SAMUeL: Efficient Vocal-Conditioned Music Generation via Soft Alignment Attention and Latent Diffusion
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915705805340672 |
|---|---|
| author | Cheung, Hei Shing Zhang, Boya Chan, Jonathan H. |
| author_facet | Cheung, Hei Shing Zhang, Boya Chan, Jonathan H. |
| contents | We present a lightweight latent diffusion model for vocal-conditioned musical accompaniment generation that addresses critical limitations in existing music AI systems. Our approach introduces a novel soft alignment attention mechanism that adaptively combines local and global temporal dependencies based on diffusion timesteps, enabling efficient capture of multi-scale musical structure. Operating in the compressed latent space of a pre-trained variational autoencoder, the model achieves a 220 times parameter reduction compared to state-of-the-art systems while delivering 52 times faster inference. Experimental evaluation demonstrates competitive performance with only 15M parameters, outperforming OpenAI Jukebox in production quality and content unity while maintaining reasonable musical coherence. The ultra-lightweight architecture enables real-time deployment on consumer hardware, making AI-assisted music creation accessible for interactive applications and resource-constrained environments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2507_19991 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SAMUeL: Efficient Vocal-Conditioned Music Generation via Soft Alignment Attention and Latent Diffusion Cheung, Hei Shing Zhang, Boya Chan, Jonathan H. Sound Machine Learning We present a lightweight latent diffusion model for vocal-conditioned musical accompaniment generation that addresses critical limitations in existing music AI systems. Our approach introduces a novel soft alignment attention mechanism that adaptively combines local and global temporal dependencies based on diffusion timesteps, enabling efficient capture of multi-scale musical structure. Operating in the compressed latent space of a pre-trained variational autoencoder, the model achieves a 220 times parameter reduction compared to state-of-the-art systems while delivering 52 times faster inference. Experimental evaluation demonstrates competitive performance with only 15M parameters, outperforming OpenAI Jukebox in production quality and content unity while maintaining reasonable musical coherence. The ultra-lightweight architecture enables real-time deployment on consumer hardware, making AI-assisted music creation accessible for interactive applications and resource-constrained environments. |
| title | SAMUeL: Efficient Vocal-Conditioned Music Generation via Soft Alignment Attention and Latent Diffusion |
| topic | Sound Machine Learning |
| url | https://arxiv.org/abs/2507.19991 |