Neural Dynamic GI: Random-Access Neural Compression for Temporal Lightmaps in Dynamic Lighting Environments

Neural Dynamic GI (NDGI) addresses a specific pain point in the graphics industry: the massive storage cost of precomputed lightmaps for dynamic lighting. While Ray Tracing (DXR) is the future, baked lightmaps remain the standard for mobile and mid-range hardware. The project's defensibility (Score: 4) is currently rooted in the technical difficulty of implementing random-access neural compression that is fast enough for shader-side sampling, but as an open-source paper implementation with 0 stars and 5 forks (likely the authors), it lacks an ecosystem moat. The 'frontier risk' is medium because while LLM labs like OpenAI don't compete here, 'Frontier Graphics Labs' like NVIDIA Research or Meta Reality Labs are actively developing similar technologies (e.g., NVIDIA's Neural Texture Compression). Platform domination risk is high because this functionality is a natural candidate for integration into Unreal Engine (Lumen/Nanite pipelines) or Unity. If Epic Games or NVIDIA releases a native equivalent, this standalone implementation becomes obsolete. The 5 forks within 3 days indicate immediate peer interest from the graphics research community, but it has yet to see industry adoption.