Pith. sign in

REVIEW 2 major objections 7 minor 1 cited by

AdaptVC: High Quality Voice Conversion with Adaptive Learning

T0 review · 2 major / 7 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 AdaptVC claims to beat prior voice conversion models in intelligibility and target-speaker similarity while running up to ten times faster.

desk verdict AdaptVC is a well-engineered VC system with a real speed advantage and an informative ablation, but its 'outperforms all models' claim outruns the evidence in Table I. read the letter →

arxiv 2501.01347 v4 pith:HRXUOKQX submitted 2025-01-02 cs.SD cs.CLeess.AS

Jaehun Kim , Ji-Hoon Kim , Yeunju Choi , Tan Dat Nguyen , Seongkyu Mun , Joon Son Chung This is my paper · ORCID
classification cs.SDcs.CLeess.AS
keywords zero-shotvoiceconversionself-supervisedspeechfeaturesHuBERTadaptertuningvectorquantizationconditionalflowmatchingcross-attentionspeakerconditioningdisentanglement
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 proposes AdaptVC, a zero-shot voice conversion system that turns one speaker's utterance into another speaker's voice while preserving the words. It claims to surpass previously published voice conversion models in intelligibility and similarity to the target speaker, and that its 5-step sampling version reaches the best balance of quality and speed. The authors' route is to learn, rather than hand-pick, which layers of a pretrained self-supervised speech model carry content versus voice identity, then to strip speaker information with a vector-quantization bottleneck and re-synthesize with a fast flow-matching decoder. If the claim holds, high-quality voice conversion can run in real time without per-speaker tuning.

What carries the argument

Three components carry the argument. First, adapters: a single softmax-weighted linear combination over all HuBERT layer outputs, one adapter for content and one for speaker, learns which self-supervised layers to use instead of requiring a heuristic selection of an intermediate layer. Second, a vector-quantization bottleneck: a single quantizer with a codebook of 512 codes is applied to the content adapter output and trained with a commitment loss and a prior loss, discretizing the representation so that speaker identity is stripped away while linguistic content survives. Third, an OT-CFM decoder: a transformer U-Net trained by optimal-transport conditional flow matching, with self-attention layers replaced by cross-attention layers that use the frame-wise speaker features as keys and values, so the speaker identity is injected at multiple depths of the decoder.

What would settle it

Train a speaker classifier on the content-encoder output before the VQ layer across LibriTTS speakers; if speaker identification accuracy is far above chance, the bottleneck is leaking speaker identity and the claimed disentanglement fails. Conversely, if converted speech on heavily accented or noisy unseen speakers shows word error rates far above the reported levels, the bottleneck is discarding content the decoder cannot recover.

Watch

Extended reading notes

Core claim

The central discovery is that a small learned adapter over all intermediate HuBERT layer outputs automatically separates linguistic content from speaker timbre, provided a single vector-quantization bottleneck of 512 codes sits after the content adapter. The content adapter concentrates its learned weights on the later HuBERT layers, while the speaker adapter favors the earlier layers, matching the known layer-wise organization of self-supervised speech features. The decoder, a transformer U-Net conditioned by cross-attention on frame-wise speaker features and trained with optimal-transport conditional flow matching, then turns the content codes plus speaker features into a mel-spectrogram. Evaluated on unseen speakers from the VCTK corpus, the system reports higher naturalness and similarity mean opinion scores than kNN-VC, DiffVC, and DDDM-VC, with word and character error rates near or below the baselines and a real-time factor of 0.02 at 5 sampling steps.

Load-bearing premise

The whole result turns on a single vector-quantization bottleneck of 512 codes: it must strip away enough speaker information from the HuBERT-derived content features to make the converted voice follow the reference, yet preserve enough linguistic detail for the decoder to reconstruct intelligible words, and the right balance is found only through training.

Editorial extensions

If this is right

  • Zero-shot voice conversion can be done without heuristic layer selection: the adapters learn the weighting of self-supervised layers directly from data.
  • The 5-step OT-CFM decoder runs at a real-time factor of about 0.02, making near-real-time conversion practical for interactive applications.
  • Without the VQ bottleneck, the model reconstructs the source speaker regardless of the reference, confirming that the discrete codes are what enforce content–speaker disentanglement.
  • Cross-attention speaker conditioning outperforms both style adaptive layer normalization and mean-pooling-plus-addition, so the way speaker features are injected into the decoder matters for similarity.
  • The system degrades gracefully when the number of sampling steps is reduced from 10 to 1, unlike the diffusion baselines, which show larger quality drops.

Reading between the lines

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

  • The learned adapter weights could serve as a general diagnostic for which layers of other self-supervised speech models encode content versus speaker attributes, not just for HuBERT.
  • Because the speaker encoder produces frame-wise features rather than a single utterance-level vector, the method may transfer dynamic prosodic detail; this could be tested on emotional or whispered speech where timbre varies within an utterance.
  • The single 512-code VQ bottleneck raises a natural scaling question the paper does not test: whether larger codebooks or multi-codebook quantization would improve target similarity without hurting intelligibility.
  • If AdaptVC is applied to languages or recording conditions far outside its LibriTTS training data, the adapter layer weights may need retraining, since layer-wise specialization of self-supervised features could shift across domains.
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

2 major / 7 minor

Summary. The paper proposes AdaptVC, a zero-shot voice conversion system that combines HuBERT feature adapters, a vector quantization (VQ) bottleneck for content disentanglement, and an optimal-transport conditional flow matching (OT-CFM) decoder with cross-attention speaker conditioning. The adapters learn weighted combinations of all intermediate HuBERT layer outputs separately for content and speaker streams. The model is trained on LibriTTS and evaluated on VCTK in a zero-shot setting, comparing against kNN-VC, DiffVC, and DDDM-VC using MOS, UTMOS, WER, CER, SECS, and RTF metrics. The authors claim that AdaptVC outperforms all existing voice conversion models in intelligibility and target speaker similarity, and that the 5-step version gives the best balance between quality and speed.

Significance. If the claims were fully supported, AdaptVC would be a valuable contribution: it demonstrates a practical combination of adapter-based SSL feature tuning, VQ-based disentanglement, and fast flow-matching decoding, with an ablation study that credibly attributes performance to each component. The reported RTF values (0.02 for 5 steps) are substantially better than the diffusion baselines, which is practically significant. However, the headline claim of universal superiority is not established by the evidence as presented: only three baselines are compared, objective metrics do not consistently favor AdaptVC, and the subjective MOS differences are not tested for significance. The core system idea is sound, but the evaluation and framing need revision before the claims can be accepted.

major comments (2)
  1. [Abstract and Section V] The central claim that AdaptVC 'outperforms all existing voice conversion models' is not supported by the objective results in Table I. Only three baselines are compared, and on WER (lower is better) DDDM-VC (10) achieves 6.40 ± 2.15, which is better than both AdaptVC (10) at 7.39 ± 1.06 and AdaptVC (5) at 6.96 ± 0.97. On SECS (higher is better), Diff-VC (30) achieves 0.828 and DDDM-VC (10) 0.823, both above AdaptVC (10) at 0.821 and AdaptVC (5) at 0.801. The paper itself states in Section IV-B that DiffVC has the highest SECS. The superiority claim therefore rests on the MOS metrics, not on the objective metrics, but the broad statement in the Abstract and Section V does not reflect this nuance.
  2. [Section III-D and Section IV-B] The subjective evaluation is too weak to support the 'significant margin' claimed in Section V. The MOS results come from only 20 domain experts and 40 generated samples, and the reported 95% confidence intervals overlap considerably; for example, MOS-S for AdaptVC (5) is 3.36 ± 0.21 versus 3.19 ± 0.23 for DDDM-VC (10), and MOS-N for AdaptVC (5) is 3.86 ± 0.14 versus 3.48 ± 0.14 for DDDM-VC (10). No significance testing (e.g., paired bootstrap, Wilcoxon signed-rank, or ANOVA) is reported for the MOS differences, so the claim that AdaptVC outperforms the baselines by a significant margin is not demonstrated. Additionally, with only three baselines, the phrase 'all existing voice conversion models' is an overgeneralization that the evaluated evidence cannot support.
minor comments (7)
  1. [Section II-A] The sentence 'the output of the adapter is guided to map similar content information from various speakers into closest embedding' is unclear; it should be reworded to specify that the VQ layer assigns adapter outputs to the nearest codebook vector, and 'closest' should be 'the closest'.
  2. [Equation (2)] The prior loss in Eq. (2) uses φ(x_i; μ_i, I) without defining I as the identity covariance matrix; please state that φ denotes a Gaussian density with identity covariance.
  3. [Section III-B] The description of the decoder architecture ('The architecture of the decoder followed that of [22]') is vague; please provide the number of transformer blocks, hidden dimensions, attention heads, and other relevant hyperparameters, or include a link to the official implementation.
  4. [Section III-D] The phrase 'MOS was conducted to 20 domain experts' should be 'MOS was conducted with 20 domain experts'; also clarify how the 40 generated samples were distributed across the systems and speakers.
  5. [Section IV-B] There is a typo: 'AdapterVC with only 5 sampling steps' should read 'AdaptVC with only 5 sampling steps'.
  6. [Figure 3] The y-axis label 'trained weights' is ambiguous; specify that these are the softmax-normalized adapter weights, and label the x-axis as HuBERT layer indices.
  7. [References] Reference [29] is cited as 'Robust speech recognition via large-scale weak supervision' without naming the model (Whisper); adding the model name would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: AdaptVC is an empirical system paper with learned adapters evaluated on held-out speakers; no prediction reduces by construction to its inputs.

full rationale

The derivation chain in AdaptVC is not a formal derivation: the adapters, VQ bottleneck, and CFM decoder are trained with commitment, prior, and flow-matching losses and then evaluated zero-shot on VCTK speakers unseen during LibriTTS training. The reported predictions (WER, MOS, SECS) are measurements of a held-out system, not quantities defined by fitted parameters. The adapter weights are learned from data and analyzed post hoc; no fitted constant is renamed as a prediction. The only overlapping-author citations (ref. [2] VoiceMixer by Ji-Hoon Kim and ref. [21] by Jaehun Kim) support design choices such as applications and frame-wise speaker conditioning, but they do not carry the central performance claim. The claim of superiority over baselines is an empirical overclaim given overlapping confidence intervals and no significance testing, but that is a correctness or evidence concern, not circularity.

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

The model relies on learned adapter weights and a hand-chosen VQ codebook as its main free parameters. No new physical entities or theoretical constructs are introduced. The axioms are standard domain assumptions about SSL features and evaluation metrics, plus the transferability of the OT-CFM framework from TTS. The central design choices are empirical rather than derived.

free parameters (5)
  • Adapter softmax weights (content and speaker) = Learned, visualized in Fig. 3
    The central learned mechanism: a weighted combination of all HuBERT intermediate layers. The weights are optimized on LibriTTS, not set by a rule.
  • VQ codebook size = 512
    Chosen by hand; affects the content-speaker disentanglement tradeoff.
  • Number of decoder sampling steps = 5 (best balance), with 1 and 10 also evaluated
    Chosen post hoc as the best tradeoff between quality and speed in Table I.
  • Loss weighting coefficients = 1.0 for all three losses (implicit)
    The total loss is a sum with no explicit weights (Eq. 4); no tuning is reported.
  • Mel-spectrogram configuration = window 1280, hop 320, 80 mel bins
    Standard configuration chosen to match HuBERT temporal resolution.
assumptions (4)
  • domain assumption HuBERT's intermediate layers encode both linguistic and speaker information in a way that a linear weighted sum can separate.
    Invoked in Section II-A/B and supported by cited works [31,32]; the whole adapter design depends on it.
  • domain assumption Training on LibriTTS generalizes to zero-shot speakers from VCTK.
    Evaluation uses 20+20 VCTK speakers not seen in training; no special adaptation is applied.
  • domain assumption WER/CER via a pretrained ASR model and SECS via Resemblyzer are valid proxies for intelligibility and speaker similarity.
    Used in Section III-D as objective metrics; MOS is also collected.
  • standard math The OT-CFM framework from Matcha-TTS [22] is applicable to VC mel-spectrograms without modification.
    The decoder architecture and training objective (Eq. 3) are taken from [22].

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdaptVC: High Quality Voice Conversion with Adaptive Learning." pith.science (2026). https://pith.science/paper/HRXUOKQX

@misc{pith2026250101347,
  author       = {Pith},
  title        = {Pith review of: AdaptVC: High Quality Voice Conversion with Adaptive Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HRXUOKQX}},
  note         = {Machine review of arXiv:2501.01347}
}
read the original abstract

The goal of voice conversion is to transform the speech of a source speaker to sound like that of a reference speaker while preserving the original content. A key challenge is to extract disentangled linguistic content from the source and voice style from the reference. While existing approaches leverage various methods to isolate the two, a generalization still requires further attention, especially for robustness in zero-shot scenarios. In this paper, we achieve successful disentanglement of content and speaker features by tuning self-supervised speech features with adapters. The adapters are trained to dynamically encode nuanced features from rich self-supervised features, and the decoder fuses them to produce speech that accurately resembles the reference with minimal loss of content. Moreover, we leverage a conditional flow matching decoder with cross-attention speaker conditioning to further boost the synthesis quality and efficiency. Subjective and objective evaluations in a zero-shot scenario demonstrate that the proposed method outperforms existing models in speech quality and similarity to the reference speech.

Figures

Figures reproduced from arXiv: 2501.01347 by the authors.

Figure 1
Figure 1. Overall architecture of AdaptVC. hcont denotes the content representation from the adapter in the content encoder, and hspk denotes the speaker features from that in the speaker encoder. Prior distribution µ is obtained by fusing the content and speaker information through cross-attention. encoded content features are passed to a U-Net based CFM decoder, conditioned with encoded speaker features, which generates the… view at source ↗
Figure 2
Figure 2. Illustration of HuBERT adapter mechanism (a) and decoder block [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Visualization of adapter weights. Numbers in the x-axis indicate layer [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. EZ-VC: Easy Zero-shot Any-to-Any Voice Conversion

    cs.SD 2025-05 reject novelty 4.0 of 10

    EZ-VC combines discrete units from a multilingual self-supervised encoder (Xeus) with an F5-TTS flow-matching decoder to achieve zero-shot any-to-any voice conversion, without text labels or multiple disentangling encoders.

Reference graph

Works this paper leans on

33 extracted references · 32 canonical work pages · cited by 1 Pith paper

  1. [1]

    YourTTS: Towards Zero-Shot Multi-Speaker TTS and Zero-Shot V oice 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 V oice Conversion for Everyone,” in Proc. ICML, 2022. 1

    work page 2022

  2. [2]

    V oiceMixer: Adver- sarial V oice Style Mixup,

    S.-H. Lee, J.-H. Kim, H. Chung, and S.-W. Lee, “V oiceMixer: Adver- sarial V oice Style Mixup,” in NeurIPS, 2021. 1

    work page 2021

  3. [3]

    Disentangled Speech Representation Learning for One-Shot Cross-Lingual V oice Con- version Using ß-V AE,

    H. Lu, D. Wang, X. Wu, Z. Wu, X. Liu, and H. Meng, “Disentangled Speech Representation Learning for One-Shot Cross-Lingual V oice Con- version Using ß-V AE,” inIEEE Spoken Language Technology workshop,

    work page

  4. [4]

    AutoVC: Zero-Shot V oice Style Transfer with Only Autoencoder Loss,

    K. Qian, Y . Zhang, S. Chang, X. Yang, and M. Hasegawa-Johnson, “AutoVC: Zero-Shot V oice Style Transfer with Only Autoencoder Loss,” in Proc. ICML, 2019. 1

    work page 2019

  5. [5]

    F0- Consistent Many-To-Many Non-Parallel V oice Conversion Via Condi- tional Autoencoder,

    K. Qian, Z. Jin, M. Hasegawa-Johnson, and G. J. Mysore, “F0- Consistent Many-To-Many Non-Parallel V oice Conversion Via Condi- tional Autoencoder,” in Proc. ICASSP, 2020. 1

    work page 2020

  6. [6]

    Diffusion-Based V oice Conversion with Fast Maximum Likelihood Sampling Scheme,

    V . Popov, I. V ovk, V . Gogoryan, T. Sadekova, M. Kudinov, and J. Wei, “Diffusion-Based V oice Conversion with Fast Maximum Likelihood Sampling Scheme,” in Proc. ICLR, 2022. 1, 3

    work page 2022

  7. [7]

    HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units,

    W.-N. Hsu, B. Bolte, Y .-H. H. Tsai, K. Lakhotia, R. Salakhutdinov, and A. Mohamed, “HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units,” IEEE/ACM Trans. on Audio, Speech, and Language Processing , 2021. 1

    work page 2021

  8. [8]

    wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations,

    A. Baevski, Y . Zhou, A. Mohamed, and M. Auli, “wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations,” in NeurIPS, 2020. 1

    work page 2020

Show all 33 references
  1. [9]

    WavLM: Large-Scale Self-Supervised Pre- Training for Full Stack Speech Processing,

    S. Chen, C. Wang, Z. Chen, Y . Wu, S. Liu, Z. Chen, J. Li, N. Kanda, T. Yoshioka, X. Xiao, J. Wu, L. Zhou, S. Ren, Y . Qian, Y . Qian, J. Wu, M. Zeng, X. Yu, and F. Wei, “WavLM: Large-Scale Self-Supervised Pre- Training for Full Stack Speech Processing,” IEEE Journal of Select...

    2022

  2. [10]

    XLS-R: Self-Supervised Cross-Lingual Speech Representation Learning at Scale,

    A. Babu, C. Wang, A. Tjandra, K. Lakhotia, Q. Xu, N. Goyal, K. Singh, P. von Platen, Y . Saraf, J. M. Pino, A. Baevski, A. Conneau, and M. Auli, “XLS-R: Self-Supervised Cross-Lingual Speech Representation Learning at Scale,” in Proc. Interspeech, 2021. 1

    2021

  3. [11]

    Neural Analysis and Synthesis: Reconstructing Speech from Self-Supervised Representations,

    H.-S. Choi, J. Lee, W. Kim, J. Lee, H. Heo, and K. Lee, “Neural Analysis and Synthesis: Reconstructing Speech from Self-Supervised Representations,” in NeurIPS, 2021. 1

    2021

  4. [12]

    DDDM-VC: Decoupled De- noising Diffusion Models with Disentangled Representation and Prior Mixup for Verified Robust V oice Conversion,

    H.-Y . Choi, S.-H. Lee, and S.-W. Lee, “DDDM-VC: Decoupled De- noising Diffusion Models with Disentangled Representation and Prior Mixup for Verified Robust V oice Conversion,” in Proc. AAAI, 2024. 1, 3

    2024

  5. [13]

    V oice Conversion With Just Nearest Neighbors,

    M. Baas, B. van Niekerk, and H. Kamper, “V oice Conversion With Just Nearest Neighbors,” in Proc. Interspeech, 2023. 1, 3

    2023

  6. [14]

    Parameter-Efficient Transfer Learning for NLP,

    N. Houlsby, A. Giurgiu, S. Jastrzebski, B. Morrone, Q. De Laroussilhe, A. Gesmundo, M. Attariyan, and S. Gelly, “Parameter-Efficient Transfer Learning for NLP,” in Proc. ICML, 2019. 1

    2019

  7. [15]

    Lora: Low-rank adaptation of large language models,

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “Lora: Low-rank adaptation of large language models,” in Proc. ICLR, 2022. 1

    2022

  8. [16]

    AdapterHub: A framework for adapting transformers,

    J. Pfeiffer, A. R ¨uckl´e, C. Poth, A. Kamath, I. Vuli ´c, S. Ruder, K. Cho, and I. Gurevych, “AdapterHub: A framework for adapting transformers,” in Proc. EMNLP, 2020. 1

    2020

  9. [17]

    Flow Matching for Generative Modeling,

    Y . Lipman, R. T. Q. Chen, H. Ben-Hamu, M. Nickel, and M. Le, “Flow Matching for Generative Modeling,” in Proc. ICLR, 2023. 1, 2

    2023

  10. [18]

    Vector-Quantized Image Modeling with Improved VQGAN,

    J. Yu, X. Li, J. Y . Koh, H. Zhang, R. Pang, J. Qin, A. Ku, Y . Xu, J. Baldridge, and Y . Wu, “Vector-Quantized Image Modeling with Improved VQGAN,” in Proc. ICLR, 2022. 2

    2022

  11. [19]

    Neural discrete representation learning,

    A. Van Den Oord, O. Vinyals et al. , “Neural discrete representation learning,” in NeurIPS, 2017. 2

    2017

  12. [20]

    NANSY++: Unified V oice Synthesis with Neural Analysis and Synthesis,

    H.-S. Choi, J. Yang, J. Lee, and H. Kim, “NANSY++: Unified V oice Synthesis with Neural Analysis and Synthesis,” in Proc. ICLR, 2023. 2

    2023

  13. [21]

    Encoding Speaker-Specific Latent Speech Feature for Speech Synthesis,

    J. Kong, J. Lee, J. Kim, B. Kim, J. Park, D. Kong, C. Lee, and S. Kim, “Encoding Speaker-Specific Latent Speech Feature for Speech Synthesis,” in Proc. ICML, 2024. 2

    2024

  14. [22]

    Matcha- TTS: A fast TTS architecture with conditional flow matching,

    S. Mehta, R. Tu, J. Beskow, ´E. Sz ´ekely, and G. E. Henter, “Matcha- TTS: A fast TTS architecture with conditional flow matching,” in Proc. ICASSP, 2024. 2, 3

    2024

  15. [23]

    High- resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” in Proc. CVPR,

  16. [24]

    Grad- tts: A diffusion probabilistic model for text-to-speech,

    V . Popov, I. V ovk, V . Gogoryan, T. Sadekova, and M. Kudinov, “Grad- tts: A diffusion probabilistic model for text-to-speech,” in Proc. ICML,

  17. [25]

    LibriTTS: A Corpus Derived from LibriSpeech for Text- to-Speech,

    H. Zen, V . Dang, R. Clark, Y . Zhang, R. J. Weiss, Y . Jia, Z. Chen, and Y . Wu, “LibriTTS: A Corpus Derived from LibriSpeech for Text- to-Speech,” in Proc. Interspeech, 2019. 3

    2019

  18. [26]

    Cstr vctk corpus: English multi-speaker corpus for cstr voice cloning toolkit,

    C. Veaux, J. Yamagishi, K. MacDonald et al. , “Cstr vctk corpus: English multi-speaker corpus for cstr voice cloning toolkit,” University of Edinburgh. The Centre for Speech Technology Research (CSTR), 2017. 3

    2017

  19. [27]

    HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis,

    J. Kong, J. Kim, and J. Bae, “HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis,” in NeurIPS,

  20. [28]

    Utmos: Utokyo-sarulab system for voicemos challenge 2022,

    T. Saeki, D. Xin, W. Nakata, T. Koriyama, S. Takamichi, and H. Saruwatari, “Utmos: Utokyo-sarulab system for voicemos challenge 2022,” in Proc. Interspeech, 2022. 3

    2022

  21. [29]

    Robust speech recognition via large-scale weak super- vision,

    A. Radford, J. W. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak super- vision,” in Proc. ICML, 2023. 3

    2023

  22. [30]

    Generalized end-to-end loss for speaker verification,

    L. Wan, Q. Wang, A. Papir, and I. L. Moreno, “Generalized end-to-end loss for speaker verification,” in Proc. ICASSP, 2018. 4

    2018

  23. [31]

    Noise-Robust Zero-Shot Text-to-Speech Synthesis Condi- tioned on Self-Supervised Speech-Representation Model with Adapters,

    K. Fujita, H. Sato, T. Ashihara, H. Kanagawa, M. Delcroix, T. Moriya, and Y . Ijima, “Noise-Robust Zero-Shot Text-to-Speech Synthesis Condi- tioned on Self-Supervised Speech-Representation Model with Adapters,” in Proc. ICASSP, 2024. 4

    2024

  24. [32]

    Investigating the’Autoencoder Behavior’in Speech Self- Supervised Models: a focus on HuBERT’s Pretraining,

    V . Vielzeuf, “Investigating the’Autoencoder Behavior’in Speech Self- Supervised Models: a focus on HuBERT’s Pretraining,” arXiv preprint arXiv:2405.08402, 2024. 4

    2024 arXiv

  25. [33]

    Meta-stylespeech : Multi- speaker adaptive text-to-speech generation,

    D. Min, D. B. Lee, E. Yang, and S. J. Hwang, “Meta-stylespeech : Multi- speaker adaptive text-to-speech generation,” in Proc. ICML, 2021. 4

    2021

Pith tools

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