Neural Amp Modeler
neuralampmodeler.com
- Free plan
- Windows
- macOS
- Linux
- Web
What Neural Amp Modeler is
Neural Amp Modeler is a free project for creating digital models of guitar amplifiers and pedals using deep learning. Musicians can train models online in Google Colab or install the Python-based trainer locally from PyPI. The Gateway plugin loads and plays snapshot models of gear, while ParametricOD shows modeling across an overdrive’s knobs and switches. The plugin repository produces VST3 and AU plugins, as well as a standalone desktop application; an impulse-response loader is also listed. Users can share models through TONE3000, a community-organized online library. NAM has three open-source code repositories under permissive licenses, and contributions are welcome. It runs on Windows, macOS, Linux, Raspberry Pi, embedded systems and websites. Existing A1 models remain supported in NeuralAmpModelerCore and NeuralAmpModelerPlugin. Builders can incorporate NAM’s modeling technology into their own products. The training code depends on PyTorch Lightning; version 0.12.3 excludes compromised Lightning versions 2.6.2 and 2.6.3.
Who it suits
NAM suits guitarists and audio creators who want to train, use or share amplifier and pedal models. Its open repositories and modeling technology may also suit builders developing their own products.
Good to know, and what to watch
- Free project with three permissively licensed repositories
- Train online in Google Colab or locally from PyPI
- Creates VST3, AU and standalone applications
- Models can be shared through TONE3000
- Training code depends on PyTorch Lightning
- Listed plugin platforms are Windows, macOS and Linux
Plans
Every plan and priceNeural Amp Modeler
Free plan
Getting started
- Open the Neural Amp Modeler website
- Train a model in Google Colab or install the Python trainer from PyPI
- Install the VST3, AU, or standalone desktop application
- Load snapshot models with Gateway or share models through TONE3000
The Club's review
Neural Amp Modeler is a deep-learning toolkit for capturing the behavior of guitar amplifiers and pedals, then playing those captures in a range of setups. It is best for guitarists willing to train or exchange models and for developers building their own modeling products. The trade-off is a technical workflow: this is more adaptable than a preset-only amp simulator, but less immediate for players who simply want ready-made sounds.
Overview
NAM brings model creation, playback and sharing into one open-source project. Training can happen online in Google Colab or locally with its Python trainer from PyPI, giving technically inclined users a choice between a hosted notebook and an installed workflow. Finished models can be shared through TONE3000, a community-organized library, so the platform is suited to collaboration as well as private experimentation.
Three code repositories are released under permissive open-source licenses and accept contributions. That makes NAM unusually relevant to builders: its modeling technology can be incorporated into other products rather than used only through the supplied applications. Existing A1 models continue to work with NeuralAmpModelerCore and NeuralAmpModelerPlugin, which protects earlier model investments as the project evolves.
Key features
Training and modeling
- Online or local trainer: Google Colab removes the need to set up a local environment, while the PyPI package supports users who want local control. Both routes require more technical involvement than selecting a commercial amp-sim preset.
- Deep-learning capture: NAM is designed to model amplifiers and pedals with high accuracy. ParametricOD extends that idea across an overdrive's complete range of knobs and switches, making it useful when a single static snapshot is not enough.
Playback and sharing
- Gateway plugin: Gateway loads and plays snapshot models of favorite gear, providing a focused way to use captures after training.
- Plugin and standalone formats: The plugin repository produces VST3 and AU plugins alongside a standalone desktop application. An integrated impulse-response loader is useful for setups that combine amp models with cabinet responses.
- Community library: TONE3000 gives users a place to exchange models, extending NAM beyond whatever one person captures.
Builder access and security
Support for websites, Raspberry Pi and embedded systems broadens NAM beyond a conventional desktop plugin. Builders can contact [email protected] for help integrating the technology. Training depends on PyTorch Lightning; version 0.12.3 excludes compromised Lightning versions 2.6.2 and 2.6.3, an important consideration when maintaining a training environment.
Pricing
Neural Amp Modeler — free, free. The project is open source and under continuous development, with no paid tier, seat count or usage quota attached to the stated plan. That makes it a strong fit for hobbyists, educators, contributors and builders who need an inspectable foundation without a licensing bill. The cost advantage does not remove the setup work: users choosing local training still need to manage a Python-based workflow, while cloud training runs through Google Colab.
Platforms
NAM supports Windows, macOS and Linux, with a standalone application and plugin formats for desktop use. The project also runs on the web, Raspberry Pi and embedded systems, giving developers options that typical desktop-only amp software does not. Website and embedded support is particularly relevant for product teams; players choosing a straightforward mobile amp app will find the supported systems less focused on phones and tablets.
Who it's for
NAM is the right choice for guitarists who want to capture their own amplifiers or pedals, explore community models, or keep control of the modeling chain. It also suits programmers and hardware builders who need permissively licensed technology they can adapt. It is not the best first stop for someone who wants a polished, preset-led experience with minimal setup, because training is organized around Colab or Python rather than an entirely in-app workflow.
Pros and cons
- Pro — No purchase cost: The free, open-source plan removes a financial barrier for players and developers.
- Pro — Broad deployment: Windows, macOS, Linux, web, Raspberry Pi and embedded support can accommodate both studio users and product builders.
- Pro — Flexible creation and exchange: Local or Colab training, TONE3000 sharing and A1 model compatibility support long-term experimentation.
- Con — Technical entry point: Python installation or a Google Colab workflow is less approachable than a conventional preset browser.
- Con — Snapshot-oriented playback: Gateway is built around snapshot models, while the deeper parametric example is demonstrated through ParametricOD rather than presented as a universally simple control scheme.
Alternatives
Choose GENOME instead if you want a commercial macOS, Windows or iOS product with a free introductory tier and a paid one-time edition that bundles amplifiers, pedals, cabinets and studio effects.
Helix Stadium Native is a better fit for players who prefer a time-unlimited free edition with a defined set of amps, cabinets and effects on macOS or Windows.
Pick Amped for a conventional freemium amp application, including a free Fluff amp and desktop plugin formats, with a paid full license.
Amp Locker suits users who want a free download and the option to buy extra modules outright on Linux, macOS or Windows.
AmpForge is another free, open-source plugin for Linux, Windows and macOS.
Choose Amplifikation 360 for Kuassa's freemium approach to amp software on macOS and Windows.
TH-U is the alternative for buyers seeking a paid commercial product with a free trial on iOS, macOS and Windows.
Amp ONE is a free iPad-focused option.
Verdict
Neural Amp Modeler is the strongest choice for guitarists and builders who value accurate custom captures, open code and wide deployment more than a turnkey preset library. Choose it for the combination of free access, model training and sharing; look elsewhere if your priority is an immediately guided amp-sim experience with a fixed catalog and minimal technical setup.
The verdict
Neural Amp Modeler brings model training, playback and sharing together in a free, open-source project. Its support for several operating systems and plugin formats gives guitarists multiple ways to use it.
Questions people ask
How much does Neural Amp Modeler cost?
The listed Neural Amp Modeler plan is free, with a free plan.
Which operating systems does it support?
The listed supported platforms are Windows, macOS, and Linux; NAM also runs on Raspberry Pi, embedded systems, and websites.
Can I train models locally?
Yes. The Python-based trainer can be installed locally from PyPI, or training can run online in Google Colab.
Which plugin formats are available?
The plugin repository creates VST3 and AU plugins, and it also provides a standalone desktop application.
Can existing A1 models be used?
Yes. Existing A1 models remain supported in NeuralAmpModelerCore and NeuralAmpModelerPlugin.
Where can models be shared?
Users can share models through TONE3000, a community-organized online library.
What the maker publishes
- A2 compatibility
- Existing A1 models remain supported in NeuralAmpModelerCore and NeuralAmpModelerPlugin. source
- Builder integration
- Builders can use NAM's open-source modeling technology in their own products. source
- Gateway plugin
- Gateway loads and plays snapshot models of users' favorite gear. source
- Model sharing
- Users can share models through TONE3000, a community-organized online library. source
- Model trainer
- Users can train models online in Google Colab or install the Python-based trainer locally from PyPI. source
- Open source
- NAM has three code repositories available under permissive open-source licenses and open to contributions. source
- Parametric modeling
- ParametricOD demonstrates NAM parametric modeling across the full range of an overdrive's knobs and switches. source
- Plugin formats
- The plugin repository creates VST3 and AU plugins plus a standalone desktop application. source
- Security notice
- NAM's training code depends on PyTorch Lightning, and version 0.12.3 excludes compromised Lightning versions 2.6.2 and 2.6.3. source
- Support
- Builders seeking help can contact [email protected]. source
- Supported systems
- NAM runs on Windows, macOS, Linux, Raspberry Pi, embedded systems, and websites. source
- What it does
- Neural Amp Modeler uses deep learning to create models of guitar amplifiers and pedals with state-of-the-art accuracy. source
Checked against the maker's pages on 1 Oct 2026.