SmartAmp
Key Features
- WaveNet neural network engine trained on real tube amplifier recordings for authentic dynamic response
- Clean and lead channels with independent bass, middle, treble EQ and gain controls for versatile tone shaping
- Master volume and presence knob for mix-ready output control and upper-harmonic clarity
- Open source under Apache 2.0 license with full training code available on GitHub
- Cross-platform support for Windows, macOS, and Linux with VST3, AU, AAX, and Standalone formats
- Lightweight CPU footprint suitable for sessions with multiple amp instances
- Machine-learning models respond dynamically to picking intensity and guitar volume changes
Description
SmartAmp is a guitar amp modeling plugin from GuitarML that uses neural networks to recreate the sound of a real tube amplifier. Built with the JUCE framework and a WaveNet inference engine, it captures the tonal character and dynamic response of hardware amps through machine learning rather than traditional circuit modeling.
The plugin features two channels — clean and lead — each with its own bass, middle, and treble EQ controls plus a gain knob. Both channels use machine-learning models trained on recordings of a small tube amp at various drive settings, producing tones that respond naturally to playing dynamics just like physical hardware.
A master volume and presence control round out the feature set, giving you broad tonal shaping from warm cleans to crunchy overdriven textures. The presence knob adds upper-harmonic shimmer that helps guitar parts cut through a dense mix without becoming harsh.
SmartAmp is open source under the Apache 2.0 license, with the full codebase available on GitHub for anyone interested in audio DSP and neural network modeling. Model training is done using PyTorch on recorded audio samples, and users can train their own models using the companion PedalNetRT repository.
The plugin runs on Windows 7 and up (32-bit or 64-bit), macOS 10.11 and newer (Intel and Apple Silicon via Rosetta), and Linux. Installers provide VST3, AU, AAX, and Standalone formats depending on your operating system.
Video Preview
Frequently Asked Questions
How does SmartAmp differ from traditional amp simulation plugins?
SmartAmp uses a WaveNet neural network trained on actual recordings of a tube amplifier, rather than component-level circuit modeling. This approach captures the nonlinear dynamic behavior of real hardware, meaning the plugin responds naturally to your picking intensity, volume knob adjustments, and playing technique.
Can I train my own amp or pedal models for SmartAmp?
Yes, the model training code is open source on GitHub using PyTorch and the PedalNetRT repository. You record input/output audio samples from your hardware, run the training script, and load the resulting model into the plugin. Note that as of version 1.3, custom model loading was moved to the companion SmartGuitarPedal plugin.
Does SmartAmp include cabinet simulation?
SmartAmp does not include a separate cabinet simulation block. The neural network model captures the combined response of the amp as it was recorded, which may include cabinet coloration depending on how the training data was captured. For additional cabinet shaping, you can pair it with a third-party IR loader plugin.
How does SmartAmp compare to GuitarML's other plugins like Proteus and Chameleon?
SmartAmp was GuitarML's first plugin and uses an earlier WaveNet architecture. Chameleon uses the newer RTNeural engine for lower CPU usage and offers three distinct amp voicings. Proteus supports loading user-captured models via a simpler capture utility. All three are complementary tools in the GuitarML ecosystem.
