HARP 3.0: Generalizing I/O and API Support for Machine Learning in Digital Audio Workstations

Frank Cwitkowitz, Christodoulos Benetatos, Qixin Deng, Huiran Yu, Nathan Pruyne, Patrick O'Reilly, Hugo Flores García, Zhiyao Duan, Bryan Pardo

Keywords: digital audio workstations, deep learning, workflow

TL;DR: We expand upon HARP to support models with arbitrary inputs/outputs and alternative APIs, and introduce new features that improve workflow.

Abstract: HARP is a framework that integrates deep learning models into digital audio workstations (DAWs) through hosted, asynchronous, remote processing, allowing users to access state-of-the-art models without leaving their production environment. In HARP 3.0, we introduce generalized input–output handling for audio, MIDI, and text; expand API compatibility beyond Gradio to include providers such as Stability AI; and add workflow features including a media clipboard for storing and reusing processed files. These improvements make HARP 3.0 a more versatile and extensible platform for creative audio applications. HARP is open source and available for download at https://github.com/TEAMuP-dev/HARP.

Track: Demo Track

Confirmation: Demo Track: I confirm that I have followed the formatting guideline and included all author names and affiliations.

Submission Number: 72