Neural Amp Modeler

Neural Amp Modeler (NAM) uses deep learning to build models of guitar amplifiers and pedals. It is best for guitarists interested in creating or sharing gear models, and for builders who want open-source modeling technology. Its reach across plugin, standalone, web and embedded environments is unusual; training is a more technical path than choosing a preset.

Overview

NAM brings model training, playback and sharing together in a free, open-source project. Users can train online in Google Colab or install the Python-based trainer from PyPI, then share models through TONE3000, a community-organized online library. Three permissively licensed repositories are open to contributions, making NAM relevant to developers as well as players.

The Gateway plugin plays snapshot models of gear. The plugin repository also produces VST3 and AU plugins and a standalone desktop application, with impulse response loading supported. This gives players several ways to work with models, though the training workflow calls for comfort with either a cloud notebook or Python.

Key features

• Model training: Google Colab provides an online option, while the PyPI trainer can be installed locally. That flexibility suits users who want to make models, but is less direct than selecting a finished amp sound.
• Playback and formats: Gateway loads and plays snapshot models. VST3 and AU plugins, plus a standalone app, make the project usable in plugin-based or standalone setups.
• Parametric modeling: ParametricOD demonstrates modeling an overdrive across its full range of knobs and switches. It points beyond fixed snapshots toward modeling controls, though the example is specifically an overdrive.
• Community and compatibility: Users can share models through TONE3000, and existing A1 models remain supported in NeuralAmpModelerCore and NeuralAmpModelerPlugin. This helps connect shared material with established NAM components.
• Builder use: Builders can incorporate NAM's open-source modeling technology into their own products. That extends its value beyond a guitarist's personal setup.

Pricing

NAM costs free as an open-source project under continuous development. There are no paid tiers described, so its appeal is strongest for users who want model creation and playback without a software purchase. The trade-off is that users must work within the project's training and playback approach rather than selecting a priced package of fixed features.

Platforms

NAM supports Windows, macOS and Linux, as well as Raspberry Pi, embedded systems and websites. Its platform range makes it a strong fit for builders and users working beyond a conventional desktop. The plugin repository offers VST3 and AU formats and a standalone app; the broader platform support does not mean each format applies to every system.

Who it's for

NAM is a good choice for guitarists who want to train models of their own gear, share models, or play snapshot models through a plugin or standalone application. Developers and product builders also have a clear reason to consider it: the code is permissively licensed, open to contributions, and usable in their own products. Players seeking a preset-only experience or a turnkey, nontechnical training workflow should look elsewhere.

One security point matters for users installing the trainer: NAM's training code depends on PyTorch Lightning, and version 0.12.3 excludes compromised Lightning versions 2.6.2 and 2.6.3. Users should check that notice when choosing a training setup.

Pros and cons

• Pro: Free, permissively licensed repositories open the project to contributions and product integration, not just personal use.
• Pro: Google Colab and local PyPI training provide two routes for making models.
• Pro: Gateway, VST3 and AU plugins, a standalone app, and TONE3000 sharing cover distinct parts of a model-based workflow.
• Con: Training requires a cloud notebook or Python installation, which adds setup friction for players who only want ready-made sounds.
• Con: NAM's breadth is tied to a model-centered workflow; it is not presented as a fixed, turnkey collection of amp presets.

Alternatives

Guitar Amp Software is the category directory for comparing options across the field. Consider Amp Locker instead if you want a free download with Prestige 1950 amp, seven FX pedals and seven rack FX, and the option to purchase extra modules individually and own them outright. GENOME is another option, with a free Intro plan covering four TSM-Ai amplifiers, 14 pedals, five DynIR cabinets and four Studio FX; its paid plan is $89.99, paid once.

IK Multimedia ReSing is also in the directory. AmpForge may suit readers seeking a free open-source plugin for Linux, Windows and macOS. Gjalla is a free, open-source option under GPLv3 or later, while Soundshed Guitar is free and identifies its public repository license as AGPL-3.0.

Amped offers a free Roots plan with one amp, 5034 Fluff, for Windows and Mac as standalone, VST3, AU and AAX for Mac; its full license is €5,999, paid once. Amplifikation 360 is another freemium option for macOS and Windows.

Verdict

Choose Neural Amp Modeler if you want a free, open-source route to training, playing and sharing guitar gear models, or you are a builder looking to integrate the technology into a product. Its combination of model workflows and broad platform reach is the main draw. Look elsewhere if you want a straightforward preset-first amp simulator without a training setup.