mikeoliphant/neural-amp-modeler-lv2

What it solves

This project provides an LV2 plugin that allows musicians and audio engineers to play back neural network-based amplifier models. It enables the use of high-fidelity machine learning models of guitar amps and other audio equipment within a Digital Audio Workstation (DAW) that supports the LV2 standard.

How it works

The plugin uses the NeuralAudio engine to load and process audio through pre-trained neural network models (such as NAM A1, A2, and RTNeural keras json models). It processes audio in real-time, requiring the host DAW to run at the same sample rate as the model was trained at (typically 48kHz), though it supports oversampling if the run-time sample rate is an even multiple of the training rate.

Who it’s for

Audio producers, guitarists, and sound designers using Linux, macOS, or Windows who use LV2-compatible DAWs (like Reaper, Carla, or Ardour) to model the sonic characteristics of captured amplifier hardware.

Highlights

• Supports Neural Amp Modeler (NAM) A1 and A2 models as well as RTNeural keras json models.
• Supports oversampling to reduce aliasing at the cost of performance.
• Includes a smart bypass feature to stop processing when the input signal is silent.
• Provides controls for input gain, output volume, and model quality selection.