Paper Archive

This paper explores the digital emulation of analog dynamic range compressors, proposing a grey-box model that uses a combination of traditional signal processing techniques and machine learning. The main idea is to use the structure of a traditional digital compressor in a machine learning framework, so it can be trained end-to-end to create a virtual analog model of a compressor from data. The complexity of the model can be adjusted, allowing a trade-off between the model accuracy and computational cost. The proposed model has interpretable components, so its behaviour can be controlled more readily after training in comparison to a black-box model. The result is a model that achieves similar accuracy to a black-box baseline, whilst requiring less than 10% of the number of operations per sample at runtime.

Download

The integration of additional side information to improve music source separation has been investigated numerous times, e.g., by adding features to the input or by adding learning targets in a multi-task learning scenario. These approaches, however, require additional annotations such as musical scores, instrument labels, etc. in training and possibly during inference. The available datasets for source separation do not usually provide these additional annotations. In this work, we explore transfer learning strategies to incorporate VGGish features with a state-of-the-art source separation model; VGGish features are known to be a very condensed representation of audio content and have been successfully used in many music information retrieval tasks. We introduce three approaches to incorporate the features, including two latent space regularization methods and one naive concatenation method. Our preliminary results show that our proposed approaches could improve some evaluation metrics for music source separation. In this work, we also include a discussion of our proposed approaches, such as the pros and cons of each approach, and the potential extension/improvement.

Download

Sound matching allows users to automatically approximate existing sounds using a synthesizer. Previous work has mostly focused on algorithms for automatically programming an existing synthesizer. This paper proposes a system for selecting between different synthesizer designs, each one with a corresponding automatic programmer. An implementation that allows designing ensembles based on a template is demonstrated. Several experiments are presented using a simple subtractive synthesis design. Using an ensemble of synthesizer-programmer pairs is shown to provide better matching than a single programmer trained for an equivalent integrated synthesizer. Scaling to hundreds of synthesizers is shown to improve match quality.

Download

We describe two submissions to Task B of the 1st DAFx Parameter Estimation Challenge, which estimates an unknown number of modal frequency, decay, and gain triples from a synthetic plate-reverb impulse response. The first method combines pooled spectral features with time-domain and absolute-scale conditioning in a real-valued convolutional count-density network, while the second uses a complex-valued Transformer count-density network. Both methods jointly infer the modal count and per-mode attributes directly from the IR. On an independently generated 100-IR comparison set, the two neural estimators achieve lower overall challenge error than the evaluated classical baselines, with frequency and decay estimation substantially more accurate than gain estimation.

Download

Using sensors to extract gestural information for control parameters of digital audio effects is common practice. There has also been research using machine learning techniques to classify specific gestures based on audio feature analysis. In this paper, we will describe our experiments in training a computer to map the appropriate audio-based features to look like sensor data, in order to potentially eliminate the need for sensors. Specifically, we will show our experiments using the ESitar, a digitally enhanced sensor based controller modeled after the traditional North Indian sitar. We utilize multivariate linear regression to map continuous audio features to continuous gestural data.

Download

We present Clean2FX, a study and demo of label-conditioned clean-to-effect transformation for electric guitar audio. Given a clean guitar input and a target effect label, the task is to synthesize the corresponding effected signal while preserving the musical content. Training and evaluation pairs are constructed from EGFxSet real, single-tone recordings by assembling matched clean/effected chords, melodies, and mixed timelines. This allows for controlled comparison across effects. We evaluate four neural approaches under a common spectrogram-based transformation setting: two variational autoencoders and two U-Net models that differ in whether they operate on linear or log-magnitude representations. Performance is measured using linear-magnitude spectrogram MSE and Fréchet Audio Distance. The U-Net models outperform the variational autoencoder variants. Per-effect results show that distortion effects are most readily improved, whereas delay and reverb effects exhibit weaker FAD gains despite substantial spectral-error reductions. A conditioning-sensitivity diagnostic provides evidence that the best model responds to target labels rather than collapsing to a single transformation. Our demo website compares two models applied on real-world guitar performances outside training and validation data, providing audio and spectrogram examples of the practical clean-to-effect behavior.

Download

We release synth1B1, a multi-modal audio corpus consisting of 1 billion 4-second synthesized sounds, paired with the synthesis parameters used to generate them. The dataset is 100x larger than any audio dataset in the literature. We also introduce torchsynth, an open source modular synthesizer that generates the synth1B1 samples on-the-fly at 16200x faster than real-time (714MHz) on a single GPU. Finally, we release two new audio datasets: FM synth timbre and subtractive synth pitch. Using these datasets, we demonstrate new rank-based evaluation criteria for existing audio representations. Finally, we propose a novel approach to synthesizer hyperparameter optimization.

Download

In this paper, we propose diffusion warm initialization as a simple yet effective approach for a range of audio-to-audio transformation tasks. To illustrate the generality of the approach, we demonstrate its use in timbre transfer, MIDI-to-Real synthesis, and multiple audio enhancement tasks. We conduct a detailed empirical analysis on timbre transfer to investigate the role of the initialization time t_init. The effect of t_init is evaluated using pitch-based Jaccard Distance and Fréchet Audio Distance to quantify faithfulness to the input signal and alignment with the target distribution. Our results provide practical guidance for selecting t_init and show that, once properly chosen, a single pretrained diffusion model combined with warm initialization can support multiple transformation objectives without task-specific training or conditioning. Despite its simplicity, this approach already achieves competitive results when compared with more complex pipelines designed specifically for these tasks.

Download

Extraction of symbolic information from signals is an active field of research enabling numerous applications especially in the Musical Information Retrieval domain. This complex task, that is also related to other topics such as pitch extraction or instrument recognition, is a demanding subject that gave birth to numerous approaches, mostly based on advanced signal processing-based algorithms. However, these techniques are often non-generic, allowing the extraction of definite physical properties of the signal (pitch, octave), but not allowing arbitrary vocabularies or more general annotations. On top of that, these techniques are one-sided, meaning that they can extract symbolic data from an audio signal, but cannot perform the reverse process and make symbol-to-signal generation. In this paper, we propose an bijective approach for signal/symbol translation by turning this problem into a density estimation task over signal and symbolic domains, considered both as related random variables. We estimate this joint distribution with two different variational auto-encoders, one for each domain, whose inner representations are forced to match with an additive constraint, allowing both models to learn and generate separately while allowing signal-to-symbol and symbol-to-signal inference. In this article, we test our models on pitch, octave and dynamics symbols, which comprise a fundamental step towards music transcription and label-constrained audio generation. In addition to its versatility, this system is rather light during training and generation while allowing several interesting creative uses that we outline at the end of the article.

Download

Relative music loudness estimation is a MIR task that consists in dividing audio in segments of three classes: Foreground Music, Background Music and No Music. Given the temporal correlation of music, in this work we approach the task using a type of network with the ability to model temporal context: the Temporal Convolutional Network (TCN). We propose two architectures: a TCN, and a novel architecture resulting from the combination of a TCN with a Convolutional Neural Network (CNN) front-end. We name this new architecture CNN-TCN. We expect the CNN front-end to work as a feature extraction strategy to achieve a more efficient usage of the network’s parameters. We use the OpenBMAT dataset to train and test 40 TCN and 80 CNN-TCN models with two grid searches over a set of hyper-parameters. We compare our models with the two best algorithms submitted to the tasks of music detection and relative music loudness estimation in MIREX 2019. All our models outperform the MIREX algorithms even when using a lower number of parameters. The CNN-TCN emerges as the best architecture as all its models outperform all TCN models. We show that adding a CNN front-end to a TCN can actually reduce the number of parameters of the network while improving performance. The CNN front-end effectively works as a feature extractor producing consistent patterns that identify different combinations of music and non-music sounds and also helps in producing a smoother output in comparison to the TCN models.

Download