How to train a machine to recognize a video it knows?

Question

Video

I don’t even know how to better formulate the question than it is formulated in the title. The essence of the task is to get a sound and a picture from the input video stream and compare them with those available in the storage and determine whether there really is such a video on the input video stream that is known to the machine. I'm not asking you to dump me a ready-made solution, I just want to ask knowledgeable people to help approach the solution.
From the existing ideas, there is already an option to apply an algorithm similar to the one that uses shazam to find songs in order to check the existence of a recording by sound.

Answer 1

ffmpeg