Research map / Image generation / Autoregressive and latent generation
BinauralGrad: A Two-Stage Conditional Diffusion Probabilistic Model for Binaural Audio Synthesis
Yichong Leng, Zehua Chen, Junliang Guo, Haohe Liu, Jiawei Chen, Xu Tan, Danilo Mandic, Lei He, Xiangyang Li, Tao Qin, sheng zhao, Tie-Yan Liu
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 |
|---|---|---|---|
| use two-stage (common then specific) instead of a single-stage diffusion model (ablation) | BinauralGrad · Binaural speech benchmark test | Wave L2 (x10^-3): 0.144 -> 0.128 | Improved |
| use two-stage instead of single-stage diffusion (ablation) | BinauralGrad · Binaural speech benchmark test | PESQ: 2.544 -> 2.759 | Improved |
| use two-stage instead of single-stage diffusion (ablation) | BinauralGrad · Binaural speech benchmark test | MRSTFT: 1.269 -> 1.278 | No clear effect |
| use two-stage instead of single-stage diffusion (ablation) | BinauralGrad · Binaural speech benchmark test | Phase L2: 0.877 -> 0.837 | Improved |
| add second stage on top of stage 1 output (ablation) | BinauralGrad · Binaural speech benchmark test | Wave L2 (x10^-3): 0.300 -> 0.128 | Improved |
| add first (common) stage output on top of physical warping (ablation) | BinauralGrad · Binaural speech benchmark test | Wave L2 (x10^-3): 0.725 -> 0.300 | Improved |
| condition second stage on golden average instead of synthesized stage 1 output (ablation) | BinauralGrad · Binaural speech benchmark test | PESQ: 2.759 -> 4.212 | Improved |
| replace diffusion single stage baseline WarpNet with BinauralGrad | BinauralGrad · Binaural speech benchmark test | Wave L2 (x10^-3): 0.157 -> 0.128 | Improved |
| BinauralGrad vs WarpNet on spatial perception | BinauralGrad · Binaural speech benchmark test | Spatial MOS: 3.77 -> 3.79 | No clear effect |
Closest papers by meaning
- BinauralFlow: A Causal and Streamable Approach for High-Quality Binaural Speech Synthesis with Flow Matching Models (ICML 2025 · similarity 0.90)
- Cyclic Learning for Binaural Audio Generation and Localization (CVPR 2024 · similarity 0.87)
- Both Ears Wide Open: Towards Language-Driven Spatial Audio Generation (ICLR 2025 · similarity 0.86)
- AV-GS: Learning Material and Geometry Aware Priors for Novel View Acoustic Synthesis (NeurIPS 2024 · similarity 0.86)
- Diffusion-based Realistic Listening Head Generation via Hybrid Motion Modeling (CVPR 2025 · similarity 0.85)
- Fast Timing-Conditioned Latent Audio Diffusion (ICML 2024 · similarity 0.85)
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.
