Skip to content
inclusionAI Repositories·· 7 d agoPicksAI score71

Inclusion AI releases Code World Model 2.0 and Invoker-Render streaming video renderer

CodeWorldModel

AI summary

0 and its renderer Invoker-Render, which takes a structural proxy video, one appearance reference image and text, and generates appearance in four steps per chunk. The causal streaming model, built on MiniMax-H3, SGLang and SGLang-Omni, keeps a fixed appearance anchor and a bounded history window, so a session runs as long as proxy frames arrive; the demo shows one continuous 30-minute racing generation.

Why it matters

The release details a four-step streaming video renderer that accepts coarse or fine structural proxies and sustains a 30-minute generation without resets, with weights on Hugging Face.

Source:inclusionAI Repositories · github.com