StoryEngine: A State-Grounded Agentic Framework for Video Storytelling
StoryEngine tracks explicit story-world state so multi-shot video stays narratively and visually consistent.
Agentic multi-shot video systems often rely on text plans or prior pixels and fail to carry story consequences between shots. StoryEngine keeps a structured world state for entities and events, builds canonical references, and compiles state and visual constraints into render plans. A bounded evaluation-guided repair loop corrects local inconsistencies before they spread. On a new benchmark spanning scenarios and visual styles, it outperforms prior methods on storytelling quality, narrative coherence, and visual consistency.
- Tracks entity placement and story-relevant world state across shots
- Separates authoritative plans from unreliable visual observations
- A repair loop stops local visual errors from propagating
- Benchmark scores storytelling, coherence, and visual consistency
From the original article
1 min read at huggingface.co
Despite recent progress in agentic multi-shot video generation, producing coherent and consistent long-form stories remains challenging. Existing agentic pipelines typically rely on textual shot plans or previously generated pixels, yet lack an explicit mechanism for propagating the consequences of story events and maintaining the video world state across shots.
By Yingrui Wang, Zeqing Wang, Yeying Jin · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.