Eli Silverstein

project Technology Sparse Neural FX

Sparse Neural FX

A 90%-pruned WaveNet guitar amp, real time on an iPhone

Overview

This is the project for the paper I worked on with Ryota Sato and submitted to DAFx26. Ryota is a model training and pruning unicorn and pruned a wavenet guitar amplifier model such that 90% of the AI computations are reduced and the neural guitar amp sounds basically identical.

We presented at DAFx26 and showed in a demo how the amp sounds the same as real guitar pedals while being very compute light. Made a custom audio USB dongle for our demo too.

PAPERWaveNet-Style Guitar Amplifier Model Pruning for Real-Time iOS Deployment, Ryota Sato & Eli Silverstein, arXiv 2607.10086 · accepted as a DAFx 2026 demo.

The vintage skin: dot-matrix readout, jewel lamps, brass
The vintage skin: dot-matrix readout, jewel lamps, brass

The app

Amp models, a reverb, and a pedal chain with presets.

Amp screen: model, modeler output gain, reverb mix and decay
Amp screen: model, modeler output gain, reverb mix and decay
Pedalboard screen: presets and the effect chain
Pedalboard screen: presets and the effect chain

← back to Technology