and where such a front end is unavailable, short-sentence collection remains necessary.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes a low-resource text-to-speech method based on ForwardTacotron, using four high-quality high-resource speakers plus as little as five minutes of target-speaker data. The method adds white Gaussian noise augmentation with a separate conditioning ID, splits the low-resource data into short segments, and uses weighted and binned sampling to reduce the class-imbalance problem. Objective metrics (MCD-DTW and ECAPA-TDNN cosine similarity) with 95% confidence intervals are reported for a male speaker (TC-Star) and a female speaker (Hi-Fi-TTS-92), and subjective MOS tests for naturalness and speaker similarity are reported for the male speaker. The authors compare against HierSpeech++ and AdapterMix, claiming improved speaker similarity with comparable naturalness at five minutes and that twenty minutes closes the naturalness gap.

Significance. If the claims hold, this would be a practically useful low-complexity alternative to zero-shot and finetuning-based low-resource TTS, requiring only four high-quality speakers and twenty minutes of target data, with fewer parameters than the compared baselines. The paper's strengths include objective metrics with confidence intervals, ablation studies isolating the contributions of noise augmentation and binned sampling, and comparison against independently published baselines. The main weakness is that the headline five-minute claims are only partially supported by the subjective evidence, and the abstract overstates the naturalness comparability for the five-minute configuration.

major comments (4)
  1. [Abstract and Section IV-B] The abstract's claim of 'maintaining comparable naturalness' for the five-minute configuration is contradicted by the paper's own results: Section IV-B states that versions using less than 20 minutes of data had lower naturalness ratings, and Figure 2 confirms that the 5-minute model is below HierSpeech++. Only the 20-minute model achieves comparable naturalness. Please either restrict the claim to the 20-minute setting or provide statistical evidence that the 5-minute naturalness difference is perceptually insignificant.
  2. [Section IV-B, Figure 2] The central speaker-similarity claim at five minutes rests on a 0.25 MOS advantage over HierSpeech++, reported without significance tests, confidence intervals, or per-listener and per-item analysis. With 15 listeners and 20 samples, the boxplots in Figure 2 show substantial overlap. The objective ECAPA-TDNN cosine similarity in Table II supports the direction, but the subjective claim of 'improved speaker similarity' for 5-minute data is not statistically supported. Please add significance testing (e.g., paired bootstrap or a mixed-effects model) or soften the claim.
  3. [Section IV-B] Subjective evaluations were conducted for the male speaker only, yet the abstract and conclusion generalize to female speakers and to new languages. The objective results for Hi-Fi-TTS-92 are encouraging, but human-level similarity and naturalness for the female speaker remain unvalidated. Please either add subjective tests for the female speaker or explicitly limit the human-evaluation claims to the male speaker.
  4. [Section IV-C] The method's reliance on splitting long utterances at speech pauses is load-bearing: the paper reports that training on long sentences without splitting substantially degrades quality. The conclusion nevertheless claims applicability to other languages without any multilingual experiment, and WhisperX coverage for arbitrary languages is not guaranteed. Please either provide cross-language evidence or remove and hedge the generalization claim.
minor comments (4)
  1. [Section III] The choices of five noise augmentations, 20 dB SNR, and a weighting factor of six are justified only by reference to preliminary experiments; a brief description of those experiments or a pointer to supplemental material would improve reproducibility.
  2. [Section IV-B] The sentence 'Both listening tests were done independently' is ambiguous; please specify whether the naturalness and similarity tests used different listener groups or the same listeners in separate sessions.
  3. [Section IV-C and Table II] The ablation results show dataset-dependent effects for noise augmentation and binned sampling; the text explains this well, but grouping the ablation rows separately from the main results in Table II would improve readability.
  4. [Section V] The conclusion's statement that the approach 'can also be applied to other languages' is speculative given that all experiments are in English; please rephrase this as a future direction rather than a demonstrated property.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the claims are empirical comparisons against independently published baselines and do not reduce to their inputs or to self-citations.

full rationale

The paper's central claims are empirical: a modified ForwardTacotron acoustic model, trained with four high-resource speakers plus 1/5/20 minutes of target-speaker data, is compared against HierSpeech++ (an external pre-trained system) and AdapterMix (independently implemented and finetuned by the authors). No derivation chain is present in which an output quantity is defined as its own input. The noise augmentation, weighted sampling, and binned sampling are described as concrete training procedures, and the objective metrics (MCD-DTW and ECAPA-TDNN cosine similarity) and subjective MOS tests are measured on held-out sentences. The cited prior work [18] introduced noise augmentation in a single-speaker autoregressive setting and is used only as a starting point; the current paper states its own extension to multi-speaker non-autoregressive training with constant-SNR WGN and reports its own ablations. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no known empirical pattern is merely relabeled. The limitations noted in the paper, such as the lack of significance testing for the 0.25 MOS similarity gain, the restriction of listening tests to one male speaker, and the informal nature of the long-sentence splitting check, are evidence-quality concerns rather than circularity. The central comparison is self-contained against external baselines, so the circularity score is 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method is an empirical configuration of known components. The main added commitments are the noise-conditioning training scheme, the fixed ~1000 sentence budget achieved through augmentation, and the sufficiency of four high-resource speakers. No new physical entities or mathematical axioms are introduced.

free parameters (5)
  • Noise augmentation count = 5 for 5 and 20 minute subsets; 10 for 1 minute subset
    Chosen to reach roughly 1000 low-resource training sentences; determined in preliminary stability experiments (Section III).
  • WGN SNR level = 20 dB
    Constant SNR for all augmented samples, selected in preliminary experiments (Section III).
  • Low-resource training sentence threshold = about 1000 sentences
    Authors state that around a thousand LR sentences provided training stability, and this drives the number of augmentations (Section III).
  • Weighting factor for 1 minute subset = 6
    Used to rebalance sampling for the one-minute subset; authors state higher factors did not improve performance (Section III).
  • Training steps and batch size = 300K steps, batch size 32
    Fixed termination criterion for all training instances (Section IV).
assumptions (4)
  • domain assumption Adding WGN to low-resource samples and conditioning on a noisy cond-ID during training, while using the clean cond-ID at inference, transfers to clean speech synthesis.
    Core mechanism of the method (Section II-B); supported only by ablations on two speakers, not by a formal argument.
  • domain assumption Short-segment training data for the low-resource speaker is necessary, and an ASR model such as WhisperX is assumed available to segment at pauses.
    Section III and Section IV-C confirm that splitting works and that long sentences without splitting degrade quality. For new languages, ASR coverage is not guaranteed.
  • domain assumption Four high-quality high-resource speakers are sufficient to provide the phonetic and contextual coverage needed for arbitrary text for a new low-resource speaker.
    Claimed in Section I and Section III, but tested with only two low-resource speakers, and two of the four HR speakers are proprietary.
  • domain assumption Unbalanced speaker distribution can be mitigated with weighted and binned sampling without degrading high-resource speaker quality.
    Section II-B draws an analogy to class imbalance; ablations show the effects differ across datasets, so the general mechanism is not fully characterized.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Low-Resource Text-to-Speech Synthesis Using Noise-Augmented Training of ForwardTacotron." pith.science (2026). https://pith.science/paper/IMHMO2UX

@misc{pith2026250105976,
  author       = {Pith},
  title        = {Pith review of: Low-Resource Text-to-Speech Synthesis Using Noise-Augmented Training of ForwardTacotron},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IMHMO2UX}},
  note         = {Machine review of arXiv:2501.05976}
}
read the original abstract

In recent years, several text-to-speech systems have been proposed to synthesize natural speech in zero-shot, few-shot, and low-resource scenarios. However, these methods typically require training with data from many different speakers. The speech quality across the speaker set typically is diverse and imposes an upper limit on the quality achievable for the low-resource speaker. In the current work, we achieve high-quality speech synthesis using as little as five minutes of speech from the desired speaker by augmenting the low-resource speaker data with noise and employing multiple sampling techniques during training. Our method requires only four high-quality, high-resource speakers, which are easy to obtain and use in practice. Our low-complexity method achieves improved speaker similarity compared to the state-of-the-art zero-shot method HierSpeech++ and the recent low-resource method AdapterMix while maintaining comparable naturalness. Our proposed approach can also reduce the data requirements for speech synthesis for new speakers and languages.

Figures

Figures reproduced from arXiv: 2501.05976 by the authors.

Figure 1
Figure 1. Block diagram of multi-speaker ForwardTacotron with the proposed extension. The [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Subjective evaluation results across naturalness and speaker similarity for the low-resource speaker. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 30 canonical work pages

  1. [1]

    FastSpeech: Fast, robust and controllable text-to-speech,

    Y . Ren, Y . Ruan, X. Tan, T. Qin, S. Zhao, Z. Zhao, and T. Liu, “FastSpeech: Fast, robust and controllable text-to-speech,” in Advances in Neural Information Processing Systems , Vancouver, Canada, 2019, pp. 3165–3174

    work page 2019

  2. [2]

    FastSpeech 2: Fast and high-quality end-to-end text-to-speech,

    Y . Ren, C. Hu, X. Tan, T. Qin, S. Zhao, Z. Zhao, and T. Liu, “FastSpeech 2: Fast and high-quality end-to-end text-to-speech,” in Proc. Intl. Conf. on Learning Representations , virtual, Austria, 2021

    work page 2021

  3. [3]

    ForwardTacotron,

    C. Sch ¨afer, O. McCarthy, and contributors, “ForwardTacotron,” https: //github.com/as-ideas/ForwardTacotron, 2020

    work page 2020

  4. [4]

    Fastpitch: Parallel text-to-speech with pitch prediction,

    A. Ła ´ncucki, “Fastpitch: Parallel text-to-speech with pitch prediction,” in Proc. IEEE Intl. Conf. on Acoustics, Speech and Signal Processing (ICASSP), 2021, pp. 6588–6592

    work page 2021

  5. [5]

    A lightweight neural TTS system for high-quality German speech synthesis,

    P. Govalkar, A. Mustafa, N. Pia, J. Bauer, M. Yurt, Y . ¨Ozer, and C. Dittmar, “A lightweight neural TTS system for high-quality German speech synthesis,” in Proc. ITG Conf. on Speech Communication , 2021

    work page 2021

  6. [6]

    YourTTS: Towards zero-shot multi-speaker TTS and zero- shot voice conversion for everyone,

    E. Casanova, J. Weber, C. D. Shulby, A. C. Junior, E. G ¨olge, and M. A. Ponti, “YourTTS: Towards zero-shot multi-speaker TTS and zero- shot voice conversion for everyone,” in Proc. Intl. Conf. on Machine Learning. PMLR, 2022, pp. 2709–2720

    work page 2022

  7. [7]

    Hierspeech++: Bridging the gap between semantic and acoustic representation of speech by hierarchical variational inference for zero-shot speech synthesis,

    S.-H. Lee, H.-Y . Choi, S.-B. Kim, and S.-W. Lee, “Hierspeech++: Bridging the gap between semantic and acoustic representation of speech by hierarchical variational inference for zero-shot speech synthesis,” arXiv preprint arXiv:2311.12454 , 2023

    arXiv 2023

  8. [8]

    Generalizable zero-shot speaker adaptive speech synthesis with disentangled representations,

    W. Wang, Y . Song, and S. Jha, “Generalizable zero-shot speaker adaptive speech synthesis with disentangled representations,” in Proc. Interspeech, 2023, pp. 4454–4458

    work page 2023

Show all 34 references
  1. [9]

    Adapter-based extension of multi-speaker text-to-speech model for new speakers,

    C.-P. Hsieh, S. Ghosh, and B. Ginsburg, “Adapter-based extension of multi-speaker text-to-speech model for new speakers,” in Proc. Inter- speech, 2023, pp. 3028–3032

    2023

  2. [10]

    AdapterMix: Exploring the efficacy of mixture of adapters for low-resource TTS adaptation,

    A. Mehrish, A. Ramesh Kashyap, L. Yingting, N. Majumder, and S. Poria, “AdapterMix: Exploring the efficacy of mixture of adapters for low-resource TTS adaptation,” in Proc. Interspeech , 2023, pp. 4284– 4288

    2023

  3. [11]

    CSTR VCTK Corpus: En- glish multi-speaker corpus for CSTR voice cloning toolkit (version 0.92), [sound]

    J. Yamagishi, C. Veaux, and K. MacDonald, “CSTR VCTK Corpus: En- glish multi-speaker corpus for CSTR voice cloning toolkit (version 0.92), [sound].” University of Edinburgh. The Centre for Speech Technology Research (CSTR), 2019

    2019

  4. [12]

    LibriTTS: A corpus derived from LibriSpeech for text-to- speech,

    H. Zen, R. Clark, R. J. Weiss, V . Dang, Y . Jia, Y . Wu, Y . Zhang, and Z. Chen, “LibriTTS: A corpus derived from LibriSpeech for text-to- speech,” in Proc. Interspeech, 2019, pp. 1526–1530

    2019

  5. [13]

    StrawNet: Self-training WaveNet for TTS in low-data regimes,

    M. Sharma, T. Kenter, and R. Clark, “StrawNet: Self-training WaveNet for TTS in low-data regimes,” in Proc. Interspeech , 2020, pp. 3550– 3554

    2020

  6. [14]

    Non-autoregressive TTS with explicit duration modelling for low-resource highly expressive speech,

    R. Shah, K. Pokora, A. Ezzerg, V . Klimkov, G. Huybrechts, D. K. Bartosz Putrycz, and T. Merritt, “Non-autoregressive TTS with explicit duration modelling for low-resource highly expressive speech,” in Proc. ISCA Speech Synthesis Workshop , 2021, pp. 96–101

    2021

  7. [15]

    Low-resource expressive text-to-speech using data augmen- tation,

    G. Huybrechts, T. Merritt, G. Comini, B. Perz, R. Shah, and J. Lorenzo- Trueba, “Low-resource expressive text-to-speech using data augmen- tation,” in Proc. IEEE Intl. Conf. on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2021, pp. 6593–6597

    2021

  8. [16]

    Distribution augmentation for low-resource expressive Text-To-Speech,

    M. Lajszczak, A. Prasad, A. van Korlaar, B. Bollepalli, A. Bona- fonte, A. Joly, M. Nicolis, A. Moinet, T. Drugman, T. Wood, and E. Sokolova, “Distribution augmentation for low-resource expressive Text-To-Speech,” in Proc. IEEE Intl. Conf. on Acoustics, Speech and Signal Proce...

    2022

  9. [17]

    Noise robust TTS for low resource speakers using pre-trained model and speech enhancement,

    D. Dai, L. Chen, Y . Wang, M. Wang, R. Xia, X. Song, Z. Wu, and Y . Wang, “Noise robust TTS for low resource speakers using pre-trained model and speech enhancement,” arXiv preprint arXiv:2005.12531 , 2020

    2005 arXiv

  10. [18]

    Low-resource text-to- speech using specific data and noise augmentation,

    K. Kayyar, C. Dittmar, N. Pia, and E. Habets, “Low-resource text-to- speech using specific data and noise augmentation,” in Proc. IEEE-SPS European Signal Processing Conf. , 2023, pp. 61–65

    2023

  11. [19]

    Tacotron: Towards end-to-end speech synthesis,

    Y . Wang, R. Skerry-Ryan, D. Stanton, Y . Wu, R. J. Weiss, N. Jaitly, Z. Yang, Y . Xiao, Z. Chen, S. Bengio, Q. Le, Y . Agiomyrgiannakis, R. Clark, and R. A. Saurous, “Tacotron: Towards end-to-end speech synthesis,” in Proc. Interspeech, 2017, pp. 4006–4010

    2017

  12. [20]

    StyleMelGAN: An efficient high- fidelity adversarial vocoder with temporal adaptive normalization,

    A. Mustafa, N. Pia, and G. Fuchs, “StyleMelGAN: An efficient high- fidelity adversarial vocoder with temporal adaptive normalization,” in Proc. IEEE Intl. Conf. on Acoustics, Speech and Signal Processing (ICASSP), Toronto, Canada, 2021, pp. 6034–6038

    2021

  13. [21]

    The AudioLabs system for the Blizzard Challenge 2023,

    F. Zalkow, P. Sani, M. Fast, J. Bauer, M. Joshaghani, K. Kayyar, E. A. P. Habets, and C. Dittmar, “The AudioLabs system for the Blizzard Challenge 2023,” in Proceedings of the Blizzard Challenge Workshop , Grenoble, France, 2023, pp. 63–68

    2023

  14. [22]

    A systematic study of the class imbalance problem in convolutional neural networks,

    M. Buda, A. Maki, and M. A. Mazurowski, “A systematic study of the class imbalance problem in convolutional neural networks,” Neural networks, vol. 106, pp. 249–259, 2018

    2018

  15. [23]

    WhisperX: Time-accurate speech transcription of long-form audio,

    M. Bain, J. Huh, T. Han, and A. Zisserman, “WhisperX: Time-accurate speech transcription of long-form audio,” in Proc. Interspeech, 2023, pp. 4489–4493

    2023

  16. [24]

    The LJ speech dataset,

    K. Ito and L. Johnson, “The LJ speech dataset,” https://keithito.com/ LJ-Speech-Dataset/, 2017

    2017

  17. [25]

    TC-STAR: Spec- ifications of language resources and evaluation for speech synthesis,

    A. Bonafonte, H. H ¨oge, I. Kiss, A. Moreno, U. Ziegenhain, H. van den Heuvel, H.-U. Hain, X. S. Wang, and M. N. Garcia, “TC-STAR: Spec- ifications of language resources and evaluation for speech synthesis,” in Proc. Fifth International Conference on Language Resources and Eva...

    2006

  18. [26]

    Hi-Fi multi- speaker english TTS dataset,

    E. Bakhturina, V . Lavrukhin, B. Ginsburg, and Y . Zhang, “Hi-Fi multi- speaker english TTS dataset,” inProc. Interspeech, 2021, pp. 2776–2780

    2021

  19. [27]

    P. 56, objective measurement of active speech level,

    International Telecommunication Union, “P. 56, objective measurement of active speech level,” ITU-T, Geneva, 2011

    2011

  20. [28]

    Subjective evaluation of text-to-speech models: Comparing absolute category rating and ranking by elimination tests,

    K. Kayyar, C. Dittmar, N. Pia, and E. Habets, “Subjective evaluation of text-to-speech models: Comparing absolute category rating and ranking by elimination tests,” in Proc. ISCA Speech Synthesis Workshop , 2023, pp. 191–196

    2023

  21. [29]

    ECAPA-TDNN: Emphasized channel attention, propagation and aggregation in TDNN based speaker verification,

    B. Desplanques, J. Thienpondt, and K. Demuynck, “ECAPA-TDNN: Emphasized channel attention, propagation and aggregation in TDNN based speaker verification,” in Proc. Interspeech . ISCA, 2020, pp. 3830–3834

    2020

  22. [30]

    SpeechBrain: A general- purpose speech toolkit,

    M. Ravanelli, T. Parcollet, P. Plantinga, A. Rouhe, S. Cornell, L. Lu- gosch, C. Subakan, N. Dawalatabad, A. Heba, J. Zhong, J.-C. Chou, S.-L. Yeh, S.-W. Fu, C.-F. Liao, E. Rastorgueva, F. Grondin, W. Aris, H. Na, Y . Gao, R. D. Mori, and Y . Bengio, “SpeechBrain: A general- p...

    2021 arXiv

  23. [31]

    Towards the next generation of web-based experiments: A case study assessing basic audio quality following the ITU-R recommendation BS.1534 (MUSHRA),

    M. Schoeffler, F.-R. St ¨oter, B. Edler, and J. Herre, “Towards the next generation of web-based experiments: A case study assessing basic audio quality following the ITU-R recommendation BS.1534 (MUSHRA),” in Proc. Web Audio Conference , Paris, France, January 2015

    2015

  24. [32]

    P. 808, Subjective evaluation of speech quality with a crowdsourcing approach,

    International Telecommunication Union, “P. 808, Subjective evaluation of speech quality with a crowdsourcing approach,” ITU-T, Geneva, 2018

    2018

  25. [33]

    Refining the evaluation of speech synthesis: A summary of the blizzard challenge 2023,

    O. Perrotin, B. Stephenson, S. Gerber, G. Bailly, and S. King, “Refining the evaluation of speech synthesis: A summary of the blizzard challenge 2023,” Computer Speech & Language , vol. 90, p. 101747, 2025

    2023

  26. [34]

    IEEE recommended practice for speech quality measurements,

    E. Rothauser, W. Chapman, N. Guttman, H. Silbiger, M. Hecker, G. Ur- banek, K. Nordby, and M. Weinstock, “IEEE recommended practice for speech quality measurements,” IEEE Transactions on Audio and Electroacoustics, vol. 17, 1969

    1969

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.