Robust Personalized Human-AI Collaboration with SmartLooper

Sageev Oore, Finlay Miller, Chandramouli Shama Sastry, Sri Harsha Dumpala, Marvin F. da Silva, Daniel Oore, Scott C. Lowe

Keywords: human-AI co-creation, music improvisation, MIDI music generation, interactive performance, AI creative tools

TL;DR: We demonstrate SmartLooper, an AI system that improvises in real time with a musician by evolving loops in the performer’s own style, based on their own data, enabling fluid human–AI co-creation.

Abstract: This demonstration can be presented either as a live improvised performance or as a recorded video of one. Like most music, it is better listened to rather than read about. We developed an AI ``smart looping'' system for collaborative human-AI real-time performance. A dataset is created by the artist who will be using the system. In performance, the artist begins by improvising a short musical fragment---say, a bass groove---and the system starts riffing on it in their style, based on the data they provided. Sometimes it repeats a phrase exactly; other times it makes small changes---shifting the rhythm, altering harmonic progressions---but not too much at once. The artist plays over this evolving loop, adding new lines or textures as the system continues to develop its own variations. The result is a fluid, co-created improvisation. It has been demonstrated in multiple live performances.

Track: Demo Track

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

(Optional) Short Video Recording Link: https://osageev.github.io/projects/looper/

Submission Number: 90