3yr anniversary of the SOTA classic: "Iron Man flying to meet his fans. With text2video."

Three years ago, a 4-second AI-generated clip of Iron Man soaring through the sky to meet adoring fans went viral, showcasing the nascent but groundbreaking capabilities of Runway's Gen-2 text-to-video model. Released in 2021, Gen-2 was among the first publicly accessible models that could translate a simple written description into a short, coherent video sequence. The Iron Man clip, created from the prompt "Iron Man flying to meet his fans," was a technical revelation at the time. It demonstrated that AI could not only generate a recognizable character but also maintain visual consistency across frames and simulate believable flight physics and camera motion, all within a cinematic style.

This viral moment served as a pivotal proof-of-concept for the entire field of generative video. Prior to Gen-2, AI video was often a disjointed slideshow of images. Runway's model, built on diffusion architecture similar to image generators like Stable Diffusion but extended across time, proved that temporal coherence was achievable. It set a new benchmark for what was possible, directly influencing the rapid development cycle that followed. The clip's popularity highlighted massive public and creative interest, fueling investment and research that led to today's state-of-the-art models from OpenAI (Sora), Luma (Dream Machine), and Runway's own more advanced successors, which can now generate longer, higher-fidelity videos.