nn~ is a Max/PureData external that bridges trained neural audio models (RAVE, vschaos2) into a patching environment

nn~ (from IRCAM’s ACIDS group) is an external object for Max 8 and PureData that loads pre-trained neural audio synthesis models and exposes them as audio objects in the patcher. It is only a bridge: you must separately download a compatible model (RAVE or vschaos2, in .ts TorchScript format). Installation on Max: unarchive the platform-specific nn~ archive and place the folder in Max’s Packages folder (or elsewhere, then register the path in Max File Preferences); on PureData, place it in the externals folder and clear the macOS quarantine. Models must live somewhere accessible to Max’s file preferences (or, in Pd, alongside the external). This lets live coders run neural synthesis models in real-time audio patches without Python or ML infrastructure.

Examples

Download the ‘vintage’ RAVE .ts model; drop the nn~ package into Max 8’s Packages folder; place vintage.ts where Max file preferences can see it. The nn~ object is now available in the patcher.

Assessment

What is the minimum setup required to run a RAVE model through Max 8? Which two artifacts must be installed and where must each be placed?