Researchers have explored the use of synthetic data to improve controllable human-centric video generation, addressing the limitations of real-world data scarcity and privacy concerns. Their diffusion-based framework allows for fine-grained control over appearance and motion, analyzing how synthetic data interacts with real data during training. Experiments revealed that synthetic and real data play complementary roles, offering insights into efficient synthetic sample selection to enhance realism, temporal consistency, and identity preservation in generated videos. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON This is a research paper exploring synthetic data augmentation for video generation.
