AI Lessons
Lesson 4: AI and Video - How does AI create videos?

Purpose: How AI generates videos by predicting sequences of visual frames and motion from text prompts.

No. of Classes

1 - (Time : 1 hour 30 minutes, Laptops/desktops : 10, Students strength : 15 to 20).

Materials Required

Laptop / Desktop with Internet connections / Wi-Fi.

Curriculum Content

1. AI Literacy

  • 1.3 Using AI tools
  • 1.4 Prompting and prompt effectiveness

2. AI and ML Concepts - A Simplified Overview

  • 2.2 Generative AI fundamentals
  • 2.4 Digital representation of media
Prior knowledge
  • Browsers
  • Basics of prompting
  • AI image generation
Exercises

Exercise (1)


  • Explore the different components of a video, such as frames and audio. Observe how a video is created by displaying a sequence of images rapidly.

Exercise (2) - Unplugged activity


  • Predict the next frame in a sequence of images. Discuss how AI generates videos by predicting one frame after another. Click here to open the activity document for this exercise.

Exercise (3)


  • Use Google Flow or any similar video generation tool to create videos from text prompts. Observe how changing the prompt changes the generated video. Click here to open the activity document for this exercise. Click here to open wild prompt ideas to try with the video tool.

Solutions


Teacher's Instruction:
  1. Check the availability of the AI tools used in this lesson and read the instruction documents for all the exercises before the class.
  2. A video is made of frames and sound, and both are stored as digital data.
  3. AI is trained by showing lots of videos. It learns patterns in the frames and sounds so it can predict what comes next or create new videos from prompts.
  4. AI generates video by predicting what will come next.
  5. AI does not understand the scenes. It predicts based on the prompts and the patterns it has been trained on.
  6. Points to Ponder:
    • Your prompt was only a few words long, but the video ran for several seconds. Where did everything you did not describe come from?
    • AI videos often go wrong with hands, or objects pass straight through each other. If the video is built one predicted frame at a time, why would that cause such mistakes?