Live Music Diffusion Models: Efficient Fine-Tuning and Post-Training of Interactive Diffusion Music Generators

Zachary Novack, Stephen Brade, Haven Kim, Hugo Flores García, Nithya Shikarpur, Chinmay Talegaonkar, Suwan Kim, Valerie K. Chen, Julian McAuley, Taylor Berg-Kirkpatrick, Cheng-Zhi Anna Huang

arXiv:2605.22717 · 2026-05-20 공개 · arXiv · PDF

diffusion-models post-training music-generation interactive-models streaming-generation generative-delay text-conditioned live-music

Abstract

Interactive streaming music generation promises the use of generative models for live performance and co-creation that is impossible with offline models. However, SOTA models exist in the discrete-AR regime, requiring industrial levels of compute for both training and inference. In this work, we investigate whether audio diffusion models, with their wide support in the open-source community but non-streaming bidirectional nature, can be repurposed efficiently into interactive models accessible on consumer hardware. By taking a critical look at the modern pipeline for block-wise outpainting diffusion, we identify critical inefficiencies during inference that result in strictly worse computational efficiency than their discrete-AR counterparts. We propose Live Music Diffusion Models (LMDMs), a simple modification of the generative diffusion process that recovers, and then outperforms, the inference complexity of the discrete Live Music Models (LMMs) through block-wise KV Caching. Unlike LMMs, LMDMs further enable stable post-training alignment through our novel ARC-Forcing paradigm, reducing error accumulation without any explicit RL or reward models. We demonstrate the application of LMDMs in a number of creative domains, including text-conditioned generation, sketch-based music synthesis, and jamming. We finally show how LMDMs can be used as a generative instrument in a real artist-AI collaboration, utilizing LMDMs as a "generative delay" to transform musicians' improvisation live for variable timbral effects while running locally on a consumer gaming laptop.

한국어 요약

한 줄 요약

**[Streaming Music Diffusion]** LMDM이 block-wise KV caching으로 audio diffusion의 inference 비효율을 회복해 discrete-AR LMM을 능가, ARC-Forcing post-training이 RL·reward model 없이 error accumulation 감소, 소비자 게이밍 노트북에서 라이브 연주 협업 지원.

핵심 기여도

핵심 아이디어

Streaming music diffusion의 효율 격차는 outpainting diffusion이 KV caching의 이점을 살리지 못해 발생하며, block-wise KV caching·ARC-Forcing post-training 두 단순 modification만으로 discrete-AR LMM의 계산 효율을 따라잡고 alignment까지 가능해 소비자 노트북의 라이브 연주 도구로 활용 가능하다.

기술적 접근법

주요 결과

의의 및 한계

**의의**: Streaming music diffusion의 효율 격차 해소, 오픈소스 커뮤니티에 친화적 audio diffusion 활용 경로 개척, ARC-Forcing이 RL 의존 없는 alignment 일반 도구, 라이브 음악 협업의 새 인터페이스. **한계**: Block-wise 처리의 chunk 경계 artifact, 매우 긴 generation에서의 누적 drift는 검증 필요, 소비자 노트북 자원의 latency 한계.

실용적 활용