Research map / Video / Video generation
HarmoniDPO: Video-guided Audio Generation via Preference-Optimized Diffusion
Wenshuo Peng, Kaipeng Zhang
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 online-DPO alignment (ablation) | HarmoniDPO · VGGSound | MKL / CLIP / CLAP: MKL 2.04 -> 1.86; CLIP 12.54 -> 13.27; CLAP 28.35 -> 31.75 | Improved |
| add online-DPO alignment (ablation) | HarmoniDPO · VGGSound | FD: 7.72 -> 6.54 | Improved |
| add Dual-scale Diffusion Search (DDS) at inference with N=4 (ablation) | HarmoniDPO aligned · VGGSound | CLAP / FD / FAD: CLAP 31.75 -> 32.57; FD 6.54 -> 6.42; FAD 1.64 -> 1.59 | Improved |
| increase online-DPO candidates from 2 to 8 (ablation) | HarmoniDPO · VGGSound online-DPO candidates (no DDS) | FID: 8.34 -> 6.54 | Improved |
| increase online-DPO candidates from 8 to 12 (ablation) | HarmoniDPO · VGGSound online-DPO candidates (no DDS) | FID / CLAP: FID 6.54 -> 6.67; CLAP 31.75 -> 32.15 | Mixed |
| use InternVid video features instead of CLIP average pooling (ablation) | HarmoniDPO base · VGGSound video-condition ablation (base model w/o alignment) | FID: 38.93 -> 32.16 | Improved |
| use InternVid video features instead of frame features only (ablation) | HarmoniDPO base · VGGSound video-condition ablation (base model w/o alignment) | FID: 38.35 -> 32.16 | Improved |
| add image frame features to video features (ablation) | HarmoniDPO base · VGGSound video-condition ablation (base model w/o alignment) | MKL / FID: MKL 1.894 -> 1.379; FID 39.45 -> 32.16 | Improved |
| HarmoniDPO versus FoleyCrafter on temporally hard clips | HarmoniDPO · AVSync15 | Onset ACC: 28.48 -> 32.53 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FoleyCrafter — FoleyCrafter: Bring Silent Videos to Life with Lifelike and Synchronized Sounds (IJCV 2026) · better on 10 of 10 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PrismAudio: Decomposed Chain-of-Thought and Multi-dimensional Rewards for Video-to-Audio Generation (ICLR 2026 · similarity 0.91)
- Seeing and Hearing: Open-domain Visual-Audio Generation with Diffusion Latent Aligners (CVPR 2024 · similarity 0.89)
- MMDisCo: Multi-Modal Discriminator-Guided Cooperative Diffusion for Joint Audio and Video Generation (ICLR 2025 · similarity 0.89)
- Model-Guided Dual-Role Alignment for High-Fidelity Open-Domain Video-to-Audio Generation (NeurIPS 2025 · similarity 0.88)
- Read, Watch and Scream! Sound Generation from Text and Video (AAAI 2025 · similarity 0.88)
- Tri-Ergon: Fine-Grained Video-to-Audio Generation with Multi-Modal Conditions and LUFS Control (AAAI 2025 · similarity 0.88)
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.
