论文解读
FlowLong: Inference-time Long Video Generation via Manifold-constrained Tweedie Matching
视频生成 | 4.9/10
视频生成
论文解读
PresentAgent-2: Towards Generalist Multimodal Presentation Agents
生成模型 | 6.5/10
会议任务专题
ICLR 2026 - 视频生成
共 2 篇 ICLR 2026 视频生成 方向论文
论文解读
InterActHuman: Multi-Concept Human Animation with Layout-Aligned Audio Conditions
视频生成 | 7.5/10
论文解读
Stable Video Infinity: Infinite-Length Video Generation with Error Recycling
视频生成 | 7.0/10
论文解读
InterActHuman: Multi-Concept Human Animation with Layout-Aligned Audio Conditions
视频生成 | 9.0/10
论文解读
Stable Video Infinity: Infinite-Length Video Generation with Error Recycling
视频生成 | 8.8/10
论文解读
Syncphony: Synchronized Audio-to-Video Generation with Diffusion Transformers
音频生成 | 7.5/10
会议任务专题
ICASSP 2026 - 视频生成
共 2 篇 ICASSP 2026 视频生成 方向论文
论文解读
MirrorTalk: Forging Personalized Avatars Via Disentangled Style and Hierarchical Motion Control
语音合成 | 7.0/10
论文解读
StyHarmo: Efficient Style-Specific Video Generation with Music Synchronization
视频生成 | 6.5/10
论文解读
VT-Heads: Voice Cloning and Talking Head Generation from Text Based on V-DiT
视频生成 | 6.5/10
论文解读
CoInteract: Physically-Consistent Human-Object Interaction Video Synthesis via Spatially-Structured Co-Generation
1. **问题**：现有视频扩散模型在生成人机交互（HOI）视频时，常出现手/脸结构崩溃和人机物理穿透等问题，根源在于模型缺乏对3D空间关系和交互结构的理解。 2. **方法核心**：提出CoInteract框架，核心是“空间结构化协同生成”范式。在一个共享的DiT骨干中联合训练RGB外观流和辅助的