Face recognition means computing which individual from some other database of people a particular face belongs to.

There’s also face tracking — detecting a face in an image and then tracking the same face across subsequent images. Which is often implemented by using a face recognition approach, but without any predefined catalog of people — you just dynamically fill up your face database as faces appear in the image sequence / video source.

Recognition implies associating the faces with an ID.

It detects objects and gives you the bounding box. Then you draw a square on it and add a label.

No fancy LLM needed, just old fashioned machine learning models.