Immediately after OpenAI released Sora, its new text-to-video model, there was rampant speculation on how it was trained. Yet, details are scarce.

On Thursday, OpenAI once again shook up the AI world with a video generation model called Sora.

The demos showed photorealistic videos with crisp detail and complexity, based off of simple text prompts. A video based on the prompt “Reflections in the window of a train traveling through the Tokyo suburbs” looked like it was filmed on a phone, shaky camera work and reflections of train passengers included. No weird distorted hands in sight.

Read the full article at: mashable.com