Learn Without Walls
AI Certificate › Day 17
Day 17 · 15 minutes

Sora 2

Generate a cinematic clip from a description.

~15 min read + 10 min hands-on

1 Why this tool matters

Sora 2 is OpenAI's text-to-video model. Its focus is on physical realism — objects bounce, fall, and collide the way things actually do. For scenes that need to look genuinely believable rather than artistic, Sora 2 is typically the stronger choice over Veo for the same prompt.

2 Walkthrough

  1. Go to sora.com. Sign in with your OpenAI account.
  2. Try a prompt where physics matters: A ceramic coffee cup slides slowly across a polished wooden table as the camera follows. It reaches the edge and tips, spilling coffee. 6 seconds, slow motion.
  3. Compare to the same prompt in Veo 3. Which feels more physically correct?
  4. Try an interior scene: A small dog running up a flight of carpeted stairs. Handheld camera. 5 seconds.
  5. Save both outputs. Notice your own taste — where do you prefer Sora, and where do you prefer Veo?

3 Your turn

Today’s exercise

Pick a scene from your life — something ordinary. Generate it in Sora. The point is not to make art; it is to see how passably your real environment can be simulated by an AI today.

4 Pro tip

Worth keeping

Sora is strict about requests involving real people, logos, and children. Describe generic roles (“a woman in her 40s in a lab coat”) rather than named individuals.