Microsoft open-sources Code Video Model: code first, pixels later for controllable video gen
xiaohu · x · 2026-10-13
Microsoft Research open-sourced Code Video Model, borrowing the film industry's previs workflow to give AI video generation an inspectable, editable intermediate step.
- A coding agent writes a Three.js program that lays out the scene, object trajectories, camera moves and even physics-based motion (spring oscillation, double pendulum) as a low-fi animated draft you can fix by editing a line of code.
- Early-step Conditioning injects the code video only into the first 25% of diffusion sampling steps to lock composition and motion, then drops it so the model restores realistic texture from reference images — no retraining needed.
- Supports targeted edits (arch to doorway, humanoid to quadruped) and independent camera/object motion planning. Paper uses GPT-6 for code and MiniMax-H3 as the video model.
Related event: Microsoft Open-Sources Code Video Model for Controllable Video Generation(3 posts)→
More from coding & agent
- Dev spins up a horde of Claudes to hit a one-week deadline — "it worked out" — rickasaurus · 2026-10-13
- Debug Multi-Agent Workflows: Permissions, Handoffs and Channels Beat Model Blame — Al_Grigor · 2026-10-13
- iOS Dev Mocks the AI-Agent Era: HomeView, agents.md, Skills and App Store Rule 2.3bc — Dimillian · 2026-10-13
- Who decides when AI evaluator and human reviewer disagree on an agent call? — AssociationNew7925 · 2026-10-13
- Dev builds real-time dictation pipeline: AssemblyAI Sync API plus LLM Gateway rewrite — tech__unicorn · 2026-10-13
- 'Spinbusy': Yacine mocks coding agents stuck in pointless fix loops — yacineMTB · 2026-10-13