Pith. sign in

REVIEW 4 major objections 5 minor 36 references

Towards Flow-Matching-based TTS without Classifier-Free Guidance

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper claims that replacing the flow-matching training target with a CFG-adjusted target lets a TTS model synthesize speech with a single conditional forward pass, matching or beating the CFG baseline at a fraction of the compute.

desk verdict Useful empirical result obscured by a wrong derivation: Eq (10) does not yield CFG as written, so the central mechanism is unsupported. read the letter →

arxiv 2504.20334 v2 pith:G4QHZ66K submitted 2025-04-29 eess.AS

Yuzhe Liang , Wenzhe Liu , Chunyu Qiang , Zhikang Niu , Yushen Chen , Ziyang Ma , Wenxi Chen , Nan Li
show 2 more authors
Chen Zhang Xie Chen
This is my paper · ORCID
classification eess.AS
keywords classifier-freeguidanceflowmatchingtext-to-speechinferenceefficiencymodelF5-TTSzero-shotTTSconditional
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 asks whether classifier-free guidance (CFG) can be removed from flow-matching-based text-to-speech at inference time without losing quality. It proposes to modify the training objective so the model learns the CFG trajectory directly: the target vector field becomes the conditional field minus a weighted, gradient-stopped difference between conditional and unconditional velocities. At inference the model needs only one conditional forward pass per sampling step instead of two, halving per-step cost. Using F5-TTS on LibriTTS, the method reports WER 2.02%, SIM-O 0.601, MOS 4.101 and RTF 0.04 at 7 steps without CFG, against the F5-TTS baseline with CFG at WER 2.28%, SIM-O 0.592, MOS 4.026 and RTF 0.31 at 32 steps, a 9x inference speed-up. If this holds, CFG's double forward pass can be dropped in real-time TTS systems.

What carries the argument

The key mechanism is the model-guidance loss for conditional flow matching, which appends to the standard conditional flow-matching loss a weighted difference between conditional and unconditional velocity fields, with a stop-gradient on the unconditional branch. It is designed so that the learned conditional field itself matches the CFG-modulated field $v^{\mathrm{CFG}}_t(x|y) = u_t(x) + w\,(u_t(x|y)-u_t(x))$; at inference the sampler solves $dx/dt = v_\theta(x_t,t)$ with only the conditional network evaluation. Gradient stopping protects the unconditional branch from back-propagated gradients, and Sway Sampling or any other ODE solver can be applied without retraining.

What would settle it

The sharpest test is to compute, for fixed $(x,t,c)$, the global minimizer of the training loss in Eq. (10). If that minimizer is not identical to the CFG target $v^{\mathrm{CFG}}_t(x|c)=u_t(x)+w\,(u_t(x|c)-u_t(x))$, the objective does not train the model to reproduce CFG, and any quality gain on the LibriSpeech-PC test set must be attributed to something else. A complementary experiment would compare the single-pass no-CFG model directly with the CFG-guided baseline on identical prompts at matched numbers of function evaluations.

Watch

Extended reading notes

Core claim

The central discovery is that a flow-matching vector field can be trained to approximate the CFG-adjusted target by adding a gradient-stopped penalty term to the conditional velocity target. The training loss becomes $L_{\mathrm{MG-CFM}} = \mathbb{E}\|v_t(x|c) + w \cdot \Delta v_t(x,c) - u_t(x|c)\|^2$, where $\Delta v_t(x,c) = \mathrm{sg}(v_t(x|c)-v_t(x))$ and $\mathrm{sg}$ is the stop-gradient operation. The stated effect is that after training, inference uses only the conditional model output, with no unconditional evaluation and no guidance-scale tuning. The paper verifies this on F5-TTS and finds that at 7 sampling steps the no-CFG model beats the 32-step CFG baseline on word error rate, speaker similarity, and predicted MOS, while running roughly 9 times faster.

Load-bearing premise

The whole fast-inference scheme stands on the premise that the modified training loss in Eq. (10) really teaches the model the classifier-free-guidance trajectory, so that one conditional pass at inference can stand in for two guided passes. If that premise is wrong, the method may still produce decent speech, but not for the reason the paper gives.

Editorial extensions

If this is right

  • Per sampling step, inference cost drops from two forward passes to one, eliminating the unconditional network evaluation and any guidance-scale tuning at decode time.
  • The modified model composes with step-reduction samplers such as Sway Sampling, so per-step savings multiply with step-count savings rather than competing with them.
  • The training budget is not increased overall: although each update needs roughly 1.5x compute, convergence takes about half the updates, so total training cost is comparable.
  • The method is architecture-agnostic within flow-matching TTS: any model trained with conditional/unconditional dropout can adopt the modified target, not only the F5-TTS case study.
  • At the reported operating point (7 steps, no CFG), the model reaches RTF 0.04 on a consumer GPU, which is comfortably real-time.

Reading between the lines

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

  • The paper's 9x speed-up stacks two effects: dropping the second forward pass (2x) and reducing sampling steps from 32 to 7 (about 4.5x). The two are independent, and Table I suggests the no-CFG model's MOS advantage over CFG appears mainly at 7 steps; at 32 steps the no-CFG model scores lower MOS than the CFG baseline. A reader should not read the 9x as attributable to CFG removal alone.
  • A testable extension is to apply the same gradient-stopped model-guidance objective to other flow-matching generators, such as video-to-audio or music generation, where the conditional and unconditional network share parameters; the same one-pass advantage should transfer if the loss is the mechanism.
  • The paper implicitly assumes that the conditional and unconditional fields can be estimated by the same network with dropout. If a future architecture uses separate heads for the two fields, the stop-gradient term and the single-pass trick would need to be re-engineered.
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 / 5 minor

Summary. The paper proposes a training objective for flow-matching-based text-to-speech that aims to eliminate classifier-free guidance at inference time by directly approximating the CFG trajectory. The method modifies the flow-matching loss with a stop-gradient difference term between conditional and unconditional outputs. Experiments on LibriTTS with F5-TTS show improved WER, SIM-O, and MOS at lower NFE, with substantial RTF reduction. However, the derivation of the loss is algebraically incorrect, and the algorithm as written does not match the equation, undermining the central claim.

Significance. If the central claim were correct, the method would be practically significant: it promises roughly a 9x inference speed-up for flow-matching TTS by removing the second forward pass required by CFG, without degrading quality. The empirical results are indeed striking—at NFE 7 without CFG, the proposed model achieves WER 2.02%, SIM-O 0.601, MAS 4.101, and RTF 0.04, compared with the baseline's WER 2.28%, SIM-O 0.592, MOS 4.026, and RTF 0.31 at NFE 32 with CFG. The paper also reports faster convergence, which is interesting. However, the theoretical justification given in Section III.B is not valid, and the training loss as written has a fixed point that moves the conditional field toward the unconditional field, the opposite of CFG's extrapolation. The stop-gradient also makes the target self-referential, so the method is better described as a self-distillation bootstrap than as approximating CFG. The empirical gains are therefore not explained by the paper's narrative, and the method's reproducibility is compromised by an inconsistency between Equation (10) and Algorithm 1.

major comments (4)
  1. [Section III.B, Eq. (10)] The derivation of Equation (10) from Equation (7) is algebraically incorrect. Substituting the model analogue of Equation (7), i.e., v_t(x) + w(v_t(x|y) - v_t(x)), into the loss in Equation (9) would yield E||v_t(x) + w(v_t(x|y) - v_t(x)) - u_t(x|y)||^2, not Equation (10). Moreover, the fixed point of Equation (10) is found by setting the gradient with respect to v_t(x|y) to zero while treating the stop-gradient copy as constant, giving (1+w) v_t(x|y) - w v_t(x) = u_t(x|y), i.e., v_t(x|y) = (u_t(x|y) + w v_t(x)) / (1+w). For w = 0.7, this is a weighted average that damps the difference between conditional and unconditional fields, not the CFG extrapolation of Equation (7), which for w > 1 amplifies the difference. Thus the paper's central claim that the trained conditional model directly approximates the CFG-optimized trajectory is contradicted by its own equations.
  2. [Algorithm 1] Algorithm 1 does not implement Equation (10). Line 6 modifies the target to u'_t = u_t(x0|c) - w·Delta v_t, and Line 7 computes the loss as ||v_t(x0) - u'_t||^2, where v_t(x0) is the unconditional model output. Thus the algorithm trains the unconditional output to match a target that depends on the conditional output and the true conditional velocity, while Equation (10) trains the conditional output. Since the stop-gradient in Delta v_t prevents gradients from flowing through the conditional output, the two losses have fundamentally different gradient flows. This inconsistency makes the method ambiguous and prevents reproduction from the description alone.
  3. [Table I and Abstract] The abstract claims a 9x inference speed-up, but the measured RTF in Table I gives a ratio of 0.31/0.04 = 7.75 for the key comparison. The 9x figure appears to be based on counting forward passes (32 steps x 2 for CFG vs 7 steps x 1), which should be stated explicitly. Additionally, the objective metrics are reported without confidence intervals or statistical significance tests; given that the method's mechanism is not supported by the derivation, these results should be interpreted cautiously.
  4. [Section IV.D.2] The ablation on stop-gradient is described too vaguely. The statement that without stop-gradient 'the model can't inference normally, and the inference results are all noise' is not quantified or analyzed. Since the stop-gradient is essential to the method and creates a self-referential training target (Delta v_t depends on the model's own outputs), the paper should provide a fixed-point or convergence analysis, or at least a clear explanation of why the stop-gradient is necessary. As written, this is a load-bearing unsupported assertion.
minor comments (5)
  1. [Figure 1 caption] The caption refers to 'w (as defined in Equation (8))', but Equation (8) defines the conditional dropout training, not w; w first appears in Equation (10).
  2. [Equation (10)] Equation (10) uses Delta v_t(x,y) without a vertical bar, while the preceding text uses Delta v_t(x|y); the notation should be consistent.
  3. [Abstract] The abstract contains a grammatical error: 'we reformulated' should be 'we reformulate' in the present tense. Also, the sentence 'Furthermore, It can be seamlessly integrated' has an unnecessary capital 'I'.
  4. [Section III.A] The text says 'establishing a linear interpolation between unconditional and conditional dynamics', but for w > 1 the expression is an extrapolation, not an interpolation; this wording is misleading.
  5. [Section IV.D.1] The paper says 'the overall training overhead is approximately 1.5 times higher than that of the previous methods' and later 'this method does not increase training overall overhead'. The first statement refers to per-update overhead, and the second to total training time; this distinction should be made explicit to avoid apparent contradiction.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is benchmarked against external metrics and the cited prior work is not by the present authors.

full rationale

The paper's central claim—that a flow-matching TTS model trained with the model-guidance loss (Eq. 10) can drop CFG at inference—is supported by external benchmark comparisons (WER, SIM-O, MOS, RTF on LibriSpeech-PC) rather than by fitting the reported metrics. The training target is not defined in terms of the evaluation quantities, and no parameter fitted to a subset of the test data is renamed as a prediction. The method is adapted from Tang et al. [20], which is not a self-citation by the present authors; the only self-citation is to the authors' own F5-TTS baseline, which is used as a case-study architecture and is not load-bearing for the theoretical claim. A separate concern, outside the scope of circularity, is that Eq. (10) does not follow algebraically from substituting Eq. (7) into Eq. (9): the fixed point of Eq. (10) is v_c = (u_c + w v_u)/(1+w), which damps the conditional-unconditional difference, whereas Eq. (7) extrapolates it. This is an internal-consistency / derivational flaw, not a circular reduction: the conclusion is not assumed in the premise, and the empirical evaluation remains externally grounded. Accordingly, the circularity score is 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The method relies on the flow-matching framework, the CFG interpolation formula, the transferability of Tang et al.'s model-guidance idea from image diffusion to flow TTS, the use of the network's own unconditional output as a surrogate for the marginal velocity field, and predictor-based metrics. The only fitted central parameter is w=0.7; dropout probabilities come from F5-TTS. No invented entities are introduced.

free parameters (3)
  • w (model-guidance weight) = 0.7
    Set from ablation in Fig. 1; performance improves with w until w >= 1 triggers training collapse. The central no-CFG results in Table I use w=0.7.
  • puncond and p_audio (condition dropout probabilities) = 0.2 all conditions, 0.3 audio condition
    Kept from F5-TTS baseline for fairness; these probabilities shape the unconditional velocity estimate used in the training target.
  • CFG inference scale for baseline = 2
    Baseline F5-TTS uses guidance scale 2; not a free parameter of the proposed method but anchors the comparison.
assumptions (5)
  • standard math Flow matching objective (Eq 1) and the OT linear interpolation path provide a valid target for training.
    Adopted from Lipman et al. [13] without modification.
  • domain assumption The Bayes-rule decomposition and CFG interpolation (Eqs 4-7) extend from score matching to flow-matching vector fields as used here.
    The paper presents this as a direct extension, but vector fields are not scores and the relation is not proven for the flow-matching setting.
  • ad hoc to paper Tang et al.'s model-guidance training transfers from class-conditional diffusion image models to text/audio-conditioned flow-matching TTS.
    Central method; no proof or analysis is provided, only citation [20].
  • domain assumption The network's unconditional output with conditioning dropped is a valid surrogate for the marginal velocity field u_t(x) in the training target.
    Needed for Algorithm 1; the paper does not analyze bias introduced by the shared-parameter estimate.
  • domain assumption Objective metrics (Whisper WER, WavLM SIM-O, NISQA MOS) are reliable proxies for the speech quality claims.
    Evaluation uses predictor models rather than human listening tests; standard in TTS but an assumption about validity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Flow-Matching-based TTS without Classifier-Free Guidance." pith.science (2026). https://pith.science/paper/G4QHZ66K

@misc{pith2026250420334,
  author       = {Pith},
  title        = {Pith review of: Towards Flow-Matching-based TTS without Classifier-Free Guidance},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G4QHZ66K}},
  note         = {Machine review of arXiv:2504.20334}
}
abstract

Flow matching has demonstrated strong generative capabilities and has become a core component in modern Text-to-Speech (TTS) systems. To ensure high-quality speech synthesis, Classifier-Free Guidance (CFG) is widely used during the inference of flow-matching-based TTS models. However, CFG incurs substantial computational cost as it requires two forward passes, which hinders its applicability in real-time scenarios. In this paper, we explore removing CFG from flow-matching-based TTS models to improve inference efficiency, while maintaining performance. Specifically, we reformulated the flow matching training target to directly approximate the CFG optimization trajectory. This training method eliminates the need for unconditional model evaluation and guided tuning during inference, effectively cutting the computational overhead in half. Furthermore, It can be seamlessly integrated with existing optimized sampling strategies. We validate our approach using the F5-TTS model on the LibriTTS dataset. Experimental results show that our method achieves a 9$\times$ inference speed-up compared to the baseline F5-TTS, while preserving comparable speech quality. We will release the code and models to support reproducibility and foster further research in this area.

Figures

Figures reproduced from arXiv: 2504.20334 by the authors.

Figure 1
Figure 1. Effect of the hyperparameter w (as defined in Equation (8)) during training. The orange dotted line and blue dotted line represent the SIM-O and WER of the baseline F5-TTS inference, respectively. We observe that at w = 0.5, the proposed approach outperforms the baseline F5-TTS in terms of WER, and at w = 0.7, it achieves improvements in both WER and SIM-O compared to baseline F5-TTS [PITH_FULL_IMAGE:figures/full_f… view at source ↗
Figure 2
Figure 2. Comparison between the baseline F5-TTS and our proposed approach [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

36 extracted references · 12 canonical work pages

  1. [1]

    Funaudiollm: V oice understanding and generation foundation models for natural interaction between humans and llms,

    K. An, Q. Chen, C. Deng, Z. Du, C. Gao, Z. Gao, Y . Gu, T. He, H. Hu, K. Hu et al., “Funaudiollm: V oice understanding and generation foundation models for natural interaction between humans and llms,” arXiv preprint arXiv:2407.04051 , 2024

    arXiv 2024

  2. [2]

    Cosyvoice 2: Scalable streaming speech synthesis with large language models,

    Z. Du, Y . Wang, Q. Chen, X. Shi, X. Lv, T. Zhao, Z. Gao, Y . Yang, C. Gao, H. Wang et al. , “Cosyvoice 2: Scalable streaming speech synthesis with large language models,” arXiv preprint arXiv:2412.10117, 2024

    arXiv 2024

  3. [3]

    Viola: Unified codec language models for speech recognition, synthesis, and translation,

    T. Wang, L. Zhou, Z. Zhang, Y . Wu, S. Liu, Y . Gaur, Z. Chen, J. Li, and F. Wei, “Viola: Unified codec language models for speech recognition, synthesis, and translation,” arXiv preprint arXiv:2305.16107 , 2023

    arXiv 2023

  4. [4]

    Naturalspeech 3: Zero-shot speech syn- thesis with factorized codec and diffusion models,

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

    arXiv 2024

  5. [5]

    E2 TTS: Embarrass- ingly 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. Tan et al. , “E2 TTS: Embarrass- ingly easy fully non-autoregressive zero-shot TTS,” arXiv preprint arXiv:2406.18009, 2024

    arXiv 2024

  6. [6]

    F5-TTS: A fairytaler that fakes fluent and faithful speech with flow matching,

    Y . Chen, Z. Niu, Z. Ma, K. Deng, C. Wang, J. Zhao, K. Yu, and X. Chen, “F5-TTS: A fairytaler that fakes fluent and faithful speech with flow matching,” arXiv preprint arXiv:2410.06885 , 2024

    arXiv 2024

  7. [7]

    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. IEEE, 2024, pp. 11 341–11 345

    work page 2024

  8. [8]

    Language mod- els are few-shot learners,

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

    1901

Show all 36 references
  1. [9]

    Diffusion models beat gans on image synthesis,

    P. Dhariwal and A. Nichol, “Diffusion models beat gans on image synthesis,” Proc. NeurIPS, vol. 34, pp. 8780–8794, 2021

    2021

  2. [10]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Proc. NeurIPS, vol. 33, pp. 6840–6851, 2020

    2020

  3. [11]

    Denoising diffusion implicit models,

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” arXiv preprint arXiv:2010.02502 , 2020

    2010 arXiv

  4. [12]

    Improved denoising diffusion probabilis- tic models,

    A. Q. Nichol and P. Dhariwal, “Improved denoising diffusion probabilis- tic models,” in Proc .ICML. PMLR, 2021, pp. 8162–8171

    2021

  5. [13]

    Flow matching for generative modeling,

    Y . Lipman, R. T. Chen, H. Ben-Hamu, M. Nickel, and M. Le, “Flow matching for generative modeling,” in ICLR 2023, 2023

    2023

  6. [14]

    FastDiff: A fast conditional diffusion model for high-quality speech synthesis,

    R. Huang, M. W. Lam, J. Wang, D. Su, D. Yu, Y . Ren, and Z. Zhao, “FastDiff: A fast conditional diffusion model for high-quality speech synthesis,” arXiv preprint arXiv:2204.09934 , 2022

    2022 arXiv

  7. [15]

    ProDiff: Progressive fast diffusion model for high-quality text-to-speech,

    R. Huang, Z. Zhao, H. Liu, J. Liu, C. Cui, and Y . Ren, “ProDiff: Progressive fast diffusion model for high-quality text-to-speech,” in Proceedings of the 30th ACM International Conference on Multimedia , 2022, pp. 2595–2605

    2022

  8. [16]

    Diff- TTS: A denoising diffusion model for text-to-speech,

    M. Jeong, H. Kim, S. J. Cheon, B. J. Choi, and N. S. Kim, “Diff- TTS: A denoising diffusion model for text-to-speech,” arXiv preprint arXiv:2104.01409, 2021

    2021 arXiv

  9. [17]

    CoMoSpeech: One-step speech and singing voice synthesis via consistency model,

    Z. Ye, W. Xue, X. Tan, J. Chen, Q. Liu, and Y . Guo, “CoMoSpeech: One-step speech and singing voice synthesis via consistency model,” in Proceedings of the 31st ACM International Conference on Multimedia , 2023, pp. 1831–1839

    2023

  10. [18]

    Reflow- TTS: A rectified flow model for high-fidelity text-to-speech,

    W. Guan, Q. Su, H. Zhou, S. Miao, X. Xie, L. Li, and Q. Hong, “Reflow- TTS: A rectified flow model for high-fidelity text-to-speech,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 10 501–10 505

    2024

  11. [19]

    FlashSpeech: Efficient zero-shot speech synthesis,

    Z. Ye, Z. Ju, H. Liu, X. Tan, J. Chen, Y . Lu, P. Sun, J. Pan, W. Bian, S. He et al. , “FlashSpeech: Efficient zero-shot speech synthesis,” in Proceedings of the 32nd ACM International Conference on Multimedia , 2024, pp. 6998–7007

    2024

  12. [20]

    Diffusion models without Classifier-Free Guidance,

    Z. Tang, J. Bao, D. Chen, and B. Guo, “Diffusion models without Classifier-Free Guidance,” arXiv preprint arXiv:2502.12154 , 2025

    2025 arXiv

  13. [21]

    Classifier-free diffusion guidance,

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

    2022 arXiv

  14. [22]

    Score-based generative modeling through stochastic differ- ential equations,

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole, “Score-based generative modeling through stochastic differ- ential equations,” in International Conference on Learning Representa- tions

  15. [23]

    Improving and generalizing flow-based generative models with minibatch optimal transport,

    A. Tong, K. Fatras, N. Malkin, G. Huguet, Y . Zhang, J. Rector- Brooks, G. Wolf, and Y . Bengio, “Improving and generalizing flow-based generative models with minibatch optimal transport,” arXiv preprint arXiv:2302.00482, 2023

    2023 arXiv

  16. [24]

    V oicebox: Text-guided mul- tilingual 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 mul- tilingual universal speech generation at scale,” Proc. NeurIPS, vol. 36, pp. 14 005–14 034, 2023

    2023

  17. [25]

    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. Gao et al., “Seed-TTS: A family of high-quality versatile speech generation models,” arXiv preprint arXiv:2406.02430 , 2024

    2024 arXiv

  18. [26]

    FireredTTS: A foundation text-to-speech framework for industry-level generative speech applications,

    H.-H. Guo, Y . Hu, K. Liu, F.-Y . Shen, X. Tang, Y .-C. Wu, F.-L. Xie, K. Xie, and K.-T. Xu, “FireredTTS: A foundation text-to-speech framework for industry-level generative speech applications,” arXiv preprint arXiv:2409.03283, 2024

    2024 arXiv

  19. [27]

    Autoregressive diffusion transformer for text-to-speech synthesis,

    Z. Liu, S. Wang, S. Inoue, Q. Bai, and H. Li, “Autoregressive diffusion transformer for text-to-speech synthesis,” arXiv preprint arXiv:2406.05551, 2024

    2024 arXiv

  20. [28]

    DiTAR: Diffusion Transformer Autoregressive Modeling for Speech Generation,

    D. Jia, Z. Chen, J. Chen, C. Du, J. Wu, J. Cong, X. Zhuang, C. Li, Z. Wei, Y . Wanget al., “DiTAR: Diffusion Transformer Autoregressive Modeling for Speech Generation,” arXiv preprint arXiv:2502.03930 , 2025

    2025

  21. [29]

    Frieren: Efficient video-to-audio generation network with rectified flow matching,

    Y . Wang, W. Guo, R. Huang, J. Huang, Z. Wang, F. You, R. Li, and Z. Zhao, “Frieren: Efficient video-to-audio generation network with rectified flow matching,” Advances in Neural Information Processing Systems, vol. 37, pp. 128 118–128 138, 2024

    2024

  22. [30]

    Null- text inversion for editing real images using guided diffusion models,

    R. Mokady, A. Hertz, K. Aberman, Y . Pritch, and D. Cohen-Or, “Null- text inversion for editing real images using guided diffusion models,” in Proc. CVPR, 2023, pp. 6038–6047

    2023

  23. [31]

    CFG++: Manifold- constrained classifier free guidance for diffusion models,

    H. Chung, J. Kim, G. Y . Park, H. Nam, and J. C. Ye, “CFG++: Manifold- constrained classifier free guidance for diffusion models,” arXiv preprint arXiv:2406.08070, 2024

    2024 arXiv

  24. [32]

    CFG-Zero*: Improved classifier-free guidance for flow matching models,

    W. Fan, A. Y . Zheng, R. A. Yeh, and Z. Liu, “CFG-Zero*: Improved classifier-free guidance for flow matching models,” arXiv preprint arXiv:2503.18886, 2025

    2025 arXiv

  25. [33]

    Classifier-Free Guidance is a predictor- corrector,

    A. Bradley and P. Nakkiran, “Classifier-Free Guidance is a predictor- corrector,” arXiv preprint arXiv:2408.09000 , 2024

    2024 arXiv

  26. [34]

    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

    1904 arXiv

  27. [35]

    Large-scale self-supervised speech representation learning for automatic speaker verification,

    Z. Chen, S. Chen, Y . Wu, Y . Qian, C. Wang, S. Liu, Y . Qian, and M. Zeng, “Large-scale self-supervised speech representation learning for automatic speaker verification,” in ICASSP 2022-2022 IEEE Interna- tional Conference on Acoustics, Speech and Signal Processing (ICASSP)....

    2022

  28. [36]

    NISQA: A deep cnn- self-attention model for multidimensional speech quality prediction with crowdsourced datasets,

    G. Mittag, B. Naderi, A. Chehadi, and S. M ¨oller, “NISQA: A deep cnn- self-attention model for multidimensional speech quality prediction with crowdsourced datasets,” arXiv preprint arXiv:2104.09494 , 2021

    2021 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.