How does AI watch video?
Last time, AI listened and talked with you like a real person. Now give it an hour-long video. A few seconds later, there's a summary. How does it watch?
FRAMES
A video is a stack of photos
A video is really a stack of still pictures, called frames, flipped very fast. Usually about 30 per second, with a soundtrack running alongside.
SAMPLING
Only a few frames per second
One hour of video is about 108,000 pictures. Watching them all is too heavy. So AI only takes samples: a few frames per second, sometimes just 1.
SPEECH
It listens to the speech too
At the same time, AI transcribes what's said into text, the same way it heard you in part 1. Frames and sentences all have timestamps, so AI knows what happened at which minute.
PUTTING IT TOGETHER
Pictures plus words make the main idea
Putting pictures and words together, AI gets the main content: who's talking, what they're doing, which slide is on the screen.
WEAK SPOT
Quick details get missed
But because it only skims, something that happens in half a second can slip between two frames. A lightning-fast goal, a small line of text flashing by. AI may miss it, or even guess.
PART 5
Can AI make videos?
AI can watch videos. But can it make them? And how do characters keep their shape halfway through? See you in part 5.
This article is based on the video AI watches video from the Mark học AI channel. Watch the video (in Vietnamese) to see the animations.