OpenAI unveils Sora: realistic video from text

On February 15, 2024, OpenAI presented Sora, an AI model that generates realistic videos of several seconds from a text prompt or image. It was one of the most talked-about AI breakthroughs of that year and marked the moment text-to-video came of age.

What Sora could do

Sora generated videos up to 1080p in various aspect ratios and produced multiple variants at once. Striking was the coherence: objects, light, and movement remained consistent throughout the entire clip — exactly where earlier video AI fell short. The model understood basic physics and could construct scenes with multiple characters and camera movements.

Why it mattered

Sora showed that generative video was no longer a niche, but a serious creative tool for marketing, film, education, and prototyping. At the same time it fueled the debate about deepfakes, copyright, and the impact on the creative industry.

What came after

The competition responded quickly: Runway, Pika, Luma AI, Kling, and Google Veo released their own video models in 2024 and 2025. Video thereby became a permanent part of the generative AI wave, alongside text and image.