Skip to main content

Neural Audio Theory

Understand How AI Music Systems Are Engineered

A practical, technically rigorous guide to creating with AI music systems, understanding their internals, and evaluating their limits.

For music makersProducer pathPrompt, arrange, troubleshoot, mix, and release.For technical readersEngineering pathStudy signals, models, training, evaluation, and deployment.

Open source · 50+ focused chapters · equations, workflows, and primary references

Inside the guide

From sound to system

Follow the complete technical path, or enter at the layer relevant to your work.

Explore →

Signal Representation Pipeline

Track the engineering path from waveform capture to STFT, mel scaling, and latent embeddings used by modern generative audio stacks.

Explore →

Model Internals and Training

Study how transformer attention, diffusion denoising, and objective functions interact to optimize timbre, rhythm, and structural coherence.

Explore →

Conditioning and Prompt Control

Learn how text tokens become conditioning vectors that steer genre, arrangement, and dynamics in text-to-music inference pipelines.

A connected curriculum

See the whole signal path​

Move from physical sound to representations, generation, and a measured output—without losing the connections between each layer.

Open the complete learning map →
  1. 01SoundWaveforms & perception
  2. 02RepresentSpectra, tokens & latents
  3. 03GenerateModels & conditioning
  4. 04EvaluateListening & measurements