FULLY CONDITIONED AND LOW-LATENCY BLACK-BOX MODELING OF ANALOG COMPRESSION

Riccardo Simionato, Stefano Fasciani
University of Oslo
DAFx 2023
Paper Code

Abstract

Neural networks have been found suitable for virtual analog modeling applications. Several analog audio effects have been successfully modeled with deep learning techniques, using low-latency and conditioned architectures suitable for real-world applications. Challenges remain with effects presenting more complex responses, such as nonlinear and time-varying input-output relationships. This paper proposes a deep-learning model for the analog compression effect. The architecture we introduce is fully conditioned by the device control parameters and it works on small audio segments, allowing low-latency real-time implementations. The architecture is used to model the CL 1B analog optical compressor, showing an overall high accuracy and ability to capture the different attack and release compression profiles. The proposed architecture' ability to model audio compression behaviors is also verified using datasets from other compressors. Limitations remain with heavy compression scenarios determined by the conditioning parameters.

Datasets

TubeTech CL 1B Dataset| PSP MicroComp Dataset | U-he Presswerk Dataset | Softube FET Dataset | LA-2A Dataset

TubeTech CL 1B - ED (Light) Audio Examples

Attack Time: 150 ms, Release Time: 0,005 s, Ratio 6:1, Threshold -20dBu.

STFT time
Input Target Prediciton

Attack Time: 0.5 ms, Release Time: 0,005 s, Ratio 10:1, Threshold -20dBu.

STFT time
Input Target Prediciton

Attack Time: 0.5 ms, Release Time: 0,005 s, Ratio 10:1, Threshold -40dBu.

STFT time
Input Target Prediciton

Teletronix LA-2A - ED (Light) Audio Examples

Mode: Compression, Peak Reduction 80.

STFT time
Input Target Prediciton

Mode: Compression, Peak Reduction 80.

STFT time
Input Target Prediciton

Mode: Compression, Peak Reduction 80.

STFT time
Input Target Prediciton

TubeTech CL 1B - ED Audio Examples

Attack Time: 0.5 ms, Release Time: 0,005 s, Ratio 10:1, Threshold -20dBu.

STFT time
Input Target Prediciton

Attack Time: 150 ms, Release Time: 0,005 s, Ratio 6:1, Threshold -20dBu.

STFT time
Input Target Prediciton

Attack Time: 0.5 ms, Release Time: 0,005 s, Ratio 2:1, Threshold -40dBu.

STFT time
Input Target Prediciton

Teletronix LA-2A - ED Audio Examples

Mode: Compression, Peak Reduction 80.

STFT time
Input Target Prediciton

Mode: Compression, Peak Reduction 80.

STFT time
Input Target Prediciton

Mode: Compression, Peak Reduction 80.

STFT time
Input Target Prediciton

Softube FET - ED Audio Examples

Attack Time: 0.5 ms, Release Time: 0,005 s, Ratio 10:1, Threshold 20

STFT time
Input Target Prediciton

U-he Presswerk - ED Audio Examples

Attack Time: 0.5 ms, Release Time: 0,005 s, Ratio 10:1, Threshold 20

STFT time
Input Target Prediciton

PSP MicroComp - ED Audio Examples

Attack Time: 0.5 ms, Release Time: 0,005 s, Ratio 10:1, Threshold 20

STFT time
Input Target Prediciton

TubeTech CL 1B - TCN Audio Examples

Attack Time: 150 ms, Release Time: 0,005 s, Ratio 6:1, Threshold -20dBu.

STFT time
Input Target Prediciton

Attack Time: 150 ms, Release Time: 0,005 s, Ratio 6:1, Threshold -20dBu.

STFT time
Input Target Prediciton

BibTeX

@inproceedings{simionato2023fully,
  title={Fully conditioned and low-latency black-box modeling of analog compression},
  author={Simionato, Riccardo and Fasciani, Stefano},
  booktitle={Proceedings of the International Conference on Digital Audio Effects},
  year={2023},
  organization={DAFx Board}
}