What it does SmartGuitarAmp is a cross-platform VST/AU plugin that models a small tube amplifier at clean and overdriven settings. It runs a WaveNet neural network in real time inside your DAW, with gain and EQ knobs to shape the modeled sound. You download an installer, reboot, and pretend you own a Fender Champ.

The interesting bit The project sits at an unusual intersection: it packages serious machine-learning inference (PyTorch-trained models serialized to JSON) into a musician-friendly binary. The training pipeline lives in a separate repo, PedalNetRT, which means the plugin itself is essentially a hardened runtime for guitar tone models.

Key highlights

• Real-time WaveNet inference in C++ via JUCE
• Pre-trained on a small tube amp; clean and drive channels included
• Cross-platform installers for Windows, Mac, and Linux
• CMake build with submodule setup; Apache 2.0 licensed
• Companion SmartGuitarPedal handles user-trained models (custom loading was removed here in v1.3)

Caveats

• Custom model loading was stripped in version 1.3; tinkerers need the sister pedal project
• The README notes you may need to reboot after install for DAW detection, which feels more 2003 than 2024

Verdict Guitarists who want credible tube tone without the maintenance, and audio developers curious about deploying ML models in JUCE, should grab this. If you need to load your own trained models out of the box, head to SmartGuitarPedal instead.