A controller on Earth, Ed Fendell, manually operated the camera by radio control, knowing the time of liftoff and the ascent trajectory expected and referring to a time-and-angle chart without watching the video feed in real time! According to Fendell:
Now, the way that worked was this. Harley Weyer, who worked for me, sat down and figured what the trajectory would be and where the lunar rover would be each second as it moved out, and what your settings would go to. That picture you see was taken without looking at it [the liftoff] at all. There was no watching it and doing anything with that picture. As the crew counted down, that's a [Apollo] 17 picture you see, as [Eugene] Cernan counted down and he knew he had to park in the right place because I was going to kill him, he didn't — and Gene and I are good friends, he'll tell you that — I actually sent the first command at liftoff minus three seconds. And each command was scripted, and all I was doing was looking at a clock, sending commands. I was not looking at the television. I really didn't see it until it was over with and played back. Those were just pre-set commands that were just punched out via time. That's the way it was followed.
The camera was mounted on the Lunar Roving Vehicle and was used during the EVAs to look at interesting things here and there -- essentially a third set of eyes in the field!
The LRV and its remote-controlled camera was flown on the last three Apollo missions, and only the third really got a satisfactory video of the ascent.
No attempt was made to track the ascent on Apollo 15 due to a mechanical problem with the camera.
On Apollo 16, it was difficult to pan fast enough to catch the ascent because the LRV was parked closer to the LM than on 17.