Lesson 58 · 2 min
AI that watches video
Some AI models can watch a video and tell you what happens in it, though they usually look at snapshots, not every moment.
A video is a stack of still pictures plus sound. To understand one, a model usually samples frames, for example one or a few each second, and reads the soundtrack or transcript alongside.
That lets it summarise a lecture, find the moment someone mentions a topic, describe a cooking clip, or pull the steps out of a how-to video.
Sampling has limits. A quick action between two snapshots can be missed, and the model may guess at what it did not see. Long videos may be cut down to fit, so details in the middle can get thin.
For anything that matters, ask for timestamps and jump to those moments to check. A summary is a map of the video, and you still look at the place yourself.
Do it now
Summarise a tutorial
- 1. Find a chatbot that accepts video uploads or video links.
- 2. Give it a short how-to video and use the prompt below.
- 3. Jump to two of the timestamps and check that they match.
Watch this video and list the main steps in order, each with a timestamp. Mark any step you are unsure about because it happened quickly or off screen.
Quick check
Keep going
Next →AI that makes videoWords from this lesson
- Multimodal
- Able to work with more than one kind of input or output, such as text, images, sound and video together.
- Computer vision
- AI that makes sense of images and video, such as reading signs, spotting faces or checking products on a factory line.