Research map / Image generation / Image restoration
SonicMaster: Towards Controllable All-in-One Music Restoration and Mastering
Jan Melechovsky, Ambuj Mehrish, Abhinaba Roy, Dorien Herremans
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add text conditioning (ablation) | SonicMaster Large · EQ Boom, snippet | average absolute error: 0.1432 -> 0.0834 | Improved |
| use correct prompts instead of shuffled prompts (ablation) | SonicMaster Large · EQ Boom, snippet | average absolute error: 0.2075 -> 0.0834 | Improved |
| use Euler 1 step instead of default ODE solver (ablation) | SonicMaster Large · EQ Boom, snippet | average absolute error: 0.0834 -> 0.1146 | Hurt |
| use Euler 100 steps instead of default solver (ablation) | SonicMaster Large · EQ Boom, snippet | average absolute error: 0.0834 -> 0.1540 | Hurt |
| use Runge-Kutta 10 steps instead of default solver (ablation) | SonicMaster Large · EQ Boom, snippet | average absolute error: 0.0834 -> 0.0810 | No clear effect |
| use 4/12 blocks instead of 6/6 (more DiT blocks) (ablation) | SonicMaster Medium · EQ Boom, snippet | average absolute error: 0.2766 -> 0.0698 | Improved |
| scale model from Small (2/6) to Large (6/18) (ablation) | SonicMaster · EQ Boom, snippet | average absolute error: 0.0819 -> 0.0834 | No clear effect |
| remove audio conditioning (basic) (ablation) | SonicMaster Large · EQ Boom, snippet | average absolute error: 0.0834 -> 0.0658 | Mixed |
| use 15s audio conditioning instead of 10s (ablation) | SonicMaster Large · EQ Muddy, snippet | average absolute error: 0.0388 -> 0.0320 | Improved |
| apply audio conditioning during inference (ablation) | SonicMaster Large · EQ Boom, snippet | average absolute error: 0.0834 -> 0.0840 | No clear effect |
| use SonicMaster instead of degraded input | SonicMaster · Snippet, all degradations | PQ: 7.026 -> 7.705 | Improved |
| use SonicMaster instead of degraded input | SonicMaster · Snippet, all degradations | SSIM: 0.777 -> 0.609 | Hurt |
Closest papers by meaning
- MuseControlLite: Multifunctional Music Generation with Lightweight Conditioners (ICML 2025 · similarity 0.86)
- Customized Condition Controllable Generation for Video Soundtrack (CVPR 2025 · similarity 0.84)
- SonicSim: A customizable simulation platform for speech processing in moving sound source scenarios (ICLR 2025 · similarity 0.83)
- Zero-Shot Unsupervised and Text-Based Audio Editing Using DDPM Inversion (ICML 2024 · similarity 0.83)
- DITTO: Diffusion Inference-Time T-Optimization for Music Generation (ICML 2024 · similarity 0.83)
- Audio Super-Resolution with Latent Bridge Models (NeurIPS 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
