Research map / Video / Video generation
Taming Text-to-Sounding Video Generation via Advanced Modality Condition and Interaction
Kaisi Guan, Xihua Wang, Zhengfeng Lai, Xin Cheng, Peng Zhang, Xiaojiang Liu, Ruihua Song, Meng Cao
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 cross-modal interaction (DCA) to independent Wan+SDA towers | Wan+SDA · AVSync15 | AV-Align: 0.205 -> 0.275 | Improved |
| use disentangled CRR captions instead of a shared audio-video caption (ablation) | BridgeDiT · AVSync15 | FVD: 1362.83 -> 765.74 | Improved |
| use disentangled CRR captions instead of video-only shared caption (ablation) | BridgeDiT · AVSync15 | CLAP: 9.67 -> 35.95 | Improved |
| add CRR rewriting instead of raw LLM captions (ablation) | BridgeDiT · AVSync15 | FAD: 19.42 -> 5.34 | Improved |
| add semantic anchors to the rewriter (ablation) | CRR · AVSync15 | FAD: 15.74 -> 5.34 | Improved |
| add semantic anchors to the rewriter (ablation) | CRR · AVSync15 | AV-Align: 0.224 -> 0.275 | Improved |
| add inference-time prompt expansion (ablation) | BridgeDiT · AVSync15 | FVD: 1463.81 -> 765.74 | Improved |
| use dual cross-attention instead of full attention fusion (ablation) | BridgeDiT · AVSync15 | AV-Align: 0.253 -> 0.275 | Improved |
| use bidirectional DCA instead of A2V cross-attention only (ablation) | BridgeDiT · AVSync15 | AV-Align: 0.236 -> 0.275 | Improved |
| use bidirectional DCA instead of A2V cross-attention only (ablation) | BridgeDiT · AVSync15 | FVD: 746.37 -> 765.74 | Hurt |
| use uniform early-bias block placement instead of late layers (ablation) | BridgeDiT · AVSync15 | AV-Align: 0.1831 -> 0.2746 | Improved |
| add DCA to SANA+AudioLDM backbones | SANA+AudioLDM · AVSync15 | FAD: 17.09 -> 9.46 | Improved |
Closest papers by meaning
- VinTAGe: Joint Video and Text Conditioning for Holistic Audio Generation (CVPR 2025 · similarity 0.90)
- AVGen-Bench: A Task-Driven Benchmark for Multi-Granular Evaluation of Text-to-Audio-Video Generation (ICML 2026 · similarity 0.89)
- Tell What You Hear From What You See - Video to Audio Generation Through Text (NeurIPS 2024 · similarity 0.88)
- Diverse and Aligned Audio-to-Video Generation via Text-to-Video Model Adaptation (AAAI 2024 · similarity 0.88)
- JavisGPT: A Unified Multi-modal LLM for Sounding-Video Comprehension and Generation (NeurIPS 2025 · similarity 0.88)
- MMAudio: Taming Multimodal Joint Training for High-Quality Video-to-Audio Synthesis (CVPR 2025 · similarity 0.87)
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.
