Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Bridge-SR: Schr\"odinger Bridge for Efficient SR

T0 review · 3 major / 6 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 Bridge-SR claims that speech super-resolution should be modeled as a Schrödinger bridge from the low-resolution waveform to the high-resolution one, and that this data-to-data path beats conditional diffusion in both quality and number of…

desk verdict Solid empirical adaptation of Schrödinger bridges to waveform speech SR; the 4-step claim is plausible but the evaluation needs a proper validation split before the numbers can be trusted. read the letter →

arxiv 2501.07897 v1 pith:JAVZ35SF submitted 2025-01-14 cs.SD eess.AS

Chang Li , Zehua Chen , Fan Bao , Jun Zhu This is my paper · ORCID
classification cs.SDeess.AS
keywords speechsuper-resolutionSchrödingerbridgescore-basedgenerativemodelswaveformgenerationdiffusionnoisescheduleaudiorestorationefficientinference
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

Bridge-SR proposes that speech super-resolution should be modeled as a data-to-data generative process rather than a noise-to-data one. The paper constructs a tractable Schrödinger bridge whose two endpoints are the low-resolution and high-resolution waveforms, so the reverse sampling trajectory starts from the observed low-resolution signal instead of Gaussian noise. On the VCTK benchmark, this approach reaches a log-spectral distance of 0.848 at 50 steps and 0.911 at 4 steps for 16 kHz-to-48 kHz upsampling, beating the 8-step conditional diffusion baseline's 0.927 under the same 1.7M network. The authors also show that an asymmetric noise schedule, data scaling, and auxiliary spectral losses each contribute to the result.

What carries the argument

The central object is the closed-form tractable Schrödinger bridge between Gaussian endpoints $\mathcal{N}(x_{HR},\epsilon_0^2 I)$ and $\mathcal{N}(x_{LR},\epsilon_T^2 I)$ with $\epsilon_0 \to 0$, which yields Gaussian marginals and the bridge objective $L_{\text{bridge}} = \mathbb{E}[\|x_\theta(x_t,t,x_T)-x_0\|^2]$. This closed form converts the Schrödinger bridge problem into a score-matching-style training loss, and the reverse SDE solvers from [22] perform generation. Three additional ingredients carry much of the empirical gain: an asymmetric noise schedule that keeps the interpolation coefficient at one so low-frequency content persists; a scaling factor $s=1/\sqrt{\operatorname{Var}(x_{LR}-x_{HR})}$ that amplifies the small high-frequency differences; and a fine-tuning loss combining multi-scale STFT magnitude and anti-wrapping phase terms.

What would settle it

One way to settle the central claim is to compute the empirical distribution of intermediate $x_t$ values produced by the actual Chebyshev, Bessel, and STFT low-pass filters and compare it with the Gaussian marginal in Eq. (6) at matching noise levels; a large divergence would show the bridge equations are only an approximation. A second check is to replace the deterministic endpoint $x_T$ with $x_{LR}+\eta$, $\eta \sim \mathcal{N}(0,\epsilon_T^2 I)$, and see whether the 4-step advantage over the diffusion baseline survives.

Watch

Extended reading notes

Core claim

On its own terms, Bridge-SR claims that the instructive content of the low-resolution waveform is best exploited by using that waveform directly as the prior distribution of a tractable Schrödinger bridge. The forward SDE carries the clean waveform toward the degraded one, the reverse SDE recovers the clean waveform, and the training objective predicts the clean waveform $x_0$ from a noisy interpolation $x_t$ and the low-resolution endpoint $x_T$. Because the bridge marginal keeps the low-frequency components constant along the trajectory, unlike a diffusion process that gradually erases them, the model can spend its capacity on high-frequency detail. With this data-to-data trajectory, a 1.7M-parameter network reaches 50-step LSD 0.848 and 4-step LSD 0.911 for 16-to-48 kHz upsampling on VCTK, beating the 8-step conditional diffusion baseline's 0.927 under the same backbone.

Load-bearing premise

The construction assumes both endpoints are Gaussian blurs around the clean and low-resolution waveforms, but the actual inputs are deterministic low-pass-filtered waveforms, and the error from that mismatch is never measured.

Editorial extensions

If this is right

  • With the same 1.7M backbone, Bridge-SR's 4-step synthesis (LSD 0.911) beats the conditional diffusion baseline's 8-step synthesis (LSD 0.927) on 16 kHz-to-48 kHz super-resolution, so the bridge trajectory roughly halves the inference budget at equal or better quality.
  • Because the data-to-data trajectory keeps low-frequency components constant, the model spends its capacity on high-frequency reconstruction, which shows up as larger gains in high-band log-spectral distance than in full-band distance.
  • The scaling factor and auxiliary spectral losses are load-bearing: removing both raises LSD from 0.848 to 0.940, indicating that variance normalization and spectral fine-tuning are part of the recipe, not optional polish.
  • A single trained model covers inputs from 8 kHz to 24 kHz upsampled to 48 kHz, matching the any-rate generality of the diffusion baseline while sampling in fewer steps.
  • The noise schedule matters independently of the bridge formalism: the asymmetric schedule that lets noise variance peak near the end beats symmetric and variance-preserving schedules, suggesting schedule design is what unlocks part of the gain.

Reading between the lines

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

  • The same data-to-data bridge could apply to other speech restoration tasks with an available degraded observation, such as denoising, dereverberation, or codec artifact removal, because the mechanism itself does not depend on super-resolution-specific structure.
  • The unquantified mismatch between deterministic low-pass filtering and the Gaussian-boundary assumption points to a concrete improvement: learning a boundary that matches the true distribution of degraded waveforms could make the closed-form bridge exact rather than approximate.
  • The asymmetric-schedule finding may transfer to image or music super-resolution, where high-frequency detail also carries low energy, so a schedule that spends more steps on the high-frequency phase could help there too.
  • The reported results come from a single clean-speech corpus, so testing on unseen speakers, noisy conditions, or real-world low-rate inputs would show whether the 4-step advantage is general or specific to the benchmark.
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

3 major / 6 minor

Summary. The paper proposes Bridge-SR, a speech super-resolution system that uses a tractable Schrödinger bridge between the low-resolution and high-resolution waveforms, with the low-resolution waveform serving as the prior in a data-to-data reverse SDE. The method uses a lightweight 1.7M-parameter network, an asymmetric noise schedule, a data-scaling factor, and auxiliary STFT-based losses. On VCTK, the authors report improved or competitive LSD and SI-SNR relative to NU-Wave2, AudioSR, NVSR, mdctGAN, and UDM+, and highlight a 4-step result (LSD 0.911) that exceeds NU-Wave2's 8-step result (LSD 0.927) for 16kHz-to-48kHz upsampling.

Significance. If the empirical claims withstand scrutiny, Bridge-SR is a practically useful result: it shows that a bridge-based conditional denoiser can match or beat a strong conditional diffusion baseline with half as many sampling steps, using the same lightweight backbone. The ablations in Table II give credit to the noise schedule, scaling factor, and auxiliary losses, and the baseline comparisons use public implementations and checkpoints. The main significance is therefore an efficiency gain in a practical speech-restoration setting, though the novelty over existing bridge formulations is incremental and the evaluation needs to be more rigorous before the headline claim is fully established.

major comments (3)
  1. [Section III-C, Table II] The central comparison against NU-Wave2 is not protected against test-set schedule selection. Section III-C states that for few-step sampling the authors 'test high-order samplers and use grid-searching algorithm for the choice of inference schedule,' and Table II reports the 4-step result (LSD 0.911) that is compared with NU-Wave2's 8-step default result (0.927). If the grid search was conducted on the VCTK test set, the Bridge-SR number is an optimistically selected statistic rather than the expected performance of a fixed method, while the NU-Wave2 arm uses the default schedule from its public implementation. Please document the train/validation/test split, select the inference schedule on a validation set (and, ideally, grid-search the baseline's schedule as well), and report the resulting fixed-schedule comparison.
  2. [Section III-A, Table II] No statistical uncertainty is reported for any of the numbers in Tables I and II, and the hyperparameters (scaling factor s = 12, noise-schedule endpoints g^2_min and g^2_max, and auxiliary-loss weights λ_mag and λ_phase) are chosen without a documented held-out validation set. Given that the headline improvement over NU-Wave2 at 4 steps is a difference of 0.016 in LSD, it is not possible to assess whether this difference is meaningful without standard deviations, multiple seeds, or a significance test. Please add these, or explicitly report the variance of the metrics across utterances and seeds.
  3. [Section II-A, Eqs. (5)-(7)] The tractable bridge solution in Eqs. (5)-(6) is derived in prior work under Gaussian endpoint distributions p0 = N(x_HR, ε_0^2 I) and p_T = N(x_LR, ε_T^2 I) with ε_0 → 0, whereas Bridge-SR uses deterministic DSP-filtered waveforms as endpoints. The paper does not quantify the mismatch between the exact Schrödinger bridge and the implemented process, even though the training loss in Eq. (7) is exact only under that Gaussian boundary assumption. This does not necessarily invalidate the method, since Eq. (7) can still be read as a conditional denoising objective, but the theoretical framing needs an explicit justification or a statement of the approximation error; otherwise the claim of 'fully exploiting' the Schrödinger bridge structure is stronger than what is demonstrated.
minor comments (6)
  1. [Section IV.A] In the sentence 'outperforming the previous gan-based method [11], conditional diffusion models [6], [15], and unconditional diffusion models [5]', the reference [5] is NU-Wave, not UDM+; UDM+ is [7] and is the unconditional diffusion model actually evaluated in Table I.
  2. [Section III.A / IV] The claim 'any-to-48kHz' is not tested outside the training support: inputs are the four rates 8/12/16/24 kHz, while training samples low-resolution inputs uniformly from 6 to 48 kHz; please clarify whether the system is evaluated on unseen rates (e.g., 6 or 32 kHz) or temper the wording.
  3. [Section III.C] The reported inference schedules have more t-values than sampling steps (e.g., the 2-step schedule lists three times and the 4-step schedule lists three times); please clarify whether these are knots of a continuous schedule or the actual discretization points.
  4. [Table I] The parameter count for mdctGAN is listed as '103.0M*4'; please explain whether this denotes four sub-networks, ensemble members, or something else.
  5. [Section II.C] The auxiliary loss in Eq. (8) is described as optimizing the STFT spectrum 'at each timestep,' but it is applied to the single-step estimate x_θ(x_t,t,x_T) only; please clarify how the loss depends on t.
  6. [Section III.A / Results] No wall-clock inference time or real-time factor is reported; 'efficient' is inferred from step count and parameter count alone. A short runtime table would make the efficiency claim concrete.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline 4-step-vs-8-step superiority is produced by an unprotected grid search over inference schedules, so that specific 'prediction' reduces to schedule selection; the bridge derivation itself is not circular.

  1. fitted input called prediction [Section III-C 'Inference Schedule'; Tables I and II]
    "In few-step sampling, we test high-order samplers and use grid-searching algorithm for the choice of inference schedule. In 4-step sampling, we employ the second-order SDE sampler with t ∈ {8 × 10−2, 5 × 10−1, 1}."

    The paper's central claim that 4-step Bridge-SR beats 8-step NU-Wave2 (LSD 0.911 vs 0.927, Table II) is obtained after grid-searching the inference schedule on the same benchmark for which the final tables are reported. No train/validation/test split is documented for this search. The 4-step LSD is therefore the selected best over the tested schedules, not the expected performance of a fixed inference method. The comparison arm uses NU-Wave2's default 8-step schedule from its public implementation, making the two arms asymmetric. The reported 'prediction' is thus partly forced by the search protocol rather than by the bridge construction.

full rationale

The bridge equations (5)-(6) and the training loss (7) are taken from the authors' prior Bridge-TTS work [22] and from standard Schrodinger-bridge references, but they are used as a conditional denoising objective whose validity does not depend on the Gaussian-boundary derivation being exact: with deterministic endpoints, Eq. (7) is still a valid denoising score-matching loss via the interpolated conditional marginal. The noise-schedule and scaling choices are supported by ablations, not by the defining equations. The only concrete circularity I can exhibit is the inference-schedule grid search in Section III-C: if that search is performed on the VCTK test set (the paper reports no validation split), the headline 4-step result is a selected optimum, not a fixed-method prediction. This makes the empirical comparison partially circular, but the model itself is not defined in terms of the target result, so the paper does not reduce entirely to its inputs.

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

All free parameters are tuned on VCTK and are load-bearing for the reported numbers because the ablation table shows material changes when scaling and auxiliary losses are removed. The Gaussian boundary assumption for the tractable Schrödinger bridge is inherited from reference [22] and is assumed to hold for low-pass-filtered waveforms. No invented entities are introduced.

free parameters (5)
  • data scaling factor s = 12 (VCTK-trained constant)
    Formula s = 1/sqrt(Var(xLR - xHR)) is dataset-dependent; fixed at 12 in experiments, and the ablation shows it is important.
  • noise schedule endpoints = g^2_min = 8e-7, g^2_max = 8e-2
    Selected empirically for SR on VCTK; changes how many sampling steps are spent on high-frequency content.
  • noise schedule asymmetry = beta0 -> 0 with t_p/T ~ 1/sqrt(2)
    The asymmetric gmax schedule is justified by ablation: gmax 0.848 vs gconst 0.869 vs SVP 0.900 at 50 steps.
  • auxiliary loss weights = lambda_mag = 4e-6, lambda_phase = 5e-6
    Tuned in the fine-tuning stage; removing auxiliary losses degrades LSD from 0.848 to 0.889 in Table II.
  • few-step inference schedules = 4-step t in {8e-2, 5e-1, 1}; 2-step t in {3e-2, 9e-1, 1}; 1-step t in {4e-2, 1}
    Section III-C states a grid-searching algorithm was used for the choice of inference schedule.
assumptions (3)
  • domain assumption Gaussian endpoint distributions p0 = N(xHR, eps0^2 I) and pT = N(xLR, epsT^2 I) with eps0 -> 0 are sufficient for the tractable closed-form SB solution in Eq. (5)-(6).
    The paper takes this boundary condition from [22] and applies it to real filter-degraded waveforms, without checking that low-pass-filtered speech satisfies the Gaussian endpoint model.
  • domain assumption The low-resolution observation xLR is an informative prior on the same manifold as xHR, so interpolating between xLR and xHR is a valid generative path for SR.
    The whole data-to-data design assumes the LR waveform's content, such as timing, pitch, and coarse spectrum, is correct and only high-frequency detail is missing.
  • standard math The bridge loss Lbridge in Eq. (7) is a valid score-matching objective for the marginal pt defined by the tractable SB.
    This follows from standard score-based generative modeling theory when the marginal is Gaussian as in Eq. (6); the paper does not re-derive it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridge-SR: Schr\"odinger Bridge for Efficient SR." pith.science (2026). https://pith.science/paper/JAVZ35SF

@misc{pith2026250107897,
  author       = {Pith},
  title        = {Pith review of: Bridge-SR: Schr\"odinger Bridge for Efficient SR},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JAVZ35SF}},
  note         = {Machine review of arXiv:2501.07897}
}
read the original abstract

Speech super-resolution (SR), which generates a waveform at a higher sampling rate from its low-resolution version, is a long-standing critical task in speech restoration. Previous works have explored speech SR in different data spaces, but these methods either require additional compression networks or exhibit limited synthesis quality and inference speed. Motivated by recent advances in probabilistic generative models, we present Bridge-SR, a novel and efficient any-to-48kHz SR system in the speech waveform domain. Using tractable Schr\"odinger Bridge models, we leverage the observed low-resolution waveform as a prior, which is intrinsically informative for the high-resolution target. By optimizing a lightweight network to learn the score functions from the prior to the target, we achieve efficient waveform SR through a data-to-data generation process that fully exploits the instructive content contained in the low-resolution observation. Furthermore, we identify the importance of the noise schedule, data scaling, and auxiliary loss functions, which further improve the SR quality of bridge-based systems. The experiments conducted on the benchmark dataset VCTK demonstrate the efficiency of our system: (1) in terms of sample quality, Bridge-SR outperforms several strong baseline methods under different SR settings, using a lightweight network backbone (1.7M); (2) in terms of inference speed, our 4-step synthesis achieves better performance than the 8-step conditional diffusion counterpart (LSD: 0.911 vs 0.927). Demo at https://bridge-sr.github.io.

Figures

Figures reproduced from arXiv: 2501.07897 by the authors.

Figure 1
Figure 1. Overview of Bridge-SR. As shown in the upper part, the forward process of the Schr [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. We show the means of the intermediate representations for the [PITH_FULL_IMAGE:figures/full_fig_p003_2.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. A2SB: Audio-to-Audio Schrodinger Bridges

    cs.SD 2025-01 conditional novelty 6.0 of 10

    A2SB applies Schrödinger bridges to music restoration, achieving state-of-the-art bandwidth extension and inpainting at 44.1kHz in a single vocoder-free model.

Reference graph

Works this paper leans on

33 extracted references · 14 canonical work pages · cited by 1 Pith paper

  1. [22]

    Schrodinger bridges beat diffusion models on text-to-speech synthesis,

    Z. Chen, G. He, K. Zheng, X. Tan, and J. Zhu, “Schrodinger bridges beat diffusion models on text-to-speech synthesis,” arXiv preprint arXiv:2312.03491, 2023

    arXiv 2023

  2. [1]

    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. Li et al., “Neural codec language models are zero-shot text to speech synthesizers,” arXiv preprint arXiv:2301.02111 , 2023

    arXiv 2023

  3. [2]

    Audioldm: text-to-audio generation with latent diffusion models,

    H. Liu, Z. Chen, Y . Yuan, X. Mei, X. Liu, D. Mandic, W. Wang, and M. D. Plumbley, “Audioldm: text-to-audio generation with latent diffusion models,” in Proceedings of the 40th International Conference on Machine Learning , 2023, pp. 21 450–21 474

    work page 2023

  4. [3]

    Quality-aware masked diffusion transformer for enhanced music generation,

    C. Li, R. Wang, L. Liu, J. Du, Y . Sun, Z. Guo, Z. Zhang, and Y . Jiang, “Quality-aware masked diffusion transformer for enhanced music generation,” arXiv preprint arXiv:2405.15863 , 2024

    arXiv 2024

  5. [4]

    Wsrglow: A glow-based waveform generative model for audio super-resolution,

    K. Zhang, Y . Ren, C. Xu, and Z. Zhao, “Wsrglow: A glow-based waveform generative model for audio super-resolution,” arXiv preprint arXiv:2106.08507, 2021

    arXiv 2021

  6. [5]

    Nu-wave: A diffusion probabilistic model for neural audio upsampling,

    J. Lee and S. Han, “Nu-wave: A diffusion probabilistic model for neural audio upsampling,” arXiv preprint arXiv:2104.02321 , 2021

    arXiv 2021

  7. [6]

    Nu-wave 2: A general neural audio upsampling model for various sampling rates,

    S. Han and J. Lee, “Nu-wave 2: A general neural audio upsampling model for various sampling rates,” arXiv preprint arXiv:2206.08545 , 2022

    arXiv 2022

  8. [7]

    Conditioning and sampling in variational diffusion models for speech super-resolution,

    C.-Y . Yu, S.-L. Yeh, G. Fazekas, and H. Tang, “Conditioning and sampling in variational diffusion models for speech super-resolution,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2023, pp. 1–5

    work page 2023

Show all 33 references
  1. [8]

    Musichifi: Fast high- fidelity stereo vocoding,

    G. Zhu, J.-P. Caceres, Z. Duan, and N. J. Bryan, “Musichifi: Fast high- fidelity stereo vocoding,” arXiv preprint arXiv:2403.10493 , 2024

    2024 arXiv

  2. [9]

    Miipher: A robust speech restoration model integrating self-supervised speech and text representations,

    Y . Koizumi, H. Zen, S. Karita, Y . Ding, K. Yatabe, N. Morioka, Y . Zhang, W. Han, A. Bapna, and M. Bacchiani, “Miipher: A robust speech restoration model integrating self-supervised speech and text representations,” in 2023 IEEE Workshop on Applications of Signal Processing ...

    2023

  3. [10]

    Generative speech foundation model pretraining for high-quality speech extraction and restoration,

    P.-J. Ku, A. H. Liu, R. Korostik, S.-F. Huang, S.-W. Fu, and A. Juki ´c, “Generative speech foundation model pretraining for high-quality speech extraction and restoration,” arXiv preprint arXiv:2409.16117 , 2024

    2024 arXiv

  4. [11]

    mdctgan: Taming transformer- based gan for speech super-resolution with modified dct spectra,

    C. Shuai, C. Shi, L. Gan, and H. Liu, “mdctgan: Taming transformer- based gan for speech super-resolution with modified dct spectra,” arXiv preprint arXiv:2305.11104, 2023

    2023 arXiv

  5. [12]

    Aero: Audio super resolution in the spectral domain,

    M. Mandel, O. Tal, and Y . Adi, “Aero: Audio super resolution in the spectral domain,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2023, pp. 1–5

    2023

  6. [13]

    Solving audio inverse prob- lems with a diffusion model,

    E. Moliner, J. Lehtinen, and V . V ¨alim¨aki, “Solving audio inverse prob- lems with a diffusion model,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2023, pp. 1–5

    2023

  7. [14]

    V oicefixer: Toward general speech restoration with neural vocoder,

    H. Liu, Q. Kong, Q. Tian, Y . Zhao, D. Wang, C. Huang, and Y . Wang, “V oicefixer: Toward general speech restoration with neural vocoder,” arXiv preprint arXiv:2109.13731 , 2021

    2021 arXiv

  8. [15]

    Audiosr: Versatile audio super-resolution at scale,

    H. Liu, K. Chen, Q. Tian, W. Wang, and M. D. Plumbley, “Audiosr: Versatile audio super-resolution at scale,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024, pp. 1076–1080

    2024

  9. [16]

    Audit: Audio editing by following instructions with latent diffusion models,

    Y . Wang, Z. Ju, X. Tan, L. He, Z. Wu, J. Bian et al. , “Audit: Audio editing by following instructions with latent diffusion models,”Advances in Neural Information Processing Systems , vol. 36, pp. 71 340–71 357, 2023

    2023

  10. [17]

    Audio super- resolution with robust speech representation learning of masked au- toencoder,

    S.-B. Kim, S.-H. Lee, H.-Y . Choi, and S.-W. Lee, “Audio super- resolution with robust speech representation learning of masked au- toencoder,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2024

    2024

  11. [18]

    Specmaskgit: Masked generative modelling of audio spectrogram for efficient audio synthesis and beyond

    M. Comunita, Z. Zhong, A. Takahashi, S. Yang, M. Zhao, K. Saito, Y . Ikemiya, T. Shibuya, S. Takahashi, and Y . Mitsufuji, “Specmaskgit: Masked generative modelling of audio spectrogram for efficient audio synthesis and beyond.”

  12. [19]

    Neural vocoder is all you need for speech super-resolution,

    H. Liu, W. Choi, X. Liu, Q. Kong, Q. Tian, and D. Wang, “Neural vocoder is all you need for speech super-resolution,” arXiv preprint arXiv:2203.14941, 2022

    2022 arXiv

  13. [20]

    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

  14. [21]

    Dif- fwave: A versatile diffusion model for audio synthesis,

    Z. Kong, W. Ping, J. Huang, K. Zhao, and B. Catanzaro, “Dif- fwave: A versatile diffusion model for audio synthesis,” arXiv preprint arXiv:2009.09761, 2020

    2009 arXiv

  15. [23]

    I2sb: Image-to-image schr ¨odinger bridge,

    G.-H. Liu, A. Vahdat, D.-A. Huang, E. A. Theodorou, W. Nie, and A. Anandkumar, “I2sb: Image-to-image schr ¨odinger bridge,” arXiv preprint arXiv:2302.05872, 2023

    2023 arXiv

  16. [24]

    Cstr vctk corpus: English multi-speaker corpus for cstr voice cloning toolkit (version 0.92),

    J. Yamagishi, C. Veaux, K. MacDonald et al., “Cstr vctk corpus: English multi-speaker corpus for cstr voice cloning toolkit (version 0.92),” University of Edinburgh. The Centre for Speech Technology Research (CSTR), pp. 271–350, 2019

    2019

  17. [25]

    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. Kavukcuoglu et al. , “Wavenet: A generative model for raw audio,” arXiv preprint arXiv:1609.03499, vol. 12, 2016

    2016 arXiv

  18. [26]

    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,” arXiv preprint arXiv:2011.13456 , 2020

    2011 arXiv

  19. [27]

    Sur la th´eorie relativiste de l’´electron et l’interpr´etation de la m ´ecanique quantique,

    E. Schr ¨odinger, “Sur la th´eorie relativiste de l’´electron et l’interpr´etation de la m ´ecanique quantique,” in Annales de l’institut Henri Poincar ´e, vol. 2, no. 4, 1932, pp. 269–310

    1932

  20. [28]

    Deep generative learning via schr¨odinger bridge,

    G. Wang, Y . Jiao, Q. Xu, Y . Wang, and C. Yang, “Deep generative learning via schr¨odinger bridge,” in International conference on machine learning. PMLR, 2021, pp. 10 794–10 804

    2021

  21. [29]

    Likelihood training of schr¨odinger bridge using forward-backward sdes theory,

    T. Chen, G.-H. Liu, and E. A. Theodorou, “Likelihood training of schr¨odinger bridge using forward-backward sdes theory,” arXiv preprint arXiv:2110.11291, 2021

    2021 arXiv

  22. [30]

    Schr ¨odinger bridge for generative speech enhancement,

    A. Juki ´c, R. Korostik, J. Balam, and B. Ginsburg, “Schr ¨odinger bridge for generative speech enhancement,” arXiv preprint arXiv:2407.16074 , 2024

    2024 arXiv

  23. [31]

    Diffusion-based speech enhancement with schr ¨odinger bridge and symmetric noise schedule,

    S. Wang, S. Liu, A. Harper, P. Kendrick, M. Salzmann, and M. Cer- nak, “Diffusion-based speech enhancement with schr ¨odinger bridge and symmetric noise schedule,” arXiv preprint arXiv:2409.05116 , 2024

    2024 arXiv

  24. [32]

    Long-form music generation with latent diffusion,

    Z. Evans, J. D. Parker, C. Carr, Z. Zukowski, J. Taylor, and J. Pons, “Long-form music generation with latent diffusion,” arXiv preprint arXiv:2404.10301, 2024

    2024 arXiv

  25. [33]

    Neural speech phase prediction based on parallel estimation architecture and anti-wrapping losses,

    Y . Ai and Z.-H. Ling, “Neural speech phase prediction based on parallel estimation architecture and anti-wrapping losses,” in ICASSP 2023- 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023, pp. 1–5

    2023

Pith tools

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