{"slug": "videococo-code-as-cot-for-physically-consistent-video-generation-via-an-agentic", "title": "VideoCoCo: Code-as-CoT for Physically-Consistent Video Generation via an Agentic Dual-Engine System", "summary": "Researchers introduced VideoCoCo, an agentic dual-engine framework that uses executable Blender code as a chain of thought to generate physically consistent videos. The system improved the OmniWeaving baseline from 0.475 to 0.558 on PhyGenBench and from 52.18 to 77.88 on VBench-2.0, achieving the best average score on both benchmarks.", "body_md": "arXiv:2607.27380v1 Announce Type: new\nAbstract: Text-to-video models have achieved remarkable visual quality, yet they still struggle to generate physically consistent dynamics because the temporal evolution of a scene must be inferred implicitly from a highly compressed text prompt. Existing chain-of-thought approaches introduce intermediate plans or visual states, but these representations are typically non-executable or temporally sparse, limiting their ability to instantiate and control the complete spatiotemporal process. To address this limitation, we introduce VideoCoCo, an agentic dual-engine framework in which executable Blender code serves as a process-level chain of thought. Given a text prompt, a coding agent synthesizes a Blender program that explicitly specifies the scene and its temporal evolution. The executable simulation engine runs the program to produce a deterministic spatiotemporal draft, which is subsequently transformed into a photorealistic video by a generative video engine through draft-conditioned editing. This decomposition separates process-level reasoning from high-fidelity visual realization. To adapt the video editor to simulated drafts, we construct VideoCoCo-3K, a curated dataset of draft-instruction-target triplets. VideoCoCo improves the OmniWeaving baseline from 0.475 to 0.558 on PhyGenBench and from 52.18 to 77.88 on VBench-2.0, achieving the best average score on both benchmarks. These results demonstrate that executable code provides an effective, controllable, and inspectable intermediate representation for physically consistent video generation.", "url": "https://wpnews.pro/news/videococo-code-as-cot-for-physically-consistent-video-generation-via-an-agentic", "canonical_source": "https://arxiv.org/abs/2607.27380", "published_at": "2026-07-31 04:00:00+00:00", "updated_at": "2026-07-31 04:39:06.845427+00:00", "lang": "en", "topics": ["artificial-intelligence", "generative-ai", "ai-research"], "entities": ["VideoCoCo", "Blender", "OmniWeaving", "PhyGenBench", "VBench-2.0"], "alternates": {"html": "https://wpnews.pro/news/videococo-code-as-cot-for-physically-consistent-video-generation-via-an-agentic", "markdown": "https://wpnews.pro/news/videococo-code-as-cot-for-physically-consistent-video-generation-via-an-agentic.md", "text": "https://wpnews.pro/news/videococo-code-as-cot-for-physically-consistent-video-generation-via-an-agentic.txt", "jsonld": "https://wpnews.pro/news/videococo-code-as-cot-for-physically-consistent-video-generation-via-an-agentic.jsonld"}}