COURSE 11L1100% FREE
Verified 2026-08-10

Image and video generation

Systems that create **images** and increasingly **video** from prompts, references, or edits—building on diffusion and related generative methods.

What it is

Systems that create images and increasingly video from prompts, references, or edits—building on diffusion and related generative methods.

Why it matters

Creative leverage is real; so are deepfake and consent harms (Course 29). Product teams must pair capability with policy.

How it works (plain)

Image: prompt → conditioned generator → optional upscalers/editors. Video: adds time consistency—harder, costlier, more failure modes (flicker, identity drift).

Always check commercial terms and likeness rights.

Everyday example

Concept art exploration vs publishing a photoreal image of a real person without consent—different ethical lanes.

Try it

Generate an image of an object you own; then refuse any prompt that depicts a real private person without permission.

Myths

⚠️ Myth: Video models “understand physics.”
✓ Reality: They approximate patterns; physical plausibility still fails.
⚠️ Myth: Watermarks solve misuse alone.
✓ Reality: Helpful layer, not a complete control.

Sources