The CRT image, when grossly slowed down, is a bright spot traveling the screen (with some considerable afterglow behind that exponentially decays).
Aim a photodiode at a small area of the screen (for simplicity say a small piece of one scan line), and you get a pulse from its output - with timing offset from the vertical sync pulse that equals vertical distance from top, and offset from horizontal sync that equals the distance from the left.
(In reality you get a burst of a few pulses as the optics will see more scanlines than just one.)
Typically there is a spiral groove, with the sound encoded as minor left-right and depth changes (for stereo). The needle travels in the groove, sensors on it (piezo or electromagnetic) are sensing its movements (the inertia of the pickup is holding it in the middle of the groove while the low-mass needle on a cantilever is following the groove including the aberrations that encode the sound). The path differences between the pickup head and the needle are directly taken from the head sensors, amplified (and equalized to correct for the higher frequency attenuation of the recording method), and sent to the speakers.
