It's fairly simple really! We are getting the output of your camera and converting it into binary. On top of that we are combining it with timestamp and then hashing that result. Thanks to all that noise and your timestamp we are getting completely random results every time.
How to improve your results?
You have to have constantly changing images out of your camera.
So how to do it?
Your computer generates pseudo-random numbers that theoretically can be guessed before time. However, this is my example of a so-called TRNG (true random number generator). Why is it better? Simple TRNGs use physical means like in this example camera output to generate a real random number.
Explanation
It is hard to predict how many cars, animals and people are gonna pass through the road in front of your house.
Combined with other variables like lighting, weather, colors of clothes that people wear, or how the leaves will move in the wind.
There are a lot of variables coming out of video output and combined with timestamps we are getting an impossible to predict Random number.