🤖 AI Summary
Optical compressors (e.g., LA-2A, CL-1B) exhibit complex, time-varying nonlinear dynamics that are challenging to model accurately in real-time audio processing. To address this, we propose a low-latency black-box modeling paradigm. We introduce Selective State Space Models (SSMs) — for the first time — into audio dynamic processing, integrating Feature-wise Linear Modulation (FiLM) and Gated Linear Units (GLUs) to enable real-time, conditional control of attack and release parameters. The model is fully end-to-end trainable and generalizes to unseen parameter configurations, significantly improving modeling fidelity—especially under challenging regimes such as fast-attack/slow-release dynamics. Quantitative evaluations (e.g., SI-SNR, PESQ) and controlled listening tests demonstrate consistent superiority over state-of-the-art methods. This work establishes a new high-fidelity, ultra-low-latency paradigm for optical compressor modeling, enabling practical deployment in live music production.
📝 Abstract
This paper presents a method for modeling optical dynamic range compressors using deep neural networks with Selective State Space models. The proposed approach surpasses previous methods based on recurrent layers by employing a Selective State Space block to encode the input audio. It features a refined technique integrating Feature-wise Linear Modulation and Gated Linear Units to adjust the network dynamically, conditioning the compression's attack and release phases according to external parameters. The proposed architecture is well-suited for low-latency and real-time applications, crucial in live audio processing. The method has been validated on the analog optical compressors TubeTech CL 1B and Teletronix LA-2A, which possess distinct characteristics. Evaluation is performed using quantitative metrics and subjective listening tests, comparing the proposed method with other state-of-the-art models. Results show that our black-box modeling methods outperform all others, achieving accurate emulation of the compression process for both seen and unseen settings during training. We further show a correlation between this accuracy and the sampling density of the control parameters in the dataset and identify settings with fast attack and slow release as the most challenging to emulate.