WordsPerspective.py is probably just a Python script he wrote himself to do that, and not something you would find on the internet.
Each letter would be comprised of a 2D block of pixels that can either be on or off, to form the shape of the letter. I would conceptually construct a 2d array of pixels from the letters of each word, and set those arrays up adjacent to the viewpoint from which each word would be visible. Then I would cast rays or construct a field-of-view volume for each pixel empty from its corresponding view point. Here is a visual example using 2D:
The colored triangles represent the field of view for a couple of empty pixels, as seen from the white points of view.
Here, I've carved out pixels from a solid chunk of pixels to represent these empty pixels (I eliminated the third point of view for simplicity). By carving out all empty pixels from each viewpoint, you should end up with only the solid pixels needed to form the two words left over. Note that the larger a volume of solid pixels you start with, the more likely that the words will both be recognizable once you are done carving.