HARP is a sample editor that allows for hosted, asynchronous, remote processing of audio with machine learning. HARP works by routing audio through Gradio endpoints. Since Gradio applications can be hosted locally or in the cloud (e.g. with HuggingFace Spaces), HARP lets users of Digital Audio Workstations (DAWs) capable of connecting with external sample editors (e.g. REAPER, Logic Pro X, or Ableton Live) access large state-of-the-art models using cloud-based services, without breaking the within-DAW workflow.
For a more formal introduction, please see our NeurIPS paper presenting an earlier version of HARP.
HARP has been tested on the following operating systems: | OS | ||||
---|---|---|---|---|---|
Version(s) | 13.0, 13.4, 14.2.1 | 10.15 | 10, 11 | 22.04 |
Download the macOS ZIP file for HARP from the releases page.
Unzip and double click on the DMG file. This will open the window shown below.
Drag HARP.app
to the Applications/
folder to install HARP.
Do NOT run harp from the installation window above. It will not run correctly if you do so.
Download the Windows ZIP file for HARP from the releases page.
Extract the contents of the ZIP file and move the directory containing HARP.exe
to a location of your choice, e.g. C:\Program Files
.
Download the Linux ZIP file for HARP from the releases page.
Extract the contents of the ZIP file and move the directory containing HARP
to a location of your choice, e.g. /usr/local/bin/
.
To work as a standalone application, you just need to open HARP and start using it.
Run HARP.app
to start the application.
Run HARP.exe
to start the application.
Run HARP
to start the application.
To set up HARP for use from within Reaper, do the following.
Choose REAPER > Preferences on the file menu for MacOS. OR hit keyboard command control+p on Windows
Scroll down to External Editors and click Add.
Click Browse to the right of the Primary Editor field.
Navigate to your HARP installation (e.g. HARP.app
) and select "OK".
(Optional) Right click the audio for the track you want to process and select Render items as new take.
Right click the audio and select Open items in editor > Open items in 'HARP.app'.
To set up HARP for use from within Logic Pro, do the following.
HARP.app
as an external sample editor following this guide.(Optional) Right click the audio for the track you want to process and select Bounce in Place.
Select any audio region and press Shift+W to open the corresponding audio file in HARP.
HARP can be used to apply deep learning models to your audio either as a stand-alone or within any DAW (e.g. Logic Pro) that supports external sample editors.
If you use it stand-alone, just load a file, load a model and apply the model to the audio file. When you're happy with the result, save the output.
In a DAW, you select the exceprt you want to process, open it in HARP, process it, and select Save from the File menu in HARP. This will return the processed file back to the DAW.
HARP is a destructive file editor. When you select Save, HARP overwrites the existing audio. After recording or loading audio into a track within your preferred DAW, it is recommended that you bounce-in-place (on Logic) or Render items as new take (on Reaper) the audio before processing it with HARP. In this way, you will avoid overwriting the original audio file and will be able to undo any changes introuced by HARP. Alternatively, overwriting can be circumvented by using the Save As functionality from the File menu in HARP.
If you would like to process only an excerpt of a track, first trim the exceprt to the portion you want to process. Then, perform a bounce-in-place of the excerpt. This will make a new file that contains only the audio you want to process with HARP. Then, open the new file in HARP.
While any algorithm or deep learning model can be deployed to HARP using the PyHARP API, at present, the following models have been made available:
Pitch Shifting: hugggof/pitch_shifter
Harmonic/Percussive Source Separation: hugggof/harmonic_percussive
Music Audio Generation: descript/vampnet
Convert Instrumental Music into 8-bit Chiptune: hugggof/nesquik
Select or type the Gradio endpoint of an available HARP-ready model (e.g. "hugggof/harmonic_percussive") in the field at the top of the application.
This will populate the window with controls for the model.
Loading can take some time if the HuggingFace Space is asleep.
Adjust the model controls to your liking and click process.
The resulting audio can be played by pressing the space bar or by clicking Play/Stop.
You can save the audio at any time using the Save and Save As features located in the File menu.
Any changes made in HARP will be automatically reflected in your DAW.
To make a model that you can use in HARP, see Adding Models with PyHARP. To modify the HARP app, see Building Harp.
The current versions of HARP and PyHARP are shown below. They are fully compatible.
HARP | PyHARP |
---|---|
1.3.0 | 0.1.1 |
1.2.0 | 0.1.0 |
We provide PyHARP, a lightweight API to build HARP-compatible Gradio apps with optional interactive controls. PyHARP allows machine learning researchers to create DAW-friendly user interfaces for virtually any audio processing code using a minimal Python wrapper.
HARP can be built from scratch with the following steps:
git clone --recurse-submodules https://github.com/TEAMuP-dev/HARP
cd HARP/
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Debug
The OSX architecture for the build can be specified explicitly by setting CMAKE_OSX_ARCHITECTURES
to either arm64
or x86_64
:
cmake .. -DCMAKE_OSX_ARCHITECTURES=x86_64
Ensure your system satisfies all JUCE dependencies.
make -j <NUM_PROCESSORS>
cmake --build . --config Debug -j <NUM_PROCESSORS>
To debug your HARP build in Visual Studio Code it is helpful to do the following.
launch.json
).{
"version": "0.2.0",
"configurations": [
{
"name": "(lldb) Standalone",
"type": "cppdbg",
"request": "launch",
"program": "${workspaceFolder}/build/HARP_artefacts/Debug/HARP.app", // macOS
// "program": "${workspaceFolder}/build/HARP_artefacts/Debug/HARP.exe", // Windows
// "program": "${workspaceFolder}/build/HARP_artefacts/Debug/HARP", // Linux
"args": ["../test.wav"],
"cwd": "${fileDirname}",
"MIMode": "lldb" // macOS
}
]
}
-DCMAKE_BUILD_TYPE=Debug
.Codesigning and packaging for distribution is done through the script located at packaging/package.sh
.
You'll need to set up a developer account with Apple and create a certificate in order to sign the plugin.
For more information on codesigning and notarization for macOS, please refer to the pamplejuce template.
The script requires the following variables to be passed:
# Retrieve values from either environment variables or command-line arguments
DEV_ID_APPLICATION # Developer ID Application certificate
ARTIFACTS_PATH # should be packaging/dmg/HARP.app
PROJECT_NAME # "HARP"
PRODUCT_NAME # "HARP"
NOTARIZATION_USERNAME # Apple ID
NOTARIZATION_PASSWORD # App-specific password for notarization
TEAM_ID # Team ID for notarization
Usage:
./HARP/packaging/package.sh <DEV_ID_APPLICATION> <ARTIFACTS_PATH> <PROJECT_NAME> <PRODUCT_NAME> <NOTARIZATION_USERNAME> <NOTARIZATION_PASSWORD> <TEAM_ID>
After running package.sh
, you should have a signed and notarized dmg file in the packaging/
directory.
If you use HARP in your research, please cite our NeurIPS paper:
@inproceedings{garcia2023harp,
title = {{HARP}: Bringing Deep Learning to the DAW with Hosted, Asynchronous, Remote Processing},
author = {Garcia, Hugo Flores and O’Reilly, Patrick and Aguilar, Aldo and Pardo, Bryan and Benetatos, Christodoulos and Duan, Zhiyao},
year = 2023,
booktitle = {NeurIPS Workshop on Machine Learning for Creativity and Design}
}