starting from a 'fork' of this very simple terrain demo, i mostly just upload my customised 'heightmap' to replace the one called 'Heightmap.jpg' in their Terrain entity.
i've been using 2048*2048 greyscale jpegs with tracings of the 1909 dublin map. pure black is the 'zero' height and i use trial and error to judge how gradually the black level should lighten. (weirdly, you need to 'flip horizontal' the map or left and right will be backwards.)
my primary goal is to give an intuitive sense of distances, so i want the building-fronts along each street to be approximately correct relative to the street widths. i have a huge problem with conical artifacts where i want flat right-angles. (update: jpegs default to color blending that causes some of the cones, but others are jaggies when the buildings aren't positioned along the right angles)
i haven't figured out how to set the 'camera's eyelevel-- usually when i lower it it falls thru the 'terrain' which seems to float slowly up into the lovely cloudy sky like graceful polygonal seagulls. (there's simulated gravity but no ground.)
the terrain claims to be divided into 250 subdivisions, and raising this number may reduce the cone problem, while slowing the rendering. (setting it to 256 causes artifacts though)
the 2048 width seems to result in serious averaging, so 256*256 may be the best compromise.
these early drafts mostly look like snow mazes: (w-a-s-d = forward-left-back-right)
eccles to butcher
sandymount
tower
tower b
tower b uses a 256*256 heightmap which can theoretically be optimised to capture tons more detail-- but it's not obvious how to efficiently attack this
