Fanny Roche

Théo Chasle Cauchy, Modan Tailleur, Lindsey Reymore et al.

Measuring neural audio synthesizers' performance is now routinely conducted using distribution based metrics such as the Fréchet Audio Distance (FAD). Although this metric can be correlated with human perception, it offers limited interpretability beyond ranking different approaches. In this paper, we introduce a deep neural timbre trait predictor composed of a pretrained audio neural embedding (CLAP), and a shallow learnable component. The latter is trained using the RWC musical instrument database and human judgments of 20 timbre descriptions (e.g., woody, percussive, rumbling, etc.) for 31 instruments. The resulting model shows strong correlation with average human ratings (r = 0.66, p < 0.001). We then demonstrate the benefit of this predictor for evaluating the performance of TokenSynth, a neural sound synthesizer. First, the Mean Absolute Error (MAE) computed over the set of generated sounds under different conditioning modalities of the model provides the same ranking as a FAD computed with the RWC database as a reference, suggesting that the proposed predictors are able to provide equivalent information on a distributional basis. Second, because the model is able to qualitatively analyze isolated sounds, we can determine which generated sounds could be improved and identify specific timbral dimensions that need adjustment.

Yann Bourdin, Pierrick Legrand, Fanny Roche

Deep learning has become a standard approach for the modeling of audio effects, yet strictly black-box modeling remains problematic for time-varying systems. Unlike time-invariant effects, training models on devices with internal modulation typically requires the recording or extraction of control signals to ensure the time-alignment required by standard loss functions. This paper introduces a Generative Adversarial Network (GAN) framework to model such effects using only input-output audio recordings, removing the need for modulation signal extraction. We propose a convolutional-recurrent architecture trained via a two-stage strategy: an initial adversarial phase allows the model to learn the distribution of the modulation behavior without strict phase constraints, followed by a supervised fine-tuning phase where a State Prediction Network (SPN) estimates the initial internal states required to synchronize the model with the target. Additionally, a new objective metric based on chirp-train signals is developed to quantify modulation accuracy. Experiments modeling a vintage hardware phaser demonstrate the method's ability to capture time-varying dynamics in a fully black-box context.

Fanny Roche, Thomas Hueber, Samuel Limier et al.

This study investigates the use of non-linear unsupervised dimensionality reduction techniques to compress a music dataset into a low-dimensional representation which can be used in turn for the synthesis of new sounds. We systematically compare (shallow) autoencoders (AEs), deep autoencoders (DAEs), recurrent autoencoders (with Long Short-Term Memory cells -- LSTM-AEs) and variational autoencoders (VAEs) with principal component analysis (PCA) for representing the high-resolution short-term magnitude spectrum of a large and dense dataset of music notes into a lower-dimensional vector (and then convert it back to a magnitude spectrum used for sound resynthesis). Our experiments were conducted on the publicly available multi-instrument and multi-pitch database NSynth. Interestingly and contrary to the recent literature on image processing, we can show that PCA systematically outperforms shallow AE. Only deep and recurrent architectures (DAEs and LSTM-AEs) lead to a lower reconstruction error. The optimization criterion in VAEs being the sum of the reconstruction error and a regularization term, it naturally leads to a lower reconstruction accuracy than DAEs but we show that VAEs are still able to outperform PCA while providing a low-dimensional latent space with nice "usability" properties. We also provide corresponding objective measures of perceptual audio quality (PEMO-Q scores), which generally correlate well with the reconstruction error.