Parameter Inference of Music Synthesizers with Deep Learning

Explore a conference talk on parameter inference of music synthesizers using deep learning techniques. Delve into the potential of automating synthesizer preset generation based on desired audio samples. Examine recent research applying deep learning to various synthesizer types, including FM and wavetable. Discover the challenges faced in this field and gain insights into neural network basics, dataset building, and advanced learning approaches like self-supervised and semi-supervised learning. Learn about differentiable DSP and its applications in sound design. Ideal for audio developers, music producers, and machine learning enthusiasts interested in the intersection of AI and music technology.

Introduction
Motivation
Demo
Additive Synthesis
Subtractive Synthesis
Wavetable Synthesis
FM Synthesis
Other Methods
Why Parameter Inference
Parameters
References
Deep Learning Basics
Neural Network Blocks
Why Deep Learning
Building a Dataset
Syntheon
Json
Cnns
Serums
Other works
Selfsupervised learning
Differentiability
Differentiable DSP
Semisupervised learning
Discussion
Summary
References Shoutouts