Source Extraction Demo (Slakh)

MGE-LDM: Simultaneous Music Generation and Extraction via Joint Latent Modeling

Source Extraction Samples

Mixture GT MSDM[1] MSG-LD[2] MGE-LDM (ours)
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel
Mel Bass Mel Mel Mel Mel
Drums Mel Mel Mel Mel
Guitar Mel Mel Mel Mel
Piano Mel Mel Mel Mel

[1] Giorgio Mariani, et al., "Multi-Source Diffusion Models for Simultaneous Music Generation and Separation."
ICLR 2024. [Paper]
[2] Tornike Karchkhadze, et al., "Simultaneous music separation and generation using multi-track latent diffusion models."
ICASSP 2025. [Paper]