Skip to content
FORGENO ADS GAMESUnlock all $8

AI Skills Academy · Unit 6 of 8

Unit 6: Images, voices and deepfakes

How AI makes images, how to check whether a picture, video or voice is real, and how to create responsibly.

Opens the first lesson of this unit in AI Skills Academy. Every lesson is open from the start.

Lesson 1: How AI makes images

Goal: Explain diffusion simply and write a strong image prompt.

Many image generators are diffusion models. During training, they see enormous numbers of images with captions. Each image is gradually covered with random noise, like TV static, and the model learns to reverse that: to remove a little noise at a time, guided by the caption.

To make a new picture, the model starts from pure random noise and removes noise step by step, steered by your text prompt, until an image appears that matches the description. That is why the same prompt gives a different picture each time: it starts from different noise.

A strong image prompt covers:

  • Subject: who or what, doing what
  • Setting: where and when
  • Style: photo, watercolor, cartoon, pixel art
  • Lighting and mood: golden hour, dramatic, soft
  • Composition: close-up, wide shot, from above
  • Shape: square, tall or wide

Worked example: weak to strong image prompt

  1. Weak: a dog
  2. Strong: A golden retriever puppy splashing in a shallow creek, late afternoon sunlight, wide shot, warm watercolor illustration style, square image

Every added detail narrows the many possible pictures down to the one you imagined.

Image models have their own weak spots: hands and fingers, readable text, exact counts ("seven apples"), and keeping a character looking the same across several images. These are improving fast, which is exactly why you cannot rely on spotting them to tell real from fake.

Diffusion model
An image generator that creates a picture by starting from random noise and removing it step by step, guided by a text prompt.

Remember: Subject, setting, style, light, composition, shape. Each detail narrows the picture down.

Lesson 2: Checking if a picture or video is real

Goal: Check where media came from instead of relying on visual clues alone.

You may have heard tips like "look at the hands" or "check whether people blink". Visual clues still catch some fakes: warped backgrounds, garbled text on signs, earrings that do not match, shadows pointing different ways, skin that looks too smooth, lip movement out of sync with speech.

But clues are a losing game, because each new generation of tools fixes the old giveaways. And a real photo can look odd too. The reliable approach is to check where it came from:

  1. Who posted it first? Trace it back. An account created last week with no history is a warning sign.
  2. Is anyone reliable reporting it? If a world leader really said something shocking, established news organizations would be covering it.
  3. Reverse image search. Search with the image itself (several search engines and apps can do this) to find older copies or the original context.
  4. Look for labels. Some tools and platforms add "made with AI" labels or invisible content credentials. Their presence tells you something. Their absence proves nothing.
  5. Pause before sharing. Fakes are often designed to make you angry or scared, so you share before you think.

Hands-on lab in the room: Real or fake? checker. Six situations, each with a choice of next move. Pick what a careful checker would do.

Deepfake
A fake video, image or audio clip, made with AI, that shows a real person saying or doing something they did not.
Reverse image search
Searching with an image instead of words to find where else it appears online.

Remember: Clues help, but sources decide. Who posted it first, and who reliable is confirming it?

Lesson 3: Voice clones and the emergency scam

Goal: Recognize voice-clone scams and set up defenses with your family.

AI can now clone a voice from a short recording, such as a few seconds pulled from a video posted online. Scammers use this for an old trick with a new twist: a frantic call from someone you love who needs money right now.

Worked example: how the scam runs

  1. Your phone rings. It sounds exactly like your cousin: "I've been in an accident. Please don't tell Mom. I need $800 sent right now."
  2. A second voice may take over, posing as a lawyer, a police officer or a hospital worker.
  3. They push for payment in a way that is hard to undo: gift cards, a payment app, a wire transfer or cryptocurrency.
  4. Pressure, secrecy and an unusual way to pay are the giveaways, not the voice, because the voice sounds real.

The U.S. Federal Trade Commission (FTC) has warned about exactly this family-emergency scam. The defenses do not depend on spotting the fake:

  • Hang up and call back on the number you already have for that person, or call another family member.
  • Agree on a family code word that a caller must say in a real emergency.
  • Slow down. Real emergencies survive a two-minute check. Scams do not.
  • Never pay with gift cards or other payment methods you cannot reverse just because a caller demands it.

Remember: Pressure + secrecy + strange payment = hang up and call back on a number you know.

Lesson 4: Consent, copyright and respect

Goal: Create with AI without harming people or misusing other people's work.

Being able to make any image does not mean every image is OK to make.

Real people need consent. Making fake images or videos of real people, especially classmates, coworkers or exes, can be cruel and can be illegal. Making or sharing fake sexual or nude images of anyone is a serious harm, and in many places, including under U.S. federal law since 2025, publishing them is a crime. If it happens to you or someone you know, tell a trusted adult and report it on the platform.

Copyright is still being worked out. Courts and lawmakers are still deciding many questions about training AI on other people's work. In the U.S., the Copyright Office has said that material generated purely by AI, without enough human creative control, cannot be copyrighted, while human-authored parts of a work can be.

Respect artists. Do not pass off AI images as hand-made art. Do not sell AI images "in the style of" a living artist as if they were theirs. For commercial use, check the tool's terms of service.

  • Would the person in this image agree to it?
  • Could this embarrass, scare or mislead anyone?
  • Am I being honest about how it was made?

Worked example: a school poster contest

Your school runs a poster contest and you would like to use an AI image tool.

  1. Read the rules first: some contests ban AI images, others allow them with a label.
  2. If it is allowed, write your own prompt, then edit, arrange and add your own text, so the design is yours.
  3. Label it honestly, for example: "Background made with an AI image tool; design, text and layout by me."

Honest about the tool, clear about your own contribution, and within the rules.

Remember: If a real person would not agree to it, or it could mislead people, do not make it.

Practise it in AI Skills Academy

Each lesson ends with two questions that check you can use the idea, not just recognise it. Wrong answers explain why and come back in Review. The unit ends with a 5-question quiz; pass it with 80% or more to master the unit. Hands-on lab: Real or fake? checker.

Every lesson is free: no ads, no lives and no paywall.