Understanding AILesson 7/8

Lesson 7 of 8 · 12 min

Images, voice and video

AI that draws, speaks and films: possibilities, limits and rights.

So far, we have mostly talked about text. But AI can also create images, voices, music and videos. These tools open up great possibilities, as long as you know their limits and the rules of the game.

What AI can create

  • Images: you describe a scene in a few sentences, and the tool produces an illustration, a realistic photo or a logo. Examples: ChatGPT, Gemini, Midjourney, Adobe Firefly.
  • Voices: text becomes natural narration, in several languages. Some tools can also copy a voice from a few minutes of recording. Example: ElevenLabs.
  • Music: a mood, a jingle, a full song from a description.
  • Videos: short clips generated from text or an image, or an edit assembled automatically.

The principle is the same as for text (lesson 2): the tool learned from huge amounts of images, sounds or videos, and it produces the most likely result for your description.

Describing what you want

The recipe from lesson 3 applies here too. The more precise your description, the better the result:

  • the subject: who or what, doing what;
  • the style: photo, illustration, watercolor, simple drawing;
  • the mood: colors, light, season, emotion;
  • the format: square for social media, horizontal for a banner.

A warm illustration of a neighborhood bakery early in the morning, watercolor style, orange tones, square format, no text.

Tip: an assistant like Claude can help you write this description, which you then paste into an image generator.

Limits to know

  • Text inside images is often misspelled: add your titles yourself, in a layout tool.
  • Details can look strange: hands, reflections, objects that blend into each other. Look closely before you use the image.
  • Consistency is hard: keeping the same character from one image to the next, or from one video shot to the next, takes practice.
  • Bias exists here too (lesson 4): ask for “a doctor” or “a nurse”, and check the diversity of what you get.

Rights and respect for people

This is the most important part.

  • Read the tool’s terms. They say whether you can use the creations commercially. Free versions are sometimes more limited.
  • Copyright on AI creations is still uncertain. In Canada as elsewhere, the law has not settled every question yet. For a logo or an important brand, get professional advice.
  • Do not imitate a specific work or artist for commercial use, and do not use protected brands or characters.
  • Never create a real person’s image or voice without their consent. It harms their privacy and reputation, and it can be illegal.
  • Say when AI made it. A simple note is enough: “Image created with AI”. Several platforms already require it for realistic content.

Key takeaways

  • AI creates images, voices, music and videos from a description.
  • A good description gives the subject, the style, the mood and the format.
  • Check the details: text, hands, consistency, bias.
  • Respect rights and people: never someone’s image or voice without their consent, and mention AI.

Quiz

1. Which description will give the best image?
2. Your generated image has a title full of spelling mistakes. What do you do?
3. A friend suggests copying your boss’s voice as a joke. What do you think?
4. You publish a realistic image created with AI. What is good practice?
Next