
DDSP
Combine neural networks with differentiable audio synthesis and effects
Pricing
Pricing not verified
- Use a neural network to control interpretable DSP components rather than generate every sample directly.
- Combine modular synthesis and effects components in audio-modeling experiments.
What is DDSP?
DDSP is a research library that combines classical digital signal processing with neural networks for audio synthesis. Instead of learning to produce every audio sample from scratch, a model learns control parameters for components such as oscillators, filters and reverberation. Developers can supply control signals or features extracted from audio to guide the resulting sound.
Its interpretable, modular components support experiments such as timbre transfer, where pitch and loudness are retained while the character of an instrument changes. The library also includes synthesis and effects components for building differentiable audio models. Tutorial notebooks provide a starting point for researchers and developers who want to explore these techniques in code.
Learn controls for an audio synthesizer
- Learned DSP controls Map input signals or audio features to parameters for sound synthesis.
- Differentiable components Train a model using standard backpropagation through DSP units.
Explore timbre transfer and audio effects
- Timbre transfer Change an instrument’s character while retaining frequency and loudness.
- Synthesis and effects modules Work with wavetable synthesis, time-varying filters and delay modulation.
User reviews
No reviews yet. Be the first to review this tool.
Log in to write a review.
Join the Aivolut newsletter
Occasional practical AI tips, app updates, and offers. Unsubscribe anytime.