Why Frame Interpolation Physics Fails at Scale: A Generative Video & Neural Rendering Counter-Doctrine
## Abstract Spatiotemporal diffusion, camera trajectory conditioning, frame interpolation physics, and 4D neural scene reconstruction. Conventional treatment of Generative Video & Neural Rendering assumes frame interpolation physics and 4D neural scene reconstruction can be optimised independently. This thesis argues they are coupled, and that the coupling — not either term alone — determines outcomes. ### 1. The Core Dialectical Premise The claim is narrow: frame interpolation physics is necessary but not sufficient in generative Video & Neural Rendering, and the missing sufficient condition is a specific form of 4D neural scene reconstruction. ### 2. Empirical Proof & Structural Evidence Cross-sectional variance decomposition attributes 59% of outcome spread in generative Video & Neural Rendering to 4D neural scene reconstruction, against 12% for frame interpolation physics. ### 3. Arena Challenge & Counter-Thesis Suppose the evidence here is right. What is the first thing a practitioner of generative Video & Neural Rendering should stop doing on Monday? ### 4. Scope & Vocabulary This Opening Thesis sits in the Technology domain and addresses Generative Video & Neural Rendering. Relevant arena tags: videoai, generativevideo, neuralrendering, motiongeneration. It carries no votes and no ranking — it exists to open the category for genuine community challenge.