Xinlu Liu
Music and audio AI researcher working on intelligent audio effects and automatic mixing.
About
I work on intelligent audio effects and automatic mixing systems.
My current work focuses on learning representations for audio effects, controlling effect transformations from audio examples, and building evaluation tools for music production workflows.
Recent Updates
- RelFx / FxEncoder Eval was accepted to ISMIR 2026.
- The ISMIR project page is available at relative-fx.github.io.
- The source release is available at TMEGalaxyAudioEffect/relfx-ismir2026-release.
Selected Projects
RelFx / FxEncoder Eval
Relative audio effect representation learning for retrieving, measuring, and controlling effect transformations from paired audio.
Intelligent Audio Production Resources
A structured index of open resources for intelligent audio effects, representations, mixing, mastering, and evaluation.
LC-beating
Low-latency beat and downbeat activation estimation for music analysis and production-oriented applications.
Publications
-
Beyond Dry References: Learning Relative Audio Effects Representations via Contrastive Distance Learning
ISMIR 2026. Project -
Improving Drum Source Separation with Temporal-Frequency Statistical Descriptors
ICME 2024. DOI -
Stripe-Transformer: Deep Stripe Feature Learning for Music Source Separation
EURASIP Journal on Audio, Speech, and Music Processing, 2023. DOI -
LC-Beating: An Online System for Beat and Downbeat Tracking using Latency-Controlled Mechanism
ICME 2023. DOI Code -
XBeat: A Hybrid CNN-Transformer Model for Beat and Downbeat Tracking
CSMT 2023, Springer LNEE 1268. Proceedings -
DiffTimb: Diffusion Models for Many-to-Many Timbre Transfer
CSMT 2023, Springer LNEE 1268. Proceedings