TorchDrum Plugin

torchdrum is an audio plugin that has audio input and output. It is a real-time drum synthesizer that receives audio as input, which is used to trigger and control the synthesizer.

The development of torchdrum is based on research on Real-time Timbre Remapping with Differentiable DSP. For more information on this research please visit the accompanying website.

To train new models to run in this plugin see the accompanying repo.

Pre-Built Plugins

See the releases for pre-built plugin installers which include some pre-trained models. Currently only includes Apple Silicon builds.

Building from Source

Clone project and pull submodules. Includes JUCE.

git clone --recurse-submodules git@github.com:jorshi/torchdrum-plugin.git
cd torchdrum-plugin

Install TorchLib dependency. We recommend doing Python. Recommended to create a new virtual environment. Instructions below use pyenv:

First, install pyenv

Install python 3.10:

pyenv install 3.10
pyenv shell 3.10

Install python dependencies:

python -m venv venv
source venv/bin/activate
python -m pip install --upgrade pip
python -m pip install torch==2.1.1 --index-url https://download.pytorch.org/whl/cpu
python -m pip install "numpy<2.0"

Save the TORCHLIB_PATH from Python

export TORCHLIB_PATH=`python3 -c 'import torch;print(torch.utils.cmake_prefix_path)'`

Build with cmake

cmake -Bbuild -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH=$TORCHLIB_PATH
cmake --build build --config Release --parallel 4

There is a warning, please ignore:

CMake Warning at venv/lib/python3.9/site-packages/torch/share/cmake/Torch/TorchConfig.cmake:22 (message):
  static library kineto_LIBRARY-NOTFOUND not found.
Call Stack (most recent call first):
  venv/lib/python3.9/site-packages/torch/share/cmake/Torch/TorchConfig.cmake:127 (append_torchlib_if_found)
  CMakeLists.txt:10 (find_package)

If it builds correctly, expect to see:

[  a%] Built target juce_vst3_helper
[  b%] Built target TorchDrum
[  c%] Built target TorchDrum_Standalone
[  d%] Built target TorchDrum_VST3
[100%] Built target TorchDrum_AU

It will replace any existing version in user audio plugin folders.

Future Development Ideas

• dry/wet
• info on model
• feedback for onset detection
• visualization of the decaying frequencies