Tap it, it's sharp.
How does it know?

The full narration of the video.

Tap, and it turns sharp. Not editing. A tiny lens actually moved. You shoot a cup: the window behind is crisp, the cup is a blur. Tap the cup, and it snaps into focus. Tap a spot, that spot gets sharp. How does the phone know? Let's follow that tap inside.

First, why things blur. Light passes through the lens and has to meet again in tiny points behind it. That's a sharp picture. Light from near things meets farther back, from far things farther forward. There's only one sensor, so the near cup and the far window often can't both be sharp. The fix is to move the lens: forward, and near things sharpen. Back, and far things do.

How far does it travel? Less than a millimeter. What pushes it is a coil of thin copper wire and a small magnet, the same trick your earbuds use to make sound. Send current and the magnet shoves the coil, more current, farther. Cut the power and a tiny spring pulls it back. It's called a voice coil motor, and it sits inside that little camera bump.

The real puzzle: the phone has to know which way, and how far. Picture the lens as a car backing into a parking spot. Older phones were like reversing without a parking sensor: creep a bit, stop, take a look. Look at what? The edges. Where it's sharp, light and dark sit right next to each other. Where it's blurry, they smear into gray.

Move a bit, edges sharpen, keep going. Go too far, it blurs, back up. That's why focusing used to look like the picture was hunting. The newer way is like having a parking sensor: no inching and checking, it just tells you how far, and which way. It relies on some special tiny squares on the sensor, each split in two: the left half only catches light from the lens's left side, the right half only from the right.

Out of focus, the two halves see the cup as two shifted copies. In focus, the copies merge into one. Which way they're shifted tells forward or back. How far they're shifted tells how much. One calculation, and the motor pushes straight there. This is called phase detection autofocus, a phrase you'll often see on spec sheets. The sensor's "this much farther" is, in the phone, how far apart the two copies sit.

That tap also sets the brightness from the spot you touched: tap a dark spot, the photo brightens. Tap a bright one, it dims. So it's sharp where you tap because you told it where to look. The rest is two copies compared, and one push from the motor. Try this: aim your phone at a blank wall and tap. No edges, nothing for the copies to compare, so many phones hunt back and forth.

Stick up a page with writing and tap again: it locks right on. Aimed at a blank wall, does your phone lock on, or hunt?