Split-then-Merge framework enhances generative video composition

Researchers have introduced Split-then-Merge (StM), a new framework for generative video composition that addresses data scarcity by learning from unlabeled videos. StM dynamically separates foreground and background layers from existing videos and then self-composes them to learn realistic subject-scene interactions. The framework incorporates a transformation-aware training pipeline with multi-layer fusion and augmentation, along with an identity-preservation loss to maintain foreground quality. Experiments indicate that StM surpasses current state-of-the-art methods in both quantitative and qualitative evaluations. AI

IMPACT Enhances control and realism in generative video, potentially improving applications in content creation and media synthesis.

RANK_REASON The item is a research paper detailing a new framework for generative video composition. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• arXiv
• arXivLabs
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Özgür Kara
• ScienceCast
• Split-then-Merge (StM)

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →