AI guitar tone, now in your DAW (with caveats)

What it does

This is the official plugin wrapper for Neural Amp Modeler, a project that uses neural networks to clone guitar amplifiers and pedals. It packages those models into a VST3 or AudioUnit you can load in a DAW, built on the iPlug2 framework. There’s also a standalone app, though the README nudges you toward the plugin version for anything serious.

The interesting bit

The project is essentially a cleanup job — a refactored, “better practices” version of an earlier iPlug2-based attempt, done with direct guidance from the iPlug2 maintainers. That makes it a case study in how research-grade audio ML slowly gets domesticated for actual musicians. The author also maintains a separate “Gateway” fork with pre-built installers, suggesting this repo is increasingly the build-it-yourself upstream.

Key highlights

• Supports Windows 10+ and macOS 10.15+; Linux users are pointed to a community LV2 fork
• Could expand to AAX, CLAP, and iOS — the README flags these as future possibilities, not promises
• Standalone version exists but has “sparser” I/O features and less routing stability than DAW-hosted use
• Graphics backend can crash on launch if Windows systems try to run it on integrated rather than discrete GPUs

Caveats

• Standalone I/O is explicitly less stable than plugin-host operation
• GPU selection issue on Windows requires manual control-panel intervention; the author is asking for help debugging

Verdict

Guitarists and producers already using Neural Amp Modeler models will want this to stop messing with Python scripts. Everyone else should probably try the pre-built “Gateway” distribution first — this repo is for people who need to build from source or want to contribute to the iPlug2 plumbing.