DDSP

DDSP is the Differentiable Digital Signal Processing (DDSP) library. It is aimed at researchers and developers working on audio synthesis and modeling and combines interpretable DSP elements with neural networks to convert control signals into DSP parameters.

It uses simple interpretable DSP elements to create complex realistic signals by precisely controlling their many parameters, and DDSP passes parameters through known sound synthesis algorithms. Using differentiable oscillators, filters, and reverberation from the DDSP library, the project can train high-quality audio synthesis models with less data and fewer parameters.