Back to versions

StreamMUSE v0

Real-Time Language Model Jamming: A Case Study for Live Music Accompaniment Generation

This work was accepted by RTAS 2026.

v0

Version Summary

What this version contributes to the StreamMUSE project.

Frames the task as real-time accompaniment generation conditioned on an external musical stream.

Uses a client-server architecture where the client sends high-frequency requests and schedules returned accompaniment.

Studies the interaction between inference interval, generation length, round-trip latency, and music quality.

Recorded demos

Real-time

A human performer plays the melody, while the system generates accompaniment in real time.

v000:40Real-time accompaniment

Real-time demo 1

v000:31Real-time accompaniment

Real-time demo 2

v000:16Real-time accompaniment

Real-time demo 3

v001:20Real-time accompaniment

Real-time demo 4

Playable assets

MIDI Examples

MIDI examples can be played directly in the browser or downloaded.

v0 MIDI accompaniment example

Playable placeholder MIDI for validating in-page playback and download.

00:05Accompaniment output
Download

Publication

Paper Information

Real-Time Language Model Jamming: A Case Study for Live Music Accompaniment Generation

2026 IEEE 32nd Real-Time and Embedded Technology and Applications Symposium (RTAS)

Bowen Zheng, Andrew H. Yang, Jiaqi Ruan, Jia He, Xinyue Li, Yuan-Hsin Chen, Ziyu Wang, Xiaosong Ma

DOI: 10.1109/RTAS68450.2026.00032

@inproceedings{zheng2026realtime,
  title = {Real-Time Language Model Jamming: A Case Study for Live Music Accompaniment Generation},
  author = {Zheng, Bowen and Yang, Andrew H. and Ruan, Jiaqi and He, Jia and Li, Xinyue and Chen, Yuan-Hsin and Wang, Ziyu and Ma, Xiaosong},
  booktitle = {2026 IEEE 32nd Real-Time and Embedded Technology and Applications Symposium (RTAS)},
  year = {2026},
  doi = {10.1109/RTAS68450.2026.00032}
}