ML//diffusion model//Sora

Sora is OpenAI's text-to-video model (February 2024): it generates coherent multi-second videos from text prompts, cinematic, physics-aware and temporally consistent. It came with no paper, only a blog post, and a DiT architecture underneath.

Sora is OpenAI's text-to-video model (February 2024): it generates coherent multi-second videos from text prompts, cinematic, physics-aware and temporally consistent. It came with no paper, only a blog post, and a DiT architecture underneath.

Its competitors include the open-weight Open-Sora and Wan 2.1, and the closed Kling and Pika. Video generation moved from research curiosity to shipping product in 2024 and 2025.