Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

SpeechAccentLLM: A Unified Framework for Foreign Accent Conversion and Text to Speech

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A CTC-regularized speech tokenizer plus joint FAC/TTS training lets an LLM-style model convert foreign-accented speech toward native pronunciation while preserving content and speaker identity, cutting accentedness from 2.48 to 1.86 and…

desk verdict A plausible system paper whose headline FAC result is undermined by training on a single-speaker synthetic TTS target; worth a referee but needs to confront natural-speech validation. read the letter →

arxiv 2507.01348 v2 pith:5J5XGWZZ submitted 2025-07-02 eess.AS cs.SD

Zhuangfei Cheng , Guangyan Zhang , Zehai Tu , Yangyang Song , Shuiyang Mao , Xiaoqi Jiao , Jingyu Li , Yiwen Guo
show 1 more author
Jiasong Wu
This is my paper
classification eess.AScs.SD
keywords foreignaccentconversiontext-to-speechdiscretespeechtokensCTCregularizationvectorquantizationmultitasklearningrestorationLLM-basedsynthesis
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 SpeechAccentLLM, a single LLM-style framework that performs both foreign accent conversion (FAC) and text-to-speech (TTS) from discrete speech tokens. The central claim is that regularizing the speech tokenizer with connectionist temporal classification (CTC) yields tokens with a locality property—each token encodes only its local speech segment—which makes LLM-based conversion stable and robust. Against the zero-shot FAC baseline, the framework reports lower accentedness (1.86 vs 2.48), lower word error rate (9.1% vs 14.4%), and higher naturalness (CMOS 4.074 vs 3.552) while preserving speaker identity. The authors attribute this to joint multitask training with abundant TTS data compensating for the scarcity of accented parallel data, plus a BERT-style SpeechRestorer module that repairs token-level errors from the LLM output.

What carries the argument

The load-bearing mechanism is the CTC-regularized SpeechCodeVAE tokenizer: a pre-trained Whisper encoder feeds a pre-VQ encoder whose outputs are supervised by CTC on IPA labels while being vector-quantized into a 1024-entry codebook. This produces discrete content tokens with locality—each token at a frame depends only on its corresponding speech segment, with minimal cross-frame dependency—which makes the token stream stable under speed changes and deduplication, and makes the subsequent LLM's accent-transfer task tractable. Around this tokenizer sit the multitask FAC/TTS transformer decoder and the BERT-style SpeechRestorer, each exploiting the locality to correct or convert tokens.

What would settle it

Collect accented speech from L2 speakers together with the same speakers reading the same texts with a native accent (or matched native speakers), train and evaluate SpeechAccentLLM against those natural targets, and compare accentedness and CMOS; if the advantage over the baseline shrinks or disappears, the synthetic-target premise is the load-bearing part of the result.

Watch

Extended reading notes

Core claim

The central claim is that SpeechAccentLLM outperforms the zero-shot foreign accent conversion baseline by making three design choices work together. SpeechCodeVAE discretizes speech into CTC-regularized content tokens plus a speaker embedding; a transformer decoder is jointly trained on FAC and TTS tasks; and SpeechRestorer, a bidirectional transformer, replaces or masks corrupted tokens before waveform reconstruction. The paper's discovery is that CTC guidance during vector-quantized codebook discretization produces content tokens that are locally constrained and temporally robust, and that this locality property is what allows an autoregressive LLM to convert nonnative content tokens into native-like content tokens without explicit phonemes. The reported 25% accentedness improvement, WER reduction from 14.4% to 9.1%, and CMOS gain from 3.552 to 4.074 are all credited to this tokenizer plus multitask training and post-restoration.

Load-bearing premise

The native-accented targets the converter learns from are synthesized by a single-speaker VITS model, so the system only ever learns to imitate a synthetic version of native pronunciation rather than natural native speech.

Editorial extensions

If this is right

  • Foreign accent conversion can be trained with far less accented parallel data because TTS data carries part of the learning load.
  • The CTC-regularized tokens give better robustness to speaking-rate variation and temporal compression, which matters for streaming and real-time speech generation.
  • A separate restoration stage can catch and fix token-level errors from LLM-based speech generation, improving fluency without retraining the whole model.
  • Jointly training FAC and TTS yields faster convergence and higher-quality speech than training FAC alone.
  • SpeechCodeVAE's disentangled representations generalize to voice conversion in unseen domains, suggesting the tokenizer is reusable beyond accent conversion.

Reading between the lines

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

  • The locality property likely makes SpeechCodeVAE useful for streaming or causal speech generation, since each token depends only on its local speech segment and could be produced incrementally.
  • Because CTC supervision uses IPA labels, the tokenizer may transfer to languages outside the three training languages more easily than phoneme-based front-ends, which typically require language-specific grapheme-to-phoneme rules.
  • The same FAC+TTS multitask recipe could be applied to arbitrary accent-to-accent conversion, training a single model to convert between any pair of accents using only text and TTS data, without parallel recordings.
  • If the synthetic-target limitation proves real, a next test is whether using multi-speaker native TTS or real native recordings as training targets closes the remaining gap to natural accentedness.
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

5 major / 6 minor

Summary. The paper proposes SpeechAccentLLM, a unified LLM-based framework for foreign accent conversion (FAC) and text-to-speech (TTS). It introduces SpeechCodeVAE, a CTC-regularized content tokenizer designed to produce speaker-agnostic speech content tokens with a claimed locality property; a multitask FAC/TTS transformer decoder trained on these tokens; and SpeechRestorer, a BERT-style postprocessor that refines LLM-generated token sequences. The central empirical claim, based on the L2-ARCTIC test set, is that SpeechAccentLLM outperforms the Quamer et al. baseline on accentedness (1.86 vs 2.48), WER (9.1% vs 14.4%), speaker similarity Sim-O (0.627 vs 0.558), and CMOS (4.074 vs 3.552). Additional experiments evaluate the tokenizer's locality and voice-conversion ability, and the TTS side is evaluated against YourTTS and NaturalSpeech2 with ablations for SpeechRestorer and the Variance Adapter.

Significance. If the reported results are correct, the paper makes a useful contribution to accent conversion by demonstrating that (i) CTC-regularized VQ tokenization can yield content tokens that are robust to speaker variation, (ii) multitask training with TTS data mitigates the small-corpus problem for FAC, and (iii) a token-level restorer can improve LLM-based speech synthesis outputs. The paper ships a thorough set of listening experiments with native raters, objective metrics, and ablations, and it compares against several baselines rather than only reporting self-comparisons. However, the main FAC conclusion rests on an untested assumption about the synthetic training target, and several supporting claims (multitask benefit, SpeechRestorer in FAC, objective metric validity) are under-evidenced. These issues are fixable within the manuscript's scope, but they currently prevent the core claim from being fully established.

major comments (5)
  1. [Section 4.1] The FAC training target is synthetic, not natural native accented speech: native counterparts for L2-ARCTIC were generated by a single-speaker VITS model trained on LJSpeech. The paper justifies this by asserting that SpeechCodeVAE 'exclusively extracts its content information without being influenced by speaker information,' but this assertion is not tested for the FAC setting, and VITS-generated targets may not represent natural native segmental, prosodic, or durational patterns. Because every headline FAC number (accentedness 1.86, WER 9.1%, CMOS 4.074) is a property of a system trained on this synthetic target, the claim that the system produces 'native-like' speech needs a direct comparison between (a) the training targets and converted outputs and (b) natural native ARCTIC recordings of the same utterances, which are available and were excluded during training. At minimum, please add an experiment reporting the same metrics on natural native reference audio, or provide a clear argument and supporting evidence that the VITS target is acoustically representative of natural native speech.
  2. [Section 5.1 and Section 5.3] The multitask claim is not supported by a standalone FAC ablation. The text states that joint FAC/TTS training 'yielded accelerated convergence and superior speech quality compared to standalone FAC training,' but Table 1 only compares the full system against the Quamer et al. baseline; no FAC-only training run is reported. Similarly, SpeechRestorer is ablated only in the TTS experiments (the 'w/o SR' row in Table 4), not in the FAC pipeline, even though the FAC inference chain in Figure 1 includes SpeechRestorer as a required component. Without these ablations, the contributions of the two central innovations to the FAC result cannot be isolated. Please add an FAC-only ablation for the joint-training strategy and an FAC ablation that removes SpeechRestorer.
  3. [Section 4.3 and Section 5.1] The objective metrics suffer from potential circularity. WER is computed with Whisper ASR, the same model family used as the frozen backbone of the Content Encoder, and speaker similarity Sim-O/Sim-R is computed with an ECAPA-TDNN-based speaker verification model, the same architecture as the frozen Speaker Encoder. The model may therefore be implicitly optimized to produce outputs that are easy for these specific encoders to recognize, which could inflate the reported WER and similarity gains relative to what a different ASR or speaker verification system would show. Please re-evaluate at least the main FAC table with an ASR system not in the Whisper family (e.g., a wav2vec 2.0 or Kaldi-based system) and a different speaker embedding extractor, or add a human transcription/intelligibility measure to complement WER.
  4. [Tables 1 and 2] Objective metrics are reported without error bars or significance tests. Table 1 gives only point estimates for Sim-O, WER, and accentedness, and Table 2 gives no dispersion at all; this makes it impossible to determine whether the reported differences (e.g., WER 9.1% vs 14.4%, De-duplication Efficiency 0.253 vs 0.159) are statistically meaningful over the 100-utterance evaluation set. Since these numbers carry the main claims, please report confidence intervals or standard deviations for all objective metrics and, where possible, significance tests for the subjective accentedness ratings.
  5. [Section 5.2] The 'locality' claim is asserted but only indirectly tested. Locality is defined as tokens per frame depending only on their corresponding speech segment, yet the evidence in Table 2 and Figure 4 uses De-duplication Efficiency, Speed Robustness, and odd-even token replacement, none of which directly measures cross-frame dependence per position. The replacement experiment is suggestive, but the spectrogram comparison is visual and not quantified. Please provide a more direct quantitative measure of locality, for example, measuring how perturbing a token at position i affects reconstruction of frames far from i, or reporting an objective distance between the original and replaced spectrograms.
minor comments (6)
  1. [Abstract] 'SpeechCodeV AE' contains an erroneous space; it should be 'SpeechCodeVAE'.
  2. [Section 1] The sentence 'We note the absence of phoneme information in accented speech' is unclear; presumably it means the absence of phoneme annotations or reliable phonemization, not absence of phonemes in the acoustic signal. Please rephrase.
  3. [Figure 1] The figure label 'Native accented speech' is confusing for the output of an FAC system that produces native-like rather than accented speech; consider 'Native-like speech'.
  4. [Section 4.3] It is unclear whether the '20 native English speakers' who judged accentedness are the same as the 'expert raters' and whether accentedness was rated in a paired or absolute setting; please clarify the listening test design.
  5. [Section 4.2] The data ratio is described as 'AC and TTS tasks 1:1'; use 'FAC and TTS tasks' for consistency with the rest of the paper.
  6. [References] The Quamer et al. reference lists 'Proc. Intespeech'; should be 'Proc. Interspeech'.

Circularity Check

1 steps flagged · score 2.0 of 10

No load-bearing circularity: the FAC target is explicitly synthetic VITS output, but the headline accentedness and CMOS results are human-judged; only the objective Sim-O (and to a lesser extent WER) metrics share pretrained backbones with the model's own encoders.

  1. other [Sections 4.2 and 4.3, Table 1 (Sim-O metric)]
    "Speaker Encoder adopts ECAPA-TDNN (Heo et al., 2020) (512-dim, frozen); ... Sim-O measures similarity between synthesized speech and the reference speaker ... This metric is computed by extracting speaker embeddings through a pre-trained speaker verification model1, followed by cosine similarity calculation between embedding pairs."

    The speaker timbre embedding S that conditions the VITS-based decoder is produced by ECAPA-TDNN, and Sim-O is computed as cosine similarity between ECAPA embeddings of the synthesized utterance and the reference speaker. Because the decoder is trained to reconstruct speech from S, a high Sim-O partly verifies that the decoder can re-encode its own conditioning vector, not that an independent perceptual identity test was passed. This is an evaluation loop on the objective Sim-O column only; the accentedness and CMOS scores in the same table are human judgements, so the central FAC claim does not reduce to this metric.

full rationale

The paper's central FAC derivation is not circular by construction. The model is trained to map L2-ARCTIC content tokens to target tokens produced by a VITS model on LJSpeech, and the reported accentedness (1.86 vs 2.48) and CMOS (4.074 vs 3.552) come from native-listener tests, which are external to both the tokenizer and the target generator. The synthetic VITS target is a data-validity assumption, not a hidden identity: the paper discloses it explicitly, and no equation in the paper forces the evaluated output to equal that target. The only in-house citation (Zuo et al., 2025) is background on TTS/VC and is not load-bearing. The real but limited concern is metric alignment: WER uses Whisper ASR while content tokens are extracted by a frozen Whisper encoder, and Sim-O uses a speaker-verification embedding space aligned with the ECAPA-TDNN speaker encoder. These objective numbers are less independent than they appear, but they are not the training loss and the main perceptual results remain external. Overall circularity is minor and does not reach the level of a prediction that reduces to its inputs.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

All numbers in the central pipeline are architectural choices or pretrained-model inheritances; no new physical or theoretical entities are postulated. The key uncharged assumptions are the disentanglement hypothesis, the adequacy of Whisper features for accented speech, and the synthetic VITS target as a proxy for natural native speech.

free parameters (6)
  • VQ codebook size = 1024 entries
    Chosen without systematic search; sets the token vocabulary capacity.
  • Token frame rate = 50 Hz (Whisper-medium)
    Determines the sequence length for the LLM; inherited from the frozen Whisper encoder.
  • FAC-to-TTS data ratio = 1:1 via oversampling
    Balances the two tasks; affects the multitask trade-off and convergence.
  • SpeechRestorer corruption rates = 10% replaced, 10% masked
    Chosen corruption schedule for training the token restorer.
  • Input perturbation ranges = f0 ±20%, formants ±15% on 50% samples
    Data augmentation intensities in the content encoder; control timbre leakage and robustness.
  • Architecture sizes = 8-layer decoder (512-dim, 8 heads), 4-layer BERT, batch 32, dropout 0.1
    Model capacities and training hyperparameters chosen without systematic tuning.
assumptions (5)
  • domain assumption Speech can be factorized into independent content, speaker, and prosody components with the proposed architecture.
    The whole pipeline relies on this disentanglement; it is evaluated indirectly through VC and FAC but not proven.
  • domain assumption Frozen Whisper encoder features are speaker-agnostic and preserve content for accented speech.
    The content encoder uses Whisper-medium frozen; the authors cite CosyVoice but provide no direct demonstration on accented data.
  • domain assumption VITS-generated single-speaker native speech is an adequate proxy for natural native target speech in FAC training.
    Section 4.1 constructs native counterparts with VITS trained on LJSpeech; no natural parallel target data are used.
  • domain assumption CTC with IPA labels provides useful alignment for accented speech without a phoneme frontend.
    CTC loss is the core of token locality; accented pronunciations can degrade IPA alignments, and this is not explicitly tested.
  • standard math Standard optimization assumptions: backpropagation, EMA codebook updates, and pretrained encoders behave as expected.
    No alternative optimization or convergence proof is provided; standard ML assumptions apply.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SpeechAccentLLM: A Unified Framework for Foreign Accent Conversion and Text to Speech." pith.science (2026). https://pith.science/paper/5J5XGWZZ

@misc{pith2026250701348,
  author       = {Pith},
  title        = {Pith review of: SpeechAccentLLM: A Unified Framework for Foreign Accent Conversion and Text to Speech},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5J5XGWZZ}},
  note         = {Machine review of arXiv:2507.01348}
}
read the original abstract

Foreign accent conversion (FAC) in speech processing remains a challenging task. Building on the remarkable success of large language models (LLMs) in Text-to-Speech (TTS) tasks, this study investigates the adaptation of LLM-based techniques for FAC, which we term SpeechAccentLLM. At the core of this framework, we introduce SpeechCodeVAE, the first model to integrate connectionist temporal classification (CTC) directly into codebook discretization for speech content tokenization. This novel architecture generates tokens with a unique "locality" property, as validated by experiments demonstrating optimal trade-offs among content faithfulness, temporal coherence, and structural recoverability. Then, to address data scarcity for the FAC module, we adopted a multitask learning strategy that jointly trains the FAC and TTS modules. Beyond mitigating data limitations, this approach yielded accelerated convergence and superior speech quality compared to standalone FAC training. Moreover, leveraging the salient properties of our discrete speech representations, we introduce SpeechRestorer, a postprocessing architecture designed to refine LLM-generated outputs. This module effectively mitigates stochastic errors prevalent in LLM inference pipelines while enhancing prosodic continuity, as validated by ablation experiments.

Figures

Figures reproduced from arXiv: 2507.01348 by the authors.

Figure 1
Figure 1. The overview of SpeechAccentLLM for FAC inference. Speech Decoupling extracts nonnative ac￾cented content tokens and speaker information from nonnative accented speech. The FAC&TTS Model takes nonnative accented content tokens as input and outputs predicted native accented content tokens. SpeechRe￾storer post-processes the speech content tokens pre￾dicted by the FAC&TTS Model. Finally, Speech Re￾construction combine… view at source ↗
Figure 2
Figure 2. The overview of SpeechCodeVAE. The Speech Decoupling stage extracts speech content tokens V and speaker timbre embeddings S, and the Speech Reconstruction stage reconstructs the decoupled infor￾mation. the Waveform Decoder synthesizing final speech. The training process incorporates adversarial guid￾ance through a discriminator, ensuring high-fidelity waveform generation while maintaining efficient content-speaker-p… view at source ↗
Figure 3
Figure 3. Training processes of the FAC&TTS Model and SpeechRestorer: the two modules are trained sepa￾rately, all speech is converted into speech content tokens via the Content Encoder for model training. The Speaker Encoder employs a pre-trained speaker recognition model to extract speaker tim￾bre embeddings S. The remainder of the architec￾tural components maintain identical configuration to VITS (Kim et al., 2021). Follow… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Comparative spectrogram analysis of odd-even token repetition replacement. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Transcript-Free Flow-Matching Text-to-Speech via Speech Feature Conditioning

    eess.AS 2026-06 unverdicted novelty 6.0 of 10

    RTFree-F5 replaces reference transcripts with mapped self-supervised speech representations in F5-TTS, cutting WER on dysarthric speech from 24.6% to 10.4% without any transcript at inference.

Reference graph

Works this paper leans on

31 extracted references · 16 canonical work pages · cited by 1 Pith paper

  1. [1]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Alexei Baevski, Wei-Ning Hsu, Qiantong Xu, Arun Babu, Jiatao Gu, and Michael Auli. 2022. Data2vec: A general framework for self-supervised learning in speech, vision and language. In Proc. ICML, pages 1298--1312. PMLR

    work page 2022

  4. [4]

    Zal \'a n Borsos, Rapha \"e l Marinier, Damien Vincent, Eugene Kharitonov, Olivier Pietquin, Matt Sharifi, Dominik Roblek, Olivier Teboul, David Grangier, Marco Tagliasacchi, and 1 others. 2023. Audiolm: a language modeling approach to audio generation. IEEE/ACM transactions on audio, speech, and language processing, 31:2523--2533

    2023

  5. [5]

    Hui Bu, Jiayu Du, Xingyu Na, Bengu Wu, and Hao Zheng. 2017. Aishell-1: An open-source mandarin speech corpus and a speech recognition baseline. In 2017 20th conference of the oriental chapter of the international coordinating committee on speech databases and speech I/O systems and assessment (O-COCOSDA), pages 1--5. IEEE

    2017

  6. [6]

    o lge, G \

    Edresson Casanova, Kelly Davis, Eren G \"o lge, G \"o rkem G \"o knar, Iulian Gulea, Logan Hart, Aya Aljafari, Joshua Meyer, Reuben Morais, Samuel Olayemi, and 1 others. 2024. Xtts: a massively multilingual zero-shot text-to-speech model. arXiv preprint arXiv:2406.04904

    arXiv 2024

  7. [7]

    Edresson Casanova, Julian Weber, Christopher D Shulby, Arnaldo Candido Junior, Eren G \"o lge, and Moacir A Ponti. 2022. Yourtts: Towards zero-shot multi-speaker tts and zero-shot voice conversion for everyone. In Proc. ICML, pages 2709--2720. PMLR

    work page 2022

  8. [8]

    Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, and 1 others. 2022. Wavlm: Large-scale self-supervised pre-training for full stack speech processing. IEEE Journal of Selected Topics in Signal Processing, 16(6):1505--1518

    2022

Show all 31 references
  1. [9]

    Hyeong-Seok Choi, Juheon Lee, Wansoo Kim, Jie Lee, Hoon Heo, and Kyogu Lee. 2021. Neural analysis and synthesis: Reconstructing speech from self-supervised representations. Proc. NIPS, 34:16251--16265

    2021

  2. [10]

    Jacob Devlin. 2018. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805

    2018 arXiv

  3. [11]

    Qianqian Dong, Zhiying Huang, Qiao Tian, Chen Xu, Tom Ko, Yunlong Zhao, Siyuan Feng, Tang Li, Kexin Wang, Xuxin Cheng, and 1 others. 2023. Polyvoice: Language models for speech to speech translation. arXiv preprint arXiv:2306.02982

    2023 arXiv

  4. [12]

    Zhihao Du, Qian Chen, Shiliang Zhang, Kai Hu, Heng Lu, Yexin Yang, Hangrui Hu, Siqi Zheng, Yue Gu, Ziyang Ma, and 1 others. 2024 a . Cosyvoice: A scalable multilingual zero-shot text-to-speech synthesizer based on supervised semantic tokens. arXiv preprint arXiv:2407.05407

    2024 arXiv

  5. [13]

    Zhihao Du, Yuxuan Wang, Qian Chen, Xian Shi, Xiang Lv, Tianyu Zhao, Zhifu Gao, Yexin Yang, Changfeng Gao, Hui Wang, and 1 others. 2024 b . Cosyvoice 2: Scalable streaming speech synthesis with large language models. arXiv preprint arXiv:2412.10117

    2024 arXiv

  6. [14]

    Abdelhamid Ezzerg, Thomas Merritt, Kayoko Yanagisawa, Piotr Bilinski, Magdalena Proszewska, Kamil Pokora, Renard Korzeniowski, Roberto Barra-Chicote, and Daniel Korzekwa. 2023. Remap, warp and attend: Non-parallel many-to-many accent conversion with normalizing flows. In 2022 ...

    2023

  7. [15]

    Hee Soo Heo, Bong-Jin Lee, Jaesung Huh, and Joon Son Chung. 2020. Clova baseline system for the voxceleb speaker recognition challenge 2020. arXiv preprint arXiv:2009.14153

    2020 arXiv

  8. [16]

    Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, and Abdelrahman Mohamed. 2021. Hubert: Self-supervised speech representation learning by masked prediction of hidden units. IEEE/ACM transactions on audio, speech, and language processin...

    2021

  9. [17]

    Jaehyeon Kim, Jungil Kong, and Juhee Son. 2021. Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech. In Proc. ICML, pages 5530--5540. PMLR

    2021

  10. [18]

    John Kominek and Alan W Black. 2004. The cmu arctic speech databases. In SSW, pages 223--224

    2004

  11. [19]

    Jingyi Li, Weiping Tu, and Li Xiao. 2023. Freevc: Towards high-quality text-free one-shot voice conversion. In Proc. ICASSP, pages 1--5. IEEE

    2023

  12. [20]

    Songxiang Liu, Disong Wang, Yuewen Cao, Lifa Sun, Xixin Wu, Shiyin Kang, Zhiyong Wu, Xunying Liu, Dan Su, Dong Yu, and 1 others. 2020. End-to-end accent conversion without using native utterances. In Proc. ICASSP, pages 6289--6293. IEEE

    2020

  13. [21]

    Vassil Panayotov, Guoguo Chen, Daniel Povey, and Sanjeev Khudanpur. 2015. Librispeech: an asr corpus based on public domain audio books. In Proc. ICASSP, pages 5206--5210. IEEE

    2015

  14. [22]

    Waris Quamer, Anurag Das, John Levis, Evgeny Chukharev-Hudilainen, and Ricardo Gutierrez-Osuna. 2022. Zero-shot foreign accent conversion without a native reference. Proc. Intespeech

    2022

  15. [23]

    Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2023. Robust speech recognition via large-scale weak supervision. In Proc. ICML, pages 28492--28518. PMLR

    2023

  16. [24]

    Kai Shen, Zeqian Ju, Xu Tan, Yanqing Liu, Yichong Leng, Lei He, Tao Qin, Sheng Zhao, and Jiang Bian. 2023. Naturalspeech 2: Latent diffusion models are natural and zero-shot speech and singing synthesizers. arXiv preprint arXiv:2304.09116

    2023 arXiv

  17. [25]

    Shinnosuke Takamichi, Kentaro Mitsui, Yuki Saito, Tomoki Koriyama, Naoko Tanji, and Hiroshi Saruwatari. 2019. Jvs corpus: free japanese multi-speaker voice corpus. arXiv preprint arXiv:1908.06248

    2019 arXiv

  18. [26]

    Shikhar Vashishth, Harman Singh, Shikhar Bharadwaj, Sriram Ganapathy, Chulayuth Asawaroengchai, Kartik Audhkhasi, Andrew Rosenberg, Ankur Bapna, and Bhuvana Ramabhadran. 2024. Stab: speech tokenizer assessment benchmark. arXiv preprint arXiv:2409.02384

    2024 arXiv

  19. [27]

    Dong Zhang, Shimin Li, Xin Zhang, Jun Zhan, Pengyu Wang, Yaqian Zhou, and Xipeng Qiu. 2023. Speechgpt: Empowering large language models with intrinsic cross-modal conversational abilities. arXiv preprint arXiv:2305.11000

    2023 arXiv

  20. [28]

    Guanlong Zhao, Shaojin Ding, and Ricardo Gutierrez-Osuna. 2021. Converting foreign accent speech without a reference. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 29:2367--2381

    2021

  21. [29]

    Guanlong Zhao and Ricardo Gutierrez-Osuna. 2019. Using phonetic posteriorgram based frame pairing for segmental accent conversion. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 27(10):1649--1660

    2019

  22. [30]

    Yi Zhou, Zhizheng Wu, Mingyang Zhang, Xiaohai Tian, and Haizhou Li. 2023. Tts-guided training for accent conversion without parallel data. IEEE Signal Processing Letters, 30:533--537

    2023

  23. [31]

    Jialong Zuo, Shengpeng Ji, Minghui Fang, Ziyue Jiang, Xize Cheng, Qian Yang, Wenrui Liu, Guangyan Zhang, Zehai Tu, Yiwen Guo, and 1 others. 2025. Enhancing expressive voice conversion with discrete pitch-conditioned flow matching model. arXiv preprint arXiv:2502.05471

    2025 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.