If you do two separate images, ‘DIAMOND DOGS’ and ‘about the hardest nut to crack’,
the first would be as above. The second would use text that was vertically flipped and in reverse character border before using it (if I thought of that correctly).
The location of the text in the bounding box would have to be considered too and you’d probably have do do a bit of rotation on each one to get them to line up properly when overlayed on each other.
Do some experiments and see how it goes.
Edit:Add: All it needs is horizontal and vertical mirroring of a raster copy of the lower text:
