OpenAI previews Sora for high-fidelity text-to-video generation

Summary

The Sora research preview used a transformer over latent spacetime patches to generate videos of up to one minute from text and image inputs. OpenAI highlighted both visual quality and failure modes involving causality, object consistency and spatial directions, initially limiting access to red teams and selected creatives. Capability, benchmark and availability claims are attributed to OpenAI unless independently verified. Follow the original source for complete technical details, current availability, licensing and later updates.