In brief: Google Photos already offers a slew of features, and while it's always been able to recognize faces in pictures to help you manage them more easily, users have been asking for the ability to do the tagging manually. The company is now rolling an update so you can add in the faces missed by Google's algorithm.

You can think of the over 1 billion users as living proof of how good Google Photos is (it's also "free" for unlimited photo and video storage).

A recent update added the ability to use Google's optical character recognition tech to search for text in photos, but users have been waiting for something a bit more pressing. If you've enabled "face grouping," Google Photos is able to automatically detect faces in your photos and tag them, but it's not perfect, so the latest update will now allow users to manually tag people in pictures, with a catch.

It still relies heavily on Google's algorithm, so you won't be able to tag people if it fails to detect that there's a face in the picture you selected. That means that you'll depend on Google to at least realize that there's a person in the photo before you can intervene to tell it who that is.

The new feature is mostly helpful in those situations where the algorithm missed faces that are slightly obscured or photographed from a strange angle. Google's algorithm isn't able to discern the face from a dark silhouette, but will tag passerby and background people, so you'll also be able to clean those up so they don't clutter the People section in the app.

Google says it doesn't share recognized faces between accounts, however it's safe to assume that they used this data to train their algorithms. If you turn "face grouping" off, the company deletes the facial models, possibly to avoid potential class-action lawsuits from privacy-conscious consumers.

By contrast, Facebook has addressed facial recognition fears by saying that it stopped the automatic face scan. However, when FTC levied a $5 billion fine against the company, it didn't explicitly ask for the removal of previously generated recognition models.