Pith. sign in

REVIEW 4 major objections 4 minor 43 references

ITO-Master: Inference-Time Optimization for Audio Effects Modeling of Music Mastering Processors

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Optimizing the reference embedding at inference time improves reference-based mastering style transfer and gives users finer control.

desk verdict A credible new application of inference-time optimization to mastering style transfer, but the headline AF gains are partly circular because the same loss drives optimization and evaluation; still worth a careful peer review. read the letter →

arxiv 2506.16889 v3 pith:PKVPK6EX submitted 2025-06-20 cs.SD eess.AS

Junghyun Koo , Marco A. Martínez-Ramírez , Wei-Hsiang Liao , Giorgio Fabbro , Michele Mancusi , Yuki Mitsufuji This is my paper · ORCID
classification cs.SDeess.AS
keywords musicmasteringstyletransferinference-timeoptimizationreferenceembeddingdifferentiableaudioeffectsfeaturelosstext-guidedprocessing
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 is trying to establish that a music mastering style-transfer system can be made more controllable and more accurate by optimizing the reference embedding at inference time, after training is complete. The proposed framework, ITO-Master, keeps the trained style converter fixed and adjusts a 2048-dimensional style vector through gradient descent on an audio-feature loss that is meant to be independent of the musical content. If the claim holds, users of an automatic mastering system would be able to fine-tune a transfer toward the target style, or toward a text description of a genre, without retraining the model or hand-adjusting dozens of effect parameters. The paper reports that this inference-time optimization lowers the audio-feature loss and dynamic-range variability, and that expert listeners rate the ITO outputs as closer to the reference mastering style, though the naturalness metrics it reports worsen after optimization.

What carries the argument

The mechanism that carries the argument is Inference-Time Optimization on the reference embedding: at inference, the style vector $z_{\mathrm{ref}}$ is treated as a free variable and updated by gradient descent under an auxiliary loss $\mathcal{L}_{\mathrm{aux}}$ while the pre-trained converter is frozen. The auxiliary loss is the Audio Feature (AF) loss, a weighted combination of audio-feature transformations intended to capture dynamics, spatialization, and spectral character without being tied to the musical content. On the white-box side, the differentiable mastering chain—six effect modules ordered as EQ, distortion, three-band compressor, makeup gain, stereo imager, and limiter, with differentiable all-pole filters in the compressor and limiter—is what makes the parameters interpretable and lets the effect of the optimized embedding be observed.

What would settle it

A decisive check would be to run ITO for a fixed input and reference over a grid of optimization steps and, with expert listeners in a blind test, determine whether the step count that minimizes the AF loss is also the one judged closest to the reference; if AF loss keeps decreasing while perceptual similarity does not increase, the auxiliary objective is not tracking the claimed mastering style.

Watch

Extended reading notes

Core claim

The central claim is that the reference embedding $z_{\mathrm{ref}}$—the representation of the reference track that conditions the mastering converter—is the right place to apply inference-time optimization for style transfer. In both a black-box waveform converter and a white-box differentiable mastering chain, updating $z_{\mathrm{ref}}$ by backpropagating the Audio Feature loss for up to 100 steps improves the objective metrics: the black-box model's AF loss falls from 0.125 to 0.099 and dynamic-range variability from 0.577 to 0.567, while the white-box model's AF loss falls from 0.186 to 0.139 and dynamic-range variability from 0.521 to 0.474. Cosine similarity between the output and reference FX embeddings stays roughly constant, and listening-test scores improve with ITO. The same optimization applied to the white-box chain's 46 parameters instead of the embedding performs worse on every metric, suggesting that the embedding is a more effective optimization target than the processor parameters when the loss is content-independent.

Load-bearing premise

The load-bearing premise is that the Audio Feature loss measures mastering style independently of musical content, so that minimizing it at inference time moves the output toward the reference's mastering style rather than toward some unrelated artifact.

Editorial extensions

If this is right

  • If ITO-Master is right, mastering style transfer can be refined after the initial inference, so a single trained model can produce many variants of one transfer from the same reference simply by choosing how long to optimize the embedding.
  • Optimizing the reference embedding with a content-independent loss is more efficient and more effective than optimizing the whole differentiable effect chain, which the paper's experiments show degrades all metrics.
  • Training the reference encoder together with the converter is a necessary ingredient: with a fixed pretrained encoder the model performs worse, while joint training improves both AF loss and dynamic-range variability.
  • The ITO objective is not tied to mastering features; the paper's text-prompt experiment shows the same embedding optimization can steer the output toward genre descriptions, so the mechanism acts as a flexible creative control.
  • Because the reported naturalness metrics worsen after ITO, the number of optimization steps is a practical tuning knob, and the style/naturalness trade-off must be managed in deployment.

Reading between the lines

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

  • Because ITO touches only the reference embedding, the same frozen converter could be steered by other differentiable objectives, such as a target loudness, a target stereo width, or a perceptual quality penalty, without any retraining; the paper demonstrates only an audio-feature objective and a text-audio objective.
  • The reported naturalness degradation after ITO suggests a multi-objective variant that adds a quality penalty to the auxiliary loss might recover the style gain while protecting audio quality, a combination the paper does not test.
  • The text-prompt steering experiment indicates the method is a generic steering mechanism for differentiable effect chains, so it should transfer to other audio effects such as reverb or EQ matching, even though the paper only studies mastering processors.
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 / 4 minor

Summary. The paper proposes ITO-Master, an inference-time optimization framework for reference-based music mastering style transfer. It trains a Master Style Converter Ψ in self-supervised fashion with random FX manipulations, and at inference time optimizes the reference embedding zref using an auxiliary content-independent Audio Feature (AF) loss while keeping Ψ fixed. The converter is instantiated in both black-box and white-box forms, the latter using a differentiable six-module mastering chain. Experiments on MTG-Jamendo compare against feature-matching and E2E baselines, reporting AF loss, DRV, cosine similarity, and FAD, plus a 10-participant MUSHRA-type listening test and a qualitative text-prompt study. The central claim is that ITO improves mastering style similarity and provides user-controllable refinement without retraining.

Significance. If the central claim were fully supported, ITO-Master would be a useful contribution to the audio-effects style transfer literature: it introduces a clean inference-time mechanism for user control, systematically compares black-box and white-box converter designs, and ships an open-source differentiable mastering chain with a demo. The qualitative text-conditioned ITO using CLAP embeddings is a promising creative tool. However, the primary objective evidence is weakened by the fact that the AF loss is both the ITO objective and the headline metric, by the absence of error bars or significance tests in Table 1, and by the generally worsening FAD scores. The DRV improvements and the subjective listening test provide independent support, but as reported they are not sufficient to fully establish the paper's claim that ITO 'improves mastering performance across different styles.'

major comments (4)
  1. [§3.4 and §4.3, Table 1] The AF loss is used both as the optimization objective for ITO (Section 3.4) and as the primary objective evaluation metric (Section 4.3, Table 1). Consequently, the reported AF improvements after ITO (e.g., Black-box + trainPhi: 0.125 to 0.099; White-box + trainPhi: 0.186 to 0.139) are expected by construction for any gradient-based optimizer and cannot serve as independent evidence that ITO improves mastering style transfer. The paper should either report AF loss only as a sanity check of the optimization and rely on non-optimized metrics for the central claim, or provide an independent, held-out style metric that is not part of the ITO objective.
  2. [Table 1 and §5.1] All objective results in Table 1 are reported as point estimates with no confidence intervals, error bars, or significance tests. Several ITO-vs-no-ITO differences are small (e.g., DRV 0.577 vs. 0.567; cos sim 0.945 vs. 0.946), and without uncertainty estimates it is impossible to tell whether the improvements are reliable. The authors should report repeated-run or bootstrap intervals and perform significance tests for each metric and condition, especially for DRV and cosine similarity.
  3. [§5.2, Subjective Evaluation] The listening test is described as showing 'significant improvements (pairwise t-test, p<0.05)', but the manuscript does not state which pairwise comparisons were tested, whether multiple-comparison correction was applied, or what the effect sizes and per-condition means/confidence intervals were. Moreover, the MUSHRA-type design lacks a hidden reference and a high anchor, making the absolute similarity scores difficult to interpret. The authors should report the full per-condition statistics, specify the exact comparisons, and discuss the impact of the missing anchors on the validity of the subjective claim.
  4. [§3.4, Figure 1(b), §5.1] The paper optimizes the raw 2048-dimensional reference embedding zref with no regularization or manifold constraint, yet claims this enables 'micro-level adjustments'. The FAD deterioration observed in most rows (e.g., Black-box + trainPhi FAD_CLAP 159.8 to 182.2; White-box + trainPhi FAD_CLAP 93.2 to 105.2) suggests that the optimized embedding may drift outside the distribution the converter was trained on. The authors should analyze embedding drift (e.g., distance from the original zref, distribution statistics before/after ITO) or add a regularization term, to support the claim that ITO makes only micro-level changes rather than arbitrary perturbations.
minor comments (4)
  1. [Title] The title in the submitted text contains an odd line-break artifact: 'INFERENCE-TIME OPTIMIZA TION' should read 'INFERENCE-TIME OPTIMIZATION'.
  2. [§4.1] The evaluation setup says 200 songs are selected from MTG-Jamendo, with 100 as xin and 100 as xref, but it is not stated whether the two sets are disjoint and how the xin/xref pairings are formed. Please clarify to avoid potential content or genre leakage.
  3. [Abstract and §5.1] The abstract states that ITO 'improves mastering performance', but the FAD trade-off is acknowledged only in Section 5.1 prose and is omitted from the abstract. The authors should qualify the abstract claim to reflect the naturalness cost.
  4. [§5.3] The qualitative text-prompt analysis is based on a single instrumental track and one run per prompt; this is fine as a qualitative demonstration, but the manuscript should explicitly label it as anecdotal rather than as evidence of general text-guided control.

Circularity Check

1 steps flagged · score 6.0 of 10

AF-loss improvement after ITO is partly circular: the same AF loss is both the ITO objective (§3.4) and the headline metric (§4.3), while FAD mostly worsens, leaving the central claim partially dependent on the optimized objective.

  1. fitted input called prediction [Section 3.4 (ITO objective) and Section 4.3 / Table 1 (evaluation metric)]
    "For ITO, the Audio Feature (AF) loss proposed by [13] is utilized as the auxiliary objective function Laux... We optimize zref iteratively using gradient descent: z(t+1)ref = z(t)ref − η∇zLaux(Ψ(xin,z(t)ref),xref)... Audio Feature (AF) Loss: As discussed in Section 3.4, AF Loss measures how well the output y′ matches the desired audio features."

    The identical functional is minimized at inference (Section 3.4) and then reported as the headline metric (Table 1, AF (↓)). Gradient descent on zref directly decreases Laux = AF loss on the evaluated samples, so the direction of the reported AF improvements (0.125→0.099 for black-box; 0.186→0.139 for white-box) is guaranteed by construction; it is an optimization trace, not an independent prediction. The independent metrics do not fully offset this: DRV improves only slightly, cos sim is nearly flat, and FAD_CLAP worsens in both ITO rows (159.8→182.2; 93.2→105.2). Only the MUSHRA listening test and DRV are genuinely independent evidence for the claim that ITO improves mastering style similarity.

full rationale

The central derivation chain in this paper is empirical, not mathematical, and most of it is self-contained. The pretrained FXencoder from [12] (same research group) is used as the reference encoder and as the cos sim metric, but the paper explicitly tests both fixed and co-trained Φ, so that self-citation is not load-bearing. The AF loss is cited to external Diff-MST work [13], not to the authors. The concrete circular step is that ITO optimizes the AF loss on zref at inference and the same AF loss is the first row of Table 1 and the primary objective result; any gradient-based optimizer will reduce its own objective on the same inputs, so the AF drop is expected rather than evidence of perceptual mastering improvement. DRV does improve without being optimized, and the MUSHRA test provides independent subjective support, so the paper is not wholly circular. However, Table 1 reports point estimates without confidence intervals, FAD mostly worsens after ITO, and the subjective ITO-vs-no-ITO contrast is reported only as an aggregate p<0.05 without effect sizes. Weighing these, the headline AF result is partially circular, but independent evidence keeps the overall circularity at a moderate level: score 6.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on three domain assumptions: consistent mastering style within a song, the validity of the AF loss as a content-independent style measure, and the faithfulness of the differentiable mastering chain. The ITO on zref also assumes the embedding can be optimized without leaving the data distribution. No new physical entities are introduced.

free parameters (2)
  • FX module application probabilities = [0.9, 0.3, 0.8, 0.85, 0.6, 1.0]
    Hand-set in Section 3.3 to balance variability and realism in the random Fx manipulation used to train the converter. These probabilities shape the training distribution but are not the central claim.
  • ITO learning rate and max steps = eta=2e-4, max 100 steps with early stopping
    Chosen in Section 4.2 to balance convergence and quality. The FAD trade-off discussion notes the step count must be calibrated, making this a manually selected hyperparameter that affects the reported improvements.
assumptions (4)
  • domain assumption A single song maintains a consistent mastering style throughout its duration.
    Section 3.1 uses this to treat segments A and B from the same song as sharing the same mastering style, which is the basis for self-supervised training pairs.
  • domain assumption The AF loss is a valid content-independent measure of mastering style covering dynamics, spatialization, and spectral characteristics.
    Section 3.4 chooses AF loss as the ITO objective and Section 4.3 uses it as the primary metric. The whole optimization and the main objective evaluation rest on this assumption.
  • domain assumption The differentiable mastering chain (EQ, distortion, 3-band compressor, makeup gain, stereo imager, limiter) sufficiently approximates real mastering processors.
    Section 3.3 claims the chain models the complexities of real-world mastering. The white-box results depend on this faithfulness.
  • ad hoc to paper Gradient-based optimization of the 2048-dimensional reference embedding stays on a meaningful manifold and does not break content.
    The paper does not regularize zref during ITO (Section 3.4), yet assumes smooth adaptation without content degradation. No explicit constraint or prior is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ITO-Master: Inference-Time Optimization for Audio Effects Modeling of Music Mastering Processors." pith.science (2026). https://pith.science/paper/PKVPK6EX

@misc{pith2026250616889,
  author       = {Pith},
  title        = {Pith review of: ITO-Master: Inference-Time Optimization for Audio Effects Modeling of Music Mastering Processors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PKVPK6EX}},
  note         = {Machine review of arXiv:2506.16889}
}
read the original abstract

Music mastering style transfer aims to model and apply the mastering characteristics of a reference track to a target track, simulating the professional mastering process. However, existing methods apply fixed processing based on a reference track, limiting users' ability to fine-tune the results to match their artistic intent. In this paper, we introduce the ITO-Master framework, a reference-based mastering style transfer system that integrates Inference-Time Optimization (ITO) to enable finer user control over the mastering process. By optimizing the reference embedding during inference, our approach allows users to refine the output dynamically, making micro-level adjustments to achieve more precise mastering results. We explore both black-box and white-box methods for modeling mastering processors and demonstrate that ITO improves mastering performance across different styles. Through objective evaluation, subjective listening tests, and qualitative analysis using text-based conditioning with CLAP embeddings, we validate that ITO enhances mastering style similarity while offering increased adaptability. Our framework provides an effective and user-controllable solution for mastering style transfer, allowing users to refine their results beyond the initial style transfer.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 39 canonical work pages

  1. [1]

    ITO-Master: Inference-Time Optimization for Audio Effects Modeling of Music Mastering Processors

    INTRODUCTION Music mastering is the final step in the audio production process, ensuring professional sound quality and consistent playback across music distribution platforms. This process involves applying a series of audio effects such as equal- ization, compression, stereo imaging, and limiting, which collectively shape the sonic characteristics and e...

    work page Pith review arXiv 2025

  2. [2]

    Recent advancements in deep learning have led to more sophisticated approaches, where neural networks are used to learn complex mappings between input and output audio signals

    RELA TED WORKS 2.1 Audio Effects Style Transfer Audio effects style transfer has become a significant area of research in automating and enhancing music production. Recent advancements in deep learning have led to more sophisticated approaches, where neural networks are used to learn complex mappings between input and output audio signals. These methods h...

    work page

  3. [3]

    The training pipeline simulates real-world master- ing scenarios by applying random Fx manipulations

    METHODOLOGY In this section, we describe the components of the proposed mastering style transfer framework: the training pipeline, Mastering Style Converter, differentiable mastering chain, and ITO. The training pipeline simulates real-world master- ing scenarios by applying random Fx manipulations. The Mastering Style Converter Ψ transfers the mastering ...

    work page

  4. [4]

    3-band compressor, 4

    distortion, 3. 3-band compressor, 4. makeup gain, 5. stereo imager, and 6. limiter. The order of these modules is fixed, with the probability of applying each Fx module for random manipulation during training set at 90%, 30%, 80%, 85%, 60%, and 100%, respectively. These probabilities are adopted to introduce greater variability while preventing the synthe...

    work page

  5. [5]

    Mixture samples from these datasets are employed, as they are not fully mastered, allowing random Fx manipulation withf to create synthetic mastered samples

    EXPERIMENTS 4.1 Dataset We utilized MoisesDB dataset [23] for training, and vali- dated using the MUSDB18 validation subset [24]. Mixture samples from these datasets are employed, as they are not fully mastered, allowing random Fx manipulation withf to create synthetic mastered samples. For Fx-Normalization, mean statistics are precomputed on the MoisesDB...

    work page

  6. [6]

    Classic Music

    RESULTS 5.1 Objective Evaluation The performance of the proposed methods, along with the baseline approaches, is summarized in Table 1. The fea- ture matching methods, specifically Fx-Normalization and Matchering, demonstrate strong performance on the AF and FAD metrics. This is expected, as these approaches directly apply Fx-related transformations to ma...

    work page

  7. [7]

    Our experiments showed that training the reference encoder Φ alongside Ψ improves performance

    CONCLUSION In this paper, we introduced the ITO-Master framework, which leverages ITO on zref for mastering style transfer. Our experiments showed that training the reference encoder Φ alongside Ψ improves performance. Optimizingzref with ITO led to meaningful improvements with few steps, outper- forming direct optimization of Θ in efficiency. Subjective ...

    work page

  8. [8]

    Zölzer, X

    U. Zölzer, X. Amatriain, D. Arfib, J. Bonada, G. De Poli, P. Dutilleux, G. Evangelista, F. Keiler, A. Loscos, D. Rocchesso et al., DAFX-Digital audio effects. John Wiley & Sons, 2002

    work page 2002

Show all 43 references
  1. [9]

    Shelvock, Audio mastering as musical practice

    M. Shelvock, Audio mastering as musical practice . The University of Western Ontario (Canada), 2012

    2012

  2. [10]

    A study on audio signal processed by

    M. Piotrowska, S. Piotrowski, and B. Kostek, “A study on audio signal processed by" instant mastering" ser- vices,” in Audio Engineering Society Convention 142 . Audio Engineering Society, 2017

    2017

  3. [11]

    Machine learning in con- text, or learning from landr: Artificial intelligence and the platformization of music mastering,

    J. Sterne and E. Razlogova, “Machine learning in con- text, or learning from landr: Artificial intelligence and the platformization of music mastering,” Social Media+ Society, vol. 5, no. 2, p. 2056305119847525, 2019

    2019

  4. [12]

    Differentiable signal processing with black- box audio effects,

    M. A. M. Ramírez, O. Wang, P. Smaragdis, and N. J. Bryan, “Differentiable signal processing with black- box audio effects,” in ICASSP 2021-2021 IEEE Inter- national Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2021, pp. 66–70

    2021

  5. [13]

    End-to-end music re- mastering system using self-supervised and adversarial training,

    J. Koo, S. Paik, and K. Lee, “End-to-end music re- mastering system using self-supervised and adversarial training,” in Proc. ICASSP, 2022, pp. 4608–4612

    2022

  6. [14]

    Matchering: Audio matching and mastering python library,

    S. Grishakov, C.-Y . Yu, and Zicklag, “Matchering: Audio matching and mastering python library,” https: //github.com/sergree/matchering

  7. [15]

    DDSP: Differen- tiable digital signal processing,

    J. Engel, C. Gu, A. Roberts et al., “DDSP: Differen- tiable digital signal processing,” in International Con- ference on Learning Representations, 2020

    2020

  8. [16]

    Reverb conversion of mixed vocal tracks using an end-to-end convolutional deep neural network,

    J. Koo, S. Paik, and K. Lee, “Reverb conversion of mixed vocal tracks using an end-to-end convolutional deep neural network,” in ICASSP 2021-2021 IEEE In- ternational Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2021, pp. 81–85

    2021

  9. [17]

    Blind estima- tion of audio processing graph,

    S. Lee, J. Park, S. Paik, and K. Lee, “Blind estima- tion of audio processing graph,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2023, pp. 1–5

    2023

  10. [18]

    Style transfer of audio effects with differentiable signal pro- cessing,

    C. J. Steinmetz, N. J. Bryan, and J. D. Reiss, “Style transfer of audio effects with differentiable signal pro- cessing,” J. Audio Eng. Soc, vol. 70, no. 9, pp. 708–721, 2022

    2022

  11. [19]

    Music mixing style transfer: A contrastive learning approach to disentangle audio effects,

    J. Koo, M. A. Martínez-Ramírez, W.-H. Liao, S. Uhlich, K. Lee, and Y . Mitsufuji, “Music mixing style transfer: A contrastive learning approach to disentangle audio effects,” in ICASSP 2023-2023 IEEE International Con- ference on Acoustics, Speech and Signal Processing (ICASSP...

    2023

  12. [20]

    Diff-MST: Differentiable mixing style transfer,

    S. S. Vanka, C. Steinmetz, J.-B. Rolland, J. Reiss, and G. Fazekas, “Diff-MST: Differentiable mixing style transfer,” in Proc. ISMIR, 2024

    2024

  13. [21]

    Towards zero-shot amplifier modeling: One-to-many amplifier modeling via tone embedding control,

    Y .-H. Chen, Y .-T. Yeh, Y .-C. Cheng, J.-T. Wu, Y .-H. Ho, J.-S. R. Jang, and Y .-H. Yang, “Towards zero-shot amplifier modeling: One-to-many amplifier modeling via tone embedding control,” in Proc. ISMIR, 2024

    2024

  14. [22]

    DITTO: Diffusion inference-time t- optimization for music generation,

    Z. Novack, J. McAuley, T. Berg-Kirkpatrick, and N. J. Bryan, “DITTO: Diffusion inference-time t- optimization for music generation,” in Proc. ICML , 2024

    2024

  15. [23]

    DITTO-2: Distilled diffusion inference-time t-optimization for music generation,

    Z. Novack, J. McAuley, T. Berg-Kirkpatrick, and N. Bryan, “DITTO-2: Distilled diffusion inference-time t-optimization for music generation,” in Proc. ISMIR, 2024

    2024

  16. [24]

    ST-ITO: Controlling audio effects for style transfer with inference-time optimization,

    C. Steinmetz, S. Singh, I. Ibnyahya, S. Yuan, E. Benetos, J. Reiss et al., “ST-ITO: Controlling audio effects for style transfer with inference-time optimization,” inProc. ISMIR, 2024

    2024

  17. [25]

    Improving inference- time optimisation for vocal effects style transfer with a gaussian prior,

    C.-Y . Yu, M. A. Martínez-Ramírez, J. Koo, W.-H. Liao, Y . Mitsufuji, and G. Fazekas, “Improving inference- time optimisation for vocal effects style transfer with a gaussian prior,” arXiv preprint arXiv:2505.11315, 2025

    2025

  18. [26]

    Automatic music mixing with deep learning and out-of-domain data,

    M. A. Martínez-Ramírez, W.-H. Liao, G. Fabbro, S. Uh- lich, C. Nagashima, and Y . Mitsufuji, “Automatic music mixing with deep learning and out-of-domain data,” in Proc. ISMIR, 2022

    2022

  19. [27]

    Active crossover networks for non- coincident drivers,

    S. H. Linkwitz, “Active crossover networks for non- coincident drivers,” Journal of the Audio Engineering Society, vol. 24, no. 1, pp. 2–8, 1976

    1976

  20. [28]

    Differentiable all-pole filters for time- varying audio systems,

    C.-y. Yu, C. Mitcheltree, A. Carson, S. Bilbao, J. Reiss, and G. Fazekas, “Differentiable all-pole filters for time- varying audio systems,” in 27th International Confer- ence on Digital Audio Effects (DAFx) , 2024

    2024

  21. [29]

    Large-scale contrastive language-audio pretraining with feature fusion and keyword-to-caption augmentation,

    Y . Wu, K. Chen, T. Zhang, Y . Hui, T. Berg-Kirkpatrick, and S. Dubnov, “Large-scale contrastive language-audio pretraining with feature fusion and keyword-to-caption augmentation,” in ICASSP 2023-2023 IEEE Interna- tional Conference on Acoustics, Speech and Signal Pro- cessin...

    2023

  22. [30]

    MoisesDB: A dataset for source separation beyond 4- stems,

    I. Pereira, F. Araújo, F. Korzeniowski, and R. V ogl, “MoisesDB: A dataset for source separation beyond 4- stems,” in Proc. ISMIR, 2023

    2023

  23. [31]

    MUSDB18-HQ - an uncompressed version of MUSDB18,

    Z. Rafii, A. Liutkus, F.-R. Stöter, S. I. Mimilakis, and R. Bittner, “MUSDB18-HQ - an uncompressed version of MUSDB18,” Aug. 2019. [Online]. Available: https://doi.org/10.5281/zenodo.3338373

    2019 doi

  24. [32]

    The mtg-jamendo dataset for automatic music tagging,

    D. Bogdanov, M. Won, P. Tovstogan, A. Porter, and X. Serra, “The mtg-jamendo dataset for automatic music tagging,” in Proc. ICML, 2019

    2019

  25. [33]

    Efficient neural net- works for real-time modeling of analog dynamic range compression,

    C. J. Steinmetz and J. D. Reiss, “Efficient neural net- works for real-time modeling of analog dynamic range compression,” arXiv preprint arXiv:2102.06200, 2021

    2021 arXiv

  26. [34]

    On the variance of the adaptive learning rate and beyond,

    L. Liu, H. Jiang, P. He, W. Chen, X. Liu, J. Gao, and J. Han, “On the variance of the adaptive learning rate and beyond,” in International Conference on Learning Representations, 2020

    2020

  27. [35]

    Fréchet audio distance: A metric for evaluat- ing music enhancement algorithms,

    K. Kilgour, M. Zuluaga, D. Roblek, and M. Shar- ifi, “Fréchet audio distance: A metric for evaluat- ing music enhancement algorithms,” arXiv preprint arXiv:1812.08466, 2018

    2018 arXiv

  28. [36]

    CLAP: Learning audio concepts from natural language supervision,

    B. Elizalde, S. Deshmukh, M. Al Ismail, and H. Wang, “CLAP: Learning audio concepts from natural language supervision,” in ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2023, pp. 1–5

    2023

  29. [37]

    High-fidelity audio compression with im- proved rvqgan,

    R. Kumar, P. Seetharaman, A. Luebs, I. Kumar, and K. Kumar, “High-fidelity audio compression with im- proved rvqgan,” Advances in Neural Information Pro- cessing Systems, vol. 36, 2024

    2024

  30. [38]

    High fidelity neural audio compression,

    A. Défossez, J. Copet, G. Synnaeve, and Y . Adi, “High fidelity neural audio compression,” Transactions on Ma- chine Learning Research, 2023

    2023

  31. [39]

    Adapting frechet audio distance for generative music evaluation,

    A. Gui, H. Gamper, S. Braun, and D. Emmanouilidou, “Adapting frechet audio distance for generative music evaluation,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2024, pp. 1331–1335

    2024

  32. [40]

    Searching for music mixing graphs: A pruning approach,

    S. Lee, M. A. Martínez-Ramírez, W.-H. Liao, S. Uhlich, G. Fabbro, K. Lee, and Y . Mitsufuji, “Searching for music mixing graphs: A pruning approach,” in 27th In- ternational Conference on Digital Audio Effects (DAFx), 2024

    2024

  33. [41]

    Reverse engineering of music mixing graphs with differentiable processors and iterative pruning,

    ——, “Reverse engineering of music mixing graphs with differentiable processors and iterative pruning,” Journal of the Audio Engineering Society , vol. 73, pp. 344–365, June 2025

    2025

  34. [42]

    Diffvox: A differentiable model for capturing and analysing pro- fessional effects distributions,

    C.-Y . Yu, M. A. Martínez-Ramírez, J. Koo, B. Hayes, W.-H. Liao, G. Fazekas, and Y . Mitsufuji, “Diffvox: A differentiable model for capturing and analysing pro- fessional effects distributions,” in 28th International Conference on Digital Audio Effects (DAFx) , 2025

    2025

  35. [43]

    Text2FX: Harnessing clap embeddings for text-guided audio ef- fects,

    A. Chu, P. O’Reilly, J. Barnett, and B. Pardo, “Text2FX: Harnessing clap embeddings for text-guided audio ef- fects,” in ICASSP 2025-2025 IEEE International Con- ference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025, pp. 1–5

    2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.