NeuralPluginTemplate

An audio plugin template based on JUCE to load an ONNX AI model.

• it uses the dynamic lib of ONNX runtime compliled for MACOS.
• the current Projucer is aimed to deploy AU and VST3 for MAC OS only.

The current controls include the input/output gain and mix level, and also a phase flip check box.

There are 2 basic level meter for input and output.

Quick instructions

• place the ONNX Model under the folder ./model

• at the constructor of PluginProcessor.cpp

  • edit the file path e.g : auto model_file = bundle.getChildFile ("Resources/model/waveunet_distort.onnx");
  • and also it's intended blocksize

• Modify OnnxModel.cpp 's process() for model's input/outputs processing.

• the current onnxmodel is a noise producing model sample.

• this repo contains a version 1.14.1 ONNX runtime library for MACOS, which the projucer is configured to run RPath with it.

Notes

• the process is currently mixed down to mono for the ML model's input.
• Do modify both OnnxModel.cpp and PluginProcessor.cpp if your model needs stereo input.

Related

• See https://github.com/cpoohee/MLPluginTemplate for the Machine learning audio template.