Download Parameter Estimation via Differentiable Modal Plate Synthesis We present our submission to Task A of the 1st DAFx Parameter Estimation Challenge, which concerns the estimation of the physical parameters of a vibrating plate from a synthetic impulse response. Our approach introduces a differentiable modal plate synthesizer and estimates the plate parameters through inference-time gradient-based optimization of the synthesizer parameters. The six target parameters are recovered by minimizing a multi-scale spectral loss via backpropagation through the differentiable plate model. To handle the non-convexity of the loss landscape, we adopt a two-phase training strategy consisting of multiple short-term probe optimizations, followed by full-scale refinement initialized from the best candidate. We evaluate the approach on eight impulse responses synthesized with the official challenge dataset generator. Compared with a constant-value predictor and the particle swarm optimization baseline provided by the challenge, the proposed method reduces the prediction error by approximately one order of magnitude.
Download ICGAN: An Implicit Conditioning Method for Interpretable Feature Control of Neural Audio Synthesis Neural audio synthesis methods can achieve high-fidelity and realistic sound generation by utilizing deep generative models. Such models typically rely on external labels which are often discrete as conditioning information to achieve guided sound generation. However, it remains difficult to control the subtle changes in sounds without appropriate and descriptive labels, especially given a limited dataset. This paper proposes an implicit conditioning method for neural audio synthesis using generative adversarial networks that allows for interpretable control of the acoustic features of synthesized sounds. Our technique creates a continuous conditioning space that enables timbre manipulation without relying on explicit labels. We further introduce an evaluation metric to explore controllability and demonstrate that our approach is effective in enabling a degree of controlled variation of different synthesized sound effects for in-domain and cross-domain sounds.
Download Piano-SSM: Diagonal State Space Models for Efficient Midi-to-Raw Audio Synthesis Deep State Space Models (SSMs) have shown remarkable performance in long-sequence reasoning tasks, such as raw audio
classification, and audio generation. This paper introduces PianoSSM, an end-to-end deep SSM neural network architecture designed to synthesize raw piano audio directly from MIDI input.
The network requires no intermediate representations or domainspecific expert knowledge, simplifying training and improving accessibility.
Quantitative evaluations on the MAESTRO dataset
show that Piano-SSM achieves a Multi-Scale Spectral Loss (MSSL)
of 7.02 at 16kHz, outperforming DDSP-Piano v1 with a MSSL of
7.09. At 24kHz, Piano-SSM maintains competitive performance
with an MSSL of 6.75, closely matching DDSP-Piano v2’s result of 6.58. Evaluations on the MAPS dataset achieve an MSSL
score of 8.23, which demonstrates the generalization capability
even when training with very limited data. Further analysis highlights Piano-SSM’s ability to train on high sampling-rate audio
while synthesizing audio at lower sampling rates, explicitly linking performance loss to aliasing effects. Additionally, the proposed model facilitates real-time causal inference through a custom C++17 header-only implementation. Using an Intel Core i712700 processor at 4.5GHz, with single core inference, allows synthesizing one second of audio at 44.1kHz in 0.44s with a workload of 23.1GFLOPS/s and an 10.1µs input/output delay with the
largest network. While the smallest network at 16kHz only needs
0.04s with 2.3GFLOP/s and 2.6µs input/output delay. These results underscore Piano-SSM’s practical utility and efficiency in
real-time audio synthesis applications.
Download A jump start for NMF with N-FINDR and NNLS Nonnegative Matrix Factorization is a popular tool for the analysis of audio spectrograms. It is usually initialized with random data, after which it iteratively converges to a local optimum. In this paper we show that N-FINDR and NNLS, popular techniques for dictionary and activation matrix learning in remote sensing, prove useful to create a better starting point for NMF. This reduces the number of iterations necessary to come to a decomposition of similar quality. Adapting algorithms from the hyperspectral image unmixing and remote sensing communities, provides an interesting direction for future research in audio spectrogram factorization.
Download Differentiable All-Pass Filters for Phase Response Estimation and Automatic Signal Alignment Virtual analog (VA) audio effects are increasingly based on neural networks and deep learning frameworks. Due to the underlying black-box methodology, a successful model will learn to approximate the data it is presented, including potential errors such as latency and audio dropouts as well as non-linear characteristics and frequency-dependent phase shifts produced by the hardware. The latter is of particular interest as the learned phase-response might cause unwanted audible artifacts when the effect is used for creative processing techniques such as dry-wet mixing or parallel compression. To overcome these artifacts we propose differentiable signal processing tools and deep optimization structures for automatically tuning all-pass filters to predict the phase response of different VA simulations, and align processed signals that are out of phase. The approaches are assessed using objective metrics while listening tests evaluate their ability to enhance the quality of parallel path processing techniques. Ultimately, an overparameterized, BiasNet-based, all-pass model is proposed for the optimization problem under consideration, resulting in models that can estimate all-pass filter coefficients to align a dry signal with its affected, wet, equivalent.
Download Enhancing Automatic Chord Recognition via Pseudo-Labeling and Knowledge Distillation Automatic Chord Recognition (ACR) is constrained by the scarcity of aligned chord annotations, which are costly to acquire. At the same time, open-weight pre-trained models are more accessible than their proprietary training data. In this work, we present a two-stage training pipeline that leverages pre-trained models together with unlabeled audio. The proposed method decouples training into two stages. In the first stage, we use the pre-trained BTC model as a teacher to generate pseudo-labels for over 1,000 hours of diverse unlabeled audio and train a student model solely on these pseudo-labels. In the second stage, the student is continually trained on ground-truth labels as they become available. To prevent catastrophic forgetting of the representations learned in the first stage, we apply selective knowledge distillation (KD) from the teacher as a regularizer. In our experiments, two models (BTC, 2E1D) were used as students. In Stage 1, using only pseudo-labels, the BTC student achieves about 99% of the teacher's performance, while the 2E1D model achieves about 97% of the teacher's performance across seven standard mir_eval metrics. After continual training with labeled data in Stage 2, the resulting BTC student model consistently surpasses both the traditional supervised learning baseline and the original pre-trained teacher model across all metrics. The resulting 2E1D student model also outperforms the supervised baseline and approaches teacher-level performance, with both models demonstrating substantial gains on rare chord qualities.
Download Differentiable Piano Model for Midi-to-Audio Performance Synthesis Recent neural-based synthesis models have achieved impressive results for musical instrument sound generation. In particular, the Differentiable Digital Signal Processing (DDSP) framework enables the usage of spectral modeling analysis and synthesis techniques in fully differentiable architectures. Yet currently, it has only been used for modeling monophonic instruments. Leveraging the interpretability and modularity of this framework, the present work introduces a polyphonic differentiable model for piano sound synthesis, conditioned on Musical Instrument Digital Interface (MIDI) inputs. The model architecture is motivated by high-level acoustic modeling knowledge of the instrument which, in tandem with the sound structure priors inherent to the DDSP components, makes for a lightweight, interpretable and realistic sounding piano model. The proposed model has been evaluated in a listening test, demonstrating improved sound quality compared to a benchmark neural-based piano model, with significantly less parameters and even with reduced training data. The same listening test indicates that physical-modeling-based models still achieve better quality, but the differentiability of our lightened approach encourages its usage in other musical tasks dealing with polyphonic audio and symbolic data.
Download Automatic Violin Synthesis Using Expressive Musical Term Features The control of interpretational properties such as duration, vibrato, and dynamics is important in music performance. Musicians continuously manipulate such properties to achieve different expressive intentions. This paper presents a synthesis system that automatically converts a mechanical, deadpan interpretation to distinct expressions by controlling these expressive factors. Extending from a prior work on expressive musical term analysis, we derive a subset of essential features as the control parameters, such as the relative time position of the energy peak in a note and the mean temporal length of the notes. An algorithm is proposed to manipulate the energy contour (i.e. for dynamics) of a note. The intended expressions of the synthesized sounds are evaluated in terms of the ability of the machine model developed in the prior work. Ten musical expressions such as Risoluto and Maestoso are considered, and the evaluation is done using held-out music pieces. Our evaluations show that it is easier for the machine to recognize the expressions of the synthetic version, comparing to those of the real recordings of an amateur student. While a listening test is under construction as a next step for further performance validation, this work represents to our best knowledge a first attempt to build and quantitatively evaluate a system for EMT analysis/synthesis.
Download Kronos VST – The Programmable Effect Plugin This paper introduces Kronos VST, an audio effect plugin conforming to the VST 3 standard that can be programmed on the fly by the user, allowing entire signal processors to be defined in real time. A brief survey of existing programmable plugins or development aids for audio effect plugins is given. Kronos VST includes a functional just in time compiler that produces high performance native machine code from high level source code. The features of the Kronos programming language are briefly covered, followed by the special considerations of integrating user programs into the VST infrastructure. Finally, introductory example programs are provided.
Download HRTF Spatial Upsampling in the Spherical Harmonics Domain Employing a Generative Adversarial Network A Head-Related Transfer Function (HRTF) is able to capture alterations a sound wave undergoes from its source before it reaches the entrances of a listener’s left and right ear canals, and is imperative for creating immersive experiences in virtual and augmented reality (VR/AR). Nevertheless, creating personalized HRTFs demands sophisticated equipment and is hindered by time-consuming data acquisition processes. To counteract these challenges, various techniques for HRTF interpolation and up-sampling have been proposed. This paper illustrates how Generative Adversarial Networks (GANs) can be applied to HRTF data upsampling in the spherical harmonics domain. We propose using Autoencoding Generative Adversarial Networks (AE-GAN) to upsample lowdegree spherical harmonics coefficients and get a more accurate representation of the full HRTF set. The proposed method is benchmarked against two baselines: barycentric interpolation and HRTF selection. Results from log-spectral distortion (LSD) evaluation suggest that the proposed AE-GAN has significant potential for upsampling very sparse HRTFs, achieving 17% improvement over baseline methods.