Tag

#text-to-video

3 articles

Microsoft Research’s World-R1 Uses Flow-GRPO and 3D-Aware Rewards to Inject Geometric Consistency Into Wan 2.1 Without Architectural Changes

This article explains how Microsoft Research's World-R1 uses reinforcement learning and 3D-aware rewards to improve geometric consistency in text-to-video generation without changing the underlying model architecture.

Apr 3014

OpenAI’s former Sora boss is leaving

Learn how to create simple AI video generation projects using Python and existing AI libraries, understanding the fundamental concepts behind systems like OpenAI's Sora.

Apr 1728

Researchers define what counts as a world model and text-to-video generators do not

This article explains what world models are in AI, how they work, and why they matter — while clarifying that systems like Sora are not world models.

Apr 1248