A Web-Based DAW for AI-Generated Music Workflow
摘要
The rapid advancement of artificial intelligence has led to significant breakthroughs in AI-generated music, enabling the production of high-quality audio at 48kHz sampling rates. However, a fundamental challenge persists in current AI music generation technologies: while end-to-end audio generation models produce high-fidelity music, their black-box nature makes musical elements (e.g., pitch, duration) difficult to edit. Conversely, symbolic generation methods (e.g., MIDI) retain structured and editable representations of music, yet rely heavily on professional post-production to achieve high-fidelity audio quality. To address this gap, this paper introduces a user-editable web-based Digital Audio Workstation (DAW) specifically designed for music generated by symbolic generation methods. The main contributions include: an intuitive visual interface supporting direct manipulation of musical elements, a browser-based real-time audio renderer, and a standardized API that integrates symbolic music generation systems, enabling a seamless end-to-end workflow from AI-generated content to user refinement. Experimental results demonstrate that the sound quality is improved and nearly 97% of users have a positive intention to use our DAW.