Differentiable rendering going LLM-based makes sense; video generation may replace path tracing

jongranskog · x · 2026-09-07

The author notes differentiable rendering is becoming LLM-based—unexpected but sensible—and predicts final rendering via video generation is next, since path tracing remains slow regardless of scene detail. A brief technical take on where graphics meets generative models.

Original post →

More from Multimodal

Multimodal channel →