drewmerc302/nam-a2a1-converter

{"i":1234,"repo":"drewmerc302/nam-a2a1-converter","stars":2,"lang":"Python","ok":true,"char_count":1571,"text":"This is a desktop application that converts one type of guitar amplifier model file into another. Specifically, it takes captures made in the newer A2 format from the Neural Amp Modeler project and produces A1 format files, which are what devices like the Valeton GP-5 and GP-50 pedalboards accept. The two formats use different neural network architectures, so you cannot just relabel the file. The conversion works by a process called distillation: the application plays a standardized test signal through the A2 model, records the result, then trains a fresh A1 model to reproduce that same output.\n\nThe result is a new A1 file that closely matches the behavior of the original A2 capture. Because the A2 model is deterministic and free of noise from a physical amplifier, the distilled A1 copy is typically a tighter match to the A2 than a real amp capture would be to its source amp. In practice, you load the resulting file onto your pedal, and it behaves like the original capture.\n\nYou can download prebuilt versions for macOS, Windows, and Linux with no Python installation required. The macOS build is signed and notarized, so it opens cleanly. The Windows build is unsigned, which means you will see a SmartScreen warning and need to click through it. The Linux build runs on recent distributions and bundles its own dependencies, so there is nothing extra to install.\n\nTraining the new model is much faster if your computer has a GPU. The macOS app already uses Apple silicon graphics by default. On Windows with an NVIDIA card, there is a separate CUDA build that can cut a conversion from about 40 minutes down to a small fraction of that time. The app checks your hardware on launch and tells you if a faster build is available.\n\nUsing it is straightforward: launch the app, drop in one or more A2 files, pick a quality preset, and click convert. A progress bar shows how long is left, and you can cancel at any point. The finished A1 files appear in a folder on your computer, ready to import into your device. The project is built on the open source Neural Amp Modeler and gives credit to related community work on the older format.","error":null}