Shortcut

Shortcut

Abstract

Independent musicians increasingly bear the burden of both creating music and producing the content needed to build an audience. Studio sessions generate a wealth of authentic, compelling moments that are rarely captured in published content, largely because the time required to locate and edit them is prohibitive. ShortCut was developed as a Max/MSP-based tool to address this problem by automating the detection and extraction of potentially engaging moments from studio session footage using transcript-based keyword matching. The system employed OpenAI's Whisper automatic speech recognition model to generate session transcripts, which were then searched using a sliding-window n-gram algorithm against a scored keyword dictionary of over 596 phrases. Detected matches were ranked by score and displayed with video loop points, allowing the user to browse and export clips directly to Adobe Premiere Pro via a generated XML file. The project demonstrated a functional proof-of-concept pipeline capable of detecting and surfacing engaging moments from real studio session footage, while identifying ASR accuracy in music-heavy environments as the primary area for future development.

Publication Date

7-1-2026

Campus

Valencia (Spain) Campus

Keywords

automatic speech recognition; content creation; keyword detection; music production; n-gram; Max/MSP; video editing automation

Shortcut

Share

COinS