Event
Microsoft proposes Code Video Model for controlling video generation with executable code
1 reports · 1 sources
The Microsoft team proposes Code Video Model: a Coding Agent converts prompts into Three.js programs, unifying geometric layout, object motion, camera trajectories, and event timing into an executable scene representation, then uses program-rendered preview videos to guide video generation. According to the team, the research examines condition composition, intervention stage, duration, and native pathways while keeping MiniMax-H3 parameters frozen; the video model remains frozen and requires no additional training of control modules; for both types of pathways, condition intervention has a greater impact on generation at approximately the first 25% of the denoising process.
Report timeline
- Microsoft proposes Code Video Model: controlling video generation with executable code
A Microsoft team proposed the Code Video Model, in which a Coding Agent turns a prompt into a Three.js program, unifying geometric layout, object motion, camera trajectories, and event timing into an executable scene representation, then uses a program-previewed video to guide video generation.
机器之心 · 微信公众号 ·