NablAFx: A Framework for Differentiable Black-box and Gray-box Modeling of Audio Effects

Comunità M, Steinmetz CJ, Reiss J. NablAFx: A Framework for Differentiable Black-box and Gray-box Modeling of Audio Effects. In: 2025 AES International Conference on Artificial Intelligence and Machine Learning for Audio. Audio Engineering Society; 2025. Paper 34. Available from: https://aes.org/publications/elibrary-page/?id=23023

@inproceedings{Comunita2025_23023,
  author = {Comunità, Marco and Steinmetz, Christian J. and Reiss, Joshua},
  title = {{NablAFx: A Framework for Differentiable Black-box and Gray-box Modeling of Audio Effects}},
  booktitle = {2025 AES International Conference on Artificial Intelligence and Machine Learning for Audio},
  note = {Paper 34},
  year = {2025},
  month = sep,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=23023}
}

TY  - CPAPER
TI  - NablAFx: A Framework for Differentiable Black-box and Gray-box Modeling of Audio Effects
AU  - Comunità, Marco
AU  - Steinmetz, Christian J.
AU  - Reiss, Joshua
T2  - 2025 AES International Conference on Artificial Intelligence and Machine Learning for Audio
M1  - Paper 34
PY  - 2025
DA  - 2025/09/04
UR  - https://aes.org/publications/elibrary-page/?id=23023
PB  - Audio Engineering Society
LA  - en
AB  - We present NablAFx, an open-source framework developed to support research in differentiable black-box and gray-box modeling of audio effects. Built in PyTorch, NablAFx offers a versatile ecosystem to configure, train, evaluate, and compare various architectural approaches. It includes classes to manage model architectures, datasets, and training, along with features to compute and log losses, metrics and media, and plotting functions to facilitate detailed analysis. It incorporates implementations of established black-box architectures and conditioning methods as well as differentiable DSP blocks and controllers, enabling the creation of both parametric and non-parametric gray-box signal chains. Beside established conditioning methods like concatenation, featurewise linear modulation (FiLM) and temporal feature-wise linear modulation (TFiLM), we propose three further methods: time-varying concatenation (TVCond), tiny TFiLM (TTFiLM) and time-varying FiLM (TVFiLM), as efficient implementations of time-varying conditioning similar to TFiLM. We also propose the Static Rational Linearity as a flexible and efficient differentiable processor to learn nonlinear functions. The code is accessible at https://github.com/mcomunita/nablafx.
ER  -