Lesson 3: Beyond Text: Images, Audio & Video

What You'll Learn

That generative AI makes more than text — it also makes images, audio, and video
Roughly how it does this: by learning patterns from huge numbers of examples
What a "deepfake" is, and why it raises questions we'll explore later

More Than Words

So far we have talked about AI that writes text. But the same basic idea — learning patterns from huge piles of examples — also lets AI create images, audio, and even video. This whole family is called generative AI, because it generates (creates) new things rather than just sorting or labeling them.

The headline feature for most people is text-to-image: you type a description, and the AI produces a picture to match. Type "a cozy cabin in a snowy forest at sunset," and out comes an image that fits those words. You did not draw anything; you described it, and the AI painted it. Tools like this have made it possible for anyone to create artwork in seconds, no drawing skills required.

How It Roughly Works

You do not need the technical details, but a simple picture helps. To make images, an AI is shown a vast number of pictures, each paired with a short description — "a red apple," "a golden retriever puppy," "a city street at night." Over time, it learns the patterns that connect words to visual details: what "red" tends to look like, what shape an apple has, how puppies are built. Then, when you give it a fresh description, it uses those learned patterns to build a brand-new image, pixel by pixel, that matches your words. It is not pasting together photos it saved; it is generating something new from patterns.

Audio works on the same principle. Trained on many hours of recorded speech, an AI can learn the patterns of a human voice and then read your text aloud in a natural-sounding way. Other tools learn from music and can generate a short tune in a given style. Video, which is really just many images shown quickly in sequence (plus sound), is the hardest of the three, but the same idea applies: learn patterns from lots of example clips, then generate new frames that flow together. Because video is so complex, AI-made video is improving fast but still often shows odd glitches.

A Quick Word on Deepfakes

There is a powerful and tricky side to all this. Because AI can imitate a real person's face or voice, it can be used to make a deepfake — a fake image, audio clip, or video that looks or sounds like a real person saying or doing something they never actually did.

Deepfakes can be harmless fun, like putting a friend's face into a movie scene as a joke. But they can also be used to deceive — to spread false information, to impersonate someone, or to damage a person's reputation. This is exactly why it is becoming so important to be a little skeptical of what you see and hear online. If a video seems shocking or out of character, it is worth pausing to ask whether it is real.

We will dig into these questions much more deeply in the ethics module. For now, the key point is simply this: generative AI can create remarkably convincing pictures, voices, and videos, so "seeing is believing" is no longer a safe rule on its own.

Key Takeaways

Generative AI creates images, audio, and video — not just text — including text-to-image, where a description becomes a picture.
It works by learning patterns from huge numbers of examples, then generating brand-new content that matches your request.
Deepfakes are AI-made fakes of real people; they show why we should be thoughtfully skeptical of online media, a topic the ethics module explores further.