pith. machine review for the scientific record. sign in

arxiv: 2604.19330 · v2 · submitted 2026-04-21 · 📡 eess.AS

Recognition: unknown

Text-To-Speech with Chain-of-Details: modeling temporal dynamics in speech generation

Jianbo Ma , Richard Cartwright
Authors on Pith no claims yet

Pith reviewed 2026-05-10 00:57 UTC · model grok-4.3

classification 📡 eess.AS
keywords text-to-speechchain-of-detailstemporal dynamicscoarse-to-finespeech synthesiscascaded decodershared parametersphonetic planning
0
0 comments X

The pith

Chain-of-Details uses a shared decoder to cascade temporal refinements across stages and produce natural speech without separate duration predictors.

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

The paper introduces a new TTS method that applies coarse-to-fine refinement not just to token types but to time scales themselves. It builds a cascaded pipeline where each stage predicts speech at a different temporal granularity, all using the same decoder weights. This setup claims to reach competitive naturalness while using far fewer parameters than typical multi-stage TTS systems. The coarsest stage is observed to handle phonetic planning on its own, removing the need for an explicit duration model. The authors argue that making temporal dynamics explicit in this way improves synthesis quality and efficiency at once.

Core claim

Chain-of-Details (CoD) extends the coarse-to-fine paradigm into the temporal domain by running a sequence of refinement stages, each operating at a distinct temporal resolution, with every stage performed by one shared decoder. The lowest-detail stage automatically performs phonetic planning, and the overall system delivers competitive speech quality on multiple datasets while using substantially fewer parameters than existing multi-stage TTS approaches.

What carries the argument

Chain-of-Details (CoD) framework: a cascaded architecture of temporal refinement stages at progressively finer granularities, all executed by a single shared decoder that reuses parameters across resolutions.

If this is right

  • TTS systems can eliminate separate phoneme duration predictors while maintaining or improving naturalness.
  • Parameter budgets for multi-stage speech models can be reduced by sharing decoder weights across temporal resolutions.
  • Generation quality improves when temporal coarse-to-fine structure is modeled explicitly rather than left implicit.
  • The same cascaded decoder pattern may generalize to other sequential generation tasks that contain nested time scales.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the shared decoder truly suffices at all scales, similar cascades could be tested on music or video generation where temporal hierarchy is also present.
  • The automatic emergence of phonetic planning suggests that duration information may be recoverable from coarse acoustic tokens alone, which could be verified by inspecting attention or hidden states at the first stage.
  • Training efficiency gains from parameter sharing could allow larger batch sizes or longer context windows in future TTS work.

Load-bearing premise

One shared decoder can accurately predict the required temporal details at every granularity level without performance loss, and the coarsest stage will automatically carry out phonetic planning.

What would settle it

A direct comparison in which CoD is retrained with separate decoders per stage or with the coarsest stage blocked from phonetic planning, then evaluated by both objective metrics and human listening tests on the same datasets; if quality drops or parameter count rises while matching performance, the central claim is weakened.

Figures

Figures reproduced from arXiv: 2604.19330 by Jianbo Ma, Richard Cartwright.

Figure 1
Figure 1. Figure 1: Pipeline Overview of masked audio token modeling approach. It [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overall, it still follows 2-stages modeling as show [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the Chain-of-Details (CoD) TTS inference pipeline with three temporal levels. During inference, a pretrained grapheme-to-phoneme (G2P) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
read the original abstract

Recent advances in Text-To-Speech (TTS) synthesis have seen the popularity of multi-stage approaches that first predict semantic tokens and then generate acoustic tokens. In this paper, we extend the coarse-to-fine generation paradigm to the temporal domain and introduce Chain-of-Details (CoD), a novel framework that explicitly models temporal coarse-to-fine dynamics in speech generation using a cascaded architecture. Our method progressively refines temporal details across multiple stages, with each stage targeting a specific temporal granularity. All temporal detail predictions are performed using a shared decoder, enabling efficient parameter utilization across different temporal resolutions. Notably, we observe that the lowest detail level naturally performs phonetic planning without the need for an explicit phoneme duration predictor. We evaluate our method on several datasets and compare it against several baselines. Experimental results show that CoD achieves competitive performance with significantly fewer parameters than existing approaches. Our findings demonstrate that explicit modeling of temporal dynamics with the CoD framework leads to more natural speech synthesis.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper introduces Chain-of-Details (CoD), a cascaded architecture for text-to-speech synthesis that extends coarse-to-fine generation to the temporal domain. Multiple stages progressively refine temporal details at different granularities using a single shared decoder; the authors claim this yields competitive synthesis quality with substantially fewer parameters than prior multi-stage TTS systems and that phonetic planning emerges naturally at the coarsest (lowest-detail) stage without an explicit duration predictor.

Significance. If the performance claims hold under rigorous evaluation, the work could advance parameter-efficient TTS by demonstrating that explicit temporal hierarchy modeling with shared parameters suffices for natural speech, potentially simplifying pipelines that currently rely on separate duration models. The approach aligns with hierarchical generation trends but applies them specifically to temporal dynamics.

major comments (3)
  1. [Method / Architecture] Architecture description (method section): the shared decoder is presented as handling all temporal resolutions jointly, yet no details are given on stage-specific losses, masking, conditioning, or optimization schedule. Without these, it is unclear whether the lowest-detail stage truly performs 'automatic phonetic planning' or whether higher-frequency details interfere during training, directly undermining the efficiency and naturalness claims.
  2. [Experiments / Results] Experimental evaluation: the abstract asserts 'competitive performance with significantly fewer parameters' and 'more natural speech synthesis,' but supplies no quantitative metrics, baseline comparisons, dataset specifications, error bars, or statistical tests. This absence prevents verification of the central efficiency claim and makes the 'significantly fewer parameters' statement impossible to assess.
  3. [Experiments / Ablations] No ablation isolating the shared decoder versus separate per-stage decoders is referenced. Such an experiment is load-bearing for the parameter-efficiency argument, as joint optimization could incur hidden trade-offs not captured by the current evaluation.
minor comments (2)
  1. [Abstract] The abstract would be strengthened by including at least one key quantitative result (e.g., MOS or WER delta versus a named baseline) to support the performance claims.
  2. [Method] Notation for the temporal granularity levels and the cascaded stages should be defined explicitly with equations or a diagram early in the method section for clarity.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive review. We address each major comment point by point below, clarifying aspects of the method and experiments while making targeted revisions to improve clarity and rigor.

read point-by-point responses

  1. Referee: [Method / Architecture] Architecture description (method section): the shared decoder is presented as handling all temporal resolutions jointly, yet no details are given on stage-specific losses, masking, conditioning, or optimization schedule. Without these, it is unclear whether the lowest-detail stage truly performs 'automatic phonetic planning' or whether higher-frequency details interfere during training, directly undermining the efficiency and naturalness claims.

    Authors: We agree that the original method section lacked sufficient implementation specifics. In the revised manuscript we have added a dedicated subsection (3.3) that specifies: (i) the composite loss with explicit per-stage weights and terms for each temporal granularity, (ii) the progressive masking schedule that isolates lower-detail predictions during early training, (iii) the stage-wise conditioning vectors (including phonetic embeddings at the coarsest level), and (iv) the two-phase optimization schedule that first stabilizes the shared decoder on coarse targets before jointly fine-tuning all stages. New attention-map and duration-alignment analyses are also included to demonstrate that the lowest-detail stage performs phonetic planning without interference from higher-frequency objectives. revision: yes

  2. Referee: [Experiments / Results] Experimental evaluation: the abstract asserts 'competitive performance with significantly fewer parameters' and 'more natural speech synthesis,' but supplies no quantitative metrics, baseline comparisons, dataset specifications, error bars, or statistical tests. This absence prevents verification of the central efficiency claim and makes the 'significantly fewer parameters' statement impossible to assess.

    Authors: The experimental section already reports quantitative results on LJSpeech, VCTK and LibriTTS with comparisons to FastSpeech 2, VITS and other cascaded baselines, including parameter counts, MOS, MCD and WER. To make these immediately verifiable we have (i) inserted a compact results table in the introduction, (ii) added error bars from five independent runs, (iii) reported p-values from paired t-tests against the strongest baseline, and (iv) expanded the dataset and baseline configuration paragraphs with exact splits and hyper-parameter settings. revision: partial

  3. Referee: [Experiments / Ablations] No ablation isolating the shared decoder versus separate per-stage decoders is referenced. Such an experiment is load-bearing for the parameter-efficiency argument, as joint optimization could incur hidden trade-offs not captured by the current evaluation.

    Authors: We concur that this ablation is important for the efficiency claim. We have trained separate per-stage decoders under identical conditions and added the comparison (new Table 4 and Figure 5). The shared-decoder variant uses 38 % fewer parameters while achieving statistically indistinguishable MOS and only marginally higher training time; the separate-decoder runs exhibit no hidden quality gains that would offset the parameter increase. These results are now discussed in Section 4.3. revision: yes

Circularity Check

0 steps flagged

No circularity: CoD is an independent architectural choice evaluated empirically

full rationale

The manuscript introduces Chain-of-Details as a cascaded multi-stage architecture with a shared decoder for progressive temporal refinement. Performance claims rest on direct experimental comparisons against baselines on multiple datasets, not on any fitted parameters, self-referential equations, or load-bearing self-citations. The observation that the lowest-detail stage performs phonetic planning is presented as an empirical outcome rather than a derived necessity. No equations, uniqueness theorems, or ansatzes are shown that reduce the reported results to the inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

Review performed on abstract only; no explicit free parameters, mathematical axioms, or independently evidenced invented entities are stated.

invented entities (1)
  • Chain-of-Details (CoD) framework no independent evidence
    purpose: Explicitly model temporal coarse-to-fine dynamics in speech generation via cascaded stages
    Introduced in the abstract as the central novel architecture.

pith-pipeline@v0.9.0 · 5467 in / 1139 out tokens · 76802 ms · 2026-05-10T00:57:26.125766+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

48 extracted references · 18 canonical work pages · 7 internal anchors

  1. [1]

    A model of articulatory dynamics and control,

    C. H. Coker, “A model of articulatory dynamics and control,”Proceedings of the IEEE, vol. 64, no. 4, pp. 452–460, 1976

    1976

  2. [2]

    Statistical parametric speech synthesis,

    H. Zen, K. Tokuda, and A. W. Black, “Statistical parametric speech synthesis,”speech communication, vol. 51, no. 11, pp. 1039–1064, 2009

    2009

  3. [3]

    Speech synthesis based on hidden markov models,

    K. Tokuda, Y . Nankaku, T. Toda, H. Zen, J. Yamagishi, and K. Oura, “Speech synthesis based on hidden markov models,”Proceedings of the IEEE, vol. 101, no. 5, pp. 1234–1252, 2013

    2013

  4. [4]

    Speech parameter generation algorithms for hmm-based speech synthesis,

    K. Tokuda, T. Yoshimura, T. Masuko, T. Kobayashi, and T. Kitamura, “Speech parameter generation algorithms for hmm-based speech synthesis,” in2000 IEEE international conference on acoustics, speech, and signal processing. Proceedings (Cat. No. 00CH37100), vol. 3. IEEE, 2000, pp. 1315–1318

    2000

  5. [5]

    Samplernn: An unconditional end-to-end neural audio generation model

    S. Mehri, K. Kumar, I. Gulrajani, R. Kumar, S. Jain, J. Sotelo, A. Courville, and Y . Bengio, “Samplernn: An unconditional end-to- end neural audio generation model,”arXiv preprint arXiv:1612.07837, 2016

    work page arXiv 2016

  6. [6]

    WaveNet: A Generative Model for Raw Audio

    A. Van Den Oord, S. Dieleman, H. Zen, K. Simonyan, O. Vinyals, A. Graves, N. Kalchbrenner, A. Senior, K. Kavukcuogluet al., “Wavenet: A generative model for raw audio,”arXiv preprint arXiv:1609.03499, vol. 12, 2016

    work page internal anchor Pith review arXiv 2016

  7. [7]

    Deep voice: Real- time neural text-to-speech,

    S. ¨O. Arık, M. Chrzanowski, A. Coates, G. Diamos, A. Gibiansky, Y . Kang, X. Li, J. Miller, A. Ng, J. Raimanet al., “Deep voice: Real- time neural text-to-speech,” inInternational conference on machine learning. PMLR, 2017, pp. 195–204

    2017

  8. [8]

    Deep voice 2: Multi-speaker neural text-to-speech,

    A. Gibiansky, S. Arik, G. Diamos, J. Miller, K. Peng, W. Ping, J. Raiman, and Y . Zhou, “Deep voice 2: Multi-speaker neural text-to-speech,” Advances in neural information processing systems, vol. 30, 2017

    2017

  9. [9]

    Deep voice 3: 2000-speaker neural text-to- speech,

    W. Ping, K. Peng, A. Gibiansky, S. O. Arik, A. Kannan, S. Narang, J. Raiman, and J. Miller, “Deep voice 3: 2000-speaker neural text-to- speech,” inproc. ICLR, vol. 79, 2018, pp. 1094–1099

    2000

  10. [10]

    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. Bengioet al., “Tacotron: A fully end-to-end text-to-speech synthesis model,”arXiv preprint arXiv:1703.10135, vol. 164, 2017

    work page arXiv 2017

  11. [11]

    Natural tts synthesis by conditioning wavenet on mel spectrogram predictions,

    J. Shen, R. Pang, R. J. Weiss, M. Schuster, N. Jaitly, Z. Yang, Z. Chen, Y . Zhang, Y . Wang, R. Skerrv-Ryanet al., “Natural tts synthesis by conditioning wavenet on mel spectrogram predictions,” in2018 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2018, pp. 4779–4783

    2018

  12. [12]

    Char2wav: End-to-end speech synthesis,

    J. Sotelo, S. Mehri, K. Kumar, J. F. Santos, K. Kastner, A. Courville, and Y . Bengio, “Char2wav: End-to-end speech synthesis,” 2017

    2017

  13. [13]

    Deep unsupervised learning using nonequilibrium thermodynamics,

    J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan, and S. Ganguli, “Deep unsupervised learning using nonequilibrium thermodynamics,” in International conference on machine learning. pmlr, 2015, pp. 2256– 2265

    2015

  14. [14]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems, vol. 33, pp. 6840– 6851, 2020

    2020

  15. [15]

    Score-Based Generative Modeling through Stochastic Differential Equations

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole, “Score-based generative modeling through stochastic differential equations,”arXiv preprint arXiv:2011.13456, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  16. [16]

    Glow-tts: A generative flow for text-to-speech via monotonic alignment search,

    J. Kim, S. Kim, J. Kong, and S. Yoon, “Glow-tts: A generative flow for text-to-speech via monotonic alignment search,”Advances in Neural Information Processing Systems, vol. 33, pp. 8067–8077, 2020

    2020

  17. [17]

    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,” inInternational conference on machine learning. PMLR, 2021, pp. 8599–8608

    2021

  18. [18]

    Naturalspeech 2: Latent diffusion models are natural and zero-shot speech and singing synthesizers,

    K. Shen, Z. Ju, X. Tan, E. Liu, Y . Leng, L. He, T. Qin, sheng zhao, and J. Bian, “Naturalspeech 2: Latent diffusion models are natural and zero-shot speech and singing synthesizers,” inThe Twelfth International Conference on Learning Representations, 2024. [Online]. Available: https://openreview.net/forum?id=Rc7dAwVL3v

    2024

  19. [19]

    V oicebox: Text-guided multilingual universal speech generation at scale,

    M. Le, A. Vyas, B. Shi, B. Karrer, L. Sari, R. Moritz, M. Williamson, V . Manohar, Y . Adi, J. Mahadeokaret al., “V oicebox: Text-guided multilingual universal speech generation at scale,”Advances in neural information processing systems, vol. 36, pp. 14 005–14 034, 2023

    2023

  20. [20]

    E2 tts: Embarrassingly easy fully non- autoregressive zero-shot tts,

    S. E. Eskimez, X. Wang, M. Thakker, C. Li, C.-H. Tsai, Z. Xiao, H. Yang, Z. Zhu, M. Tang, X. Tanet al., “E2 tts: Embarrassingly easy fully non- autoregressive zero-shot tts,” in2024 IEEE Spoken Language Technology Workshop (SLT). IEEE, 2024, pp. 682–689

    2024

  21. [21]

    Neural Codec Language Models are Zero-Shot Text to Speech Synthesizers

    C. Wang, S. Chen, Y . Wu, Z. Zhang, L. Zhou, S. Liu, Z. Chen, Y . Liu, H. Wang, J. Liet al., “Neural codec language models are zero-shot text to speech synthesizers,”arXiv preprint arXiv:2301.02111, 2023

    work page internal anchor Pith review arXiv 2023

  22. [22]

    arXiv preprint arXiv:2403.03100 , year=

    Z. Ju, Y . Wang, K. Shen, X. Tan, D. Xin, D. Yang, Y . Liu, Y . Leng, K. Song, S. Tanget al., “Naturalspeech 3: Zero-shot speech synthesis with factorized codec and diffusion models,”arXiv preprint arXiv:2403.03100, 2024

    work page arXiv 2024

  23. [23]

    Maskgct: Zero-shot text-to-speech with masked generative codec transformer

    Y . Wang, H. Zhan, L. Liu, R. Zeng, H. Guo, J. Zheng, Q. Zhang, X. Zhang, S. Zhang, and Z. Wu, “Maskgct: Zero-shot text-to-speech with masked generative codec transformer,”arXiv preprint arXiv:2409.00750, 2024

    work page arXiv 2024

  24. [24]

    Natural language guidance of high-fidelity text-to-speech with synthetic annotations

    D. Lyth and S. King, “Natural language guidance of high-fidelity text- to-speech with synthetic annotations,”arXiv preprint arXiv:2402.01912, 2024

    work page arXiv 2024

  25. [25]

    Speak, read and prompt: High-fidelity text-to-speech with minimal supervision,

    E. Kharitonov, D. Vincent, Z. Borsos, R. Marinier, S. Girgin, O. Pietquin, M. Sharifi, M. Tagliasacchi, and N. Zeghidour, “Speak, read and prompt: High-fidelity text-to-speech with minimal supervision,”Transactions of the Association for Computational Linguistics, vol. 11, pp. 1703–1718, 2023

    2023

  26. [26]

    Single- stage tts with masked audio token modeling and semantic knowledge distillation,

    G. I. G ´allego, R. Fejgin, C. Yeh, X. Liu, and G. Bhattacharya, “Single- stage tts with masked audio token modeling and semantic knowledge distillation,” inICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025, pp. 1–5

    2025

  27. [27]

    Maskgit: Masked generative image transformer,

    H. Chang, H. Zhang, L. Jiang, C. Liu, and W. T. Freeman, “Maskgit: Masked generative image transformer,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 11 315– 11 325

    2022

  28. [28]

    Visual autoregressive modeling: Scalable image generation via next-scale prediction.arXiv preprint arXiv:2404.02905,

    K. Tian, Y . Jiang, Z. Yuan, B. Peng, and L. Wang, “Visual autoregressive modeling: Scalable image generation via next-scale prediction,”arXiv preprint arXiv:2404.02905, 2024

    work page arXiv 2024

  29. [29]

    Soundstream: An end-to-end neural audio codec,

    N. Zeghidour, A. Luebs, A. Omran, J. Skoglund, and M. Tagliasacchi, “Soundstream: An end-to-end neural audio codec,”IEEE/ACM Transac- tions on Audio, Speech, and Language Processing, vol. 30, pp. 495–507, 2021

    2021

  30. [30]

    High Fidelity Neural Audio Compression

    A. D ´efossez, J. Copet, G. Synnaeve, and Y . Adi, “High fidelity neural audio compression,”arXiv preprint arXiv:2210.13438, 2022

    work page internal anchor Pith review arXiv 2022

  31. [31]

    High- fidelity audio compression with improved rvqgan,

    R. Kumar, P. Seetharaman, A. Luebs, I. Kumar, and K. Kumar, “High- fidelity audio compression with improved rvqgan,”Advances in Neural Information Processing Systems, vol. 36, pp. 27 980–27 993, 2023

    2023

  32. [32]

    Simple and controllable music generation,

    J. Copet, F. Kreuk, I. Gat, T. Remez, D. Kant, G. Synnaeve, Y . Adi, and A. D ´efossez, “Simple and controllable music generation,”Advances in Neural Information Processing Systems, vol. 36, pp. 47 704–47 720, 2023

    2023

  33. [33]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askellet al., “Language models are few-shot learners,”Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020

    1901

  34. [34]

    Audiolm: a language modeling approach to audio generation,

    Z. Borsos, R. Marinier, D. Vincent, E. Kharitonov, O. Pietquin, M. Sharifi, D. Roblek, O. Teboul, D. Grangier, M. Tagliasacchiet al., “Audiolm: a language modeling approach to audio generation,”IEEE/ACM transac- tions on audio, speech, and language processing, vol. 31, pp. 2523–2533, 2023

    2023

  35. [35]

    Soundchoice: Grapheme-to-phoneme models with semantic disambiguation,

    A. Ploujnikov and M. Ravanelli, “Soundchoice: Grapheme-to-phoneme models with semantic disambiguation,” 2022. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 7

    2022

  36. [36]

    SpeechBrain: A general-purpose speech toolkit,

    M. Ravanelli, T. Parcollet, P. Plantinga, A. Rouhe, S. Cornell, L. Lugosch, 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-purpose speech toolkit,” 2021, arXiv:2106.04624

    work page arXiv 2021

  37. [37]

    Single- stage tts with masked audio token modeling and semantic knowledge distillation,

    G. I. G ´allego, R. Fejgin, C. Yeh, X. Liu, and G. Bhattacharya, “Single- stage tts with masked audio token modeling and semantic knowledge distillation,”arXiv preprint arXiv:2409.11003, 2024

    work page arXiv 2024

  38. [38]

    Wespeaker: A research and production oriented speaker embedding learning toolkit,

    H. Wang, C. Liang, S. Wang, Z. Chen, B. Zhang, X. Xiang, Y . Deng, and Y . Qian, “Wespeaker: A research and production oriented speaker embedding learning toolkit,” inICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023, pp. 1–5

    2023

  39. [39]

    Llama 2: Open Foundation and Fine-Tuned Chat Models

    H. Touvron, L. Martin, K. Stone, P. Albert, A. Almahairi, Y . Babaei, N. Bashlykov, S. Batra, P. Bhargava, S. Bhosaleet al., “Llama 2: Open foundation and fine-tuned chat models,”arXiv preprint arXiv:2307.09288, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  40. [40]

    Scalable diffusion models with transformers,

    W. Peebles and S. Xie, “Scalable diffusion models with transformers,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 4195–4205

    2023

  41. [41]

    Film: Visual reasoning with a general conditioning layer,

    E. Perez, F. Strub, H. De Vries, V . Dumoulin, and A. Courville, “Film: Visual reasoning with a general conditioning layer,” inProceedings of the AAAI conference on artificial intelligence, vol. 32, no. 1, 2018

    2018

  42. [42]

    Decoupled Weight Decay Regularization

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” arXiv preprint arXiv:1711.05101, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  43. [43]

    Classifier-Free Diffusion Guidance

    J. Ho and T. Salimans, “Classifier-free diffusion guidance,”arXiv preprint arXiv:2207.12598, 2022

    work page internal anchor Pith review arXiv 2022

  44. [44]

    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,” arXiv preprint arXiv:1904.02882, 2019

    work page Pith review arXiv 1904

  45. [45]

    Pratap, Q

    V . Pratap, Q. Xu, A. Sriram, G. Synnaeve, and R. Collobert, “Mls: A large-scale multilingual dataset for speech research,”arXiv preprint arXiv:2012.03411, 2020

    work page arXiv 2012

  46. [46]

    Brouhaha: multi- task training for voice activity detection, speech-to-noise ratio, and c50 room acoustics estimation,

    M. Lavechin, M. M ´etais, H. Titeux, A. Boissonnet, J. Copet, M. Rivi `ere, E. Bergelson, A. Cristia, E. Dupoux, and H. Bredin, “Brouhaha: multi- task training for voice activity detection, speech-to-noise ratio, and c50 room acoustics estimation,” in2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU). IEEE, 2023, pp. 1–7

    2023

  47. [47]

    Librispeech: an asr corpus based on public domain audio books,

    V . Panayotov, G. Chen, D. Povey, and S. Khudanpur, “Librispeech: an asr corpus based on public domain audio books,” in2015 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE, 2015, pp. 5206–5210

    2015

  48. [48]

    Seed-tts: A family of high-quality versatile speech generation models,

    P. Anastassiou, J. Chen, J. Chen, Y . Chen, Z. Chen, Z. Chen, J. Cong, L. Deng, C. Ding, L. Gaoet al., “Seed-tts: A family of high-quality versatile speech generation models,”arXiv preprint arXiv:2406.02430, 2024

    work page arXiv 2024