Back with another question about modeling a terrain from a 2D drawing of it, using contours, etc.
I’m using the map of a favorite spot of mine from Final Fantasy XI, Qufim Island, as the source image to work from.
The approach I’m taking is starting with a large flat mesh that covers the entire area to be modeled, with some “overhang”. I subdivide it several times to give myself some vertices to play with and then just start moving the vertices around to match the contours. Next step would be to extrude those areas deemed as cliffs and then go through, “fill in” and refine each area 'til it’s done.
I’ve done a couple runs at it, but find that I’m running into the trouble of having vertices or edges in the way, which makes me wonder how others might approach the same project. Perhaps starting with fewer subdivisions so it’s a really blocky starting point, or perhaps taking a different approach altogether?
Given the attached map image and asked to make a full-scale 3D environment out of it, what approach would you take to build it out?
Some info about the map, assuming none will know what this area looks like in-game:
The areas in the darker orange/yellow are the navigable areas.
The long, winding, white strip you see leading to where it says “Grand Dutchy of Jeuno” is a tunnel, which I imagine would be created via extrusion and moving vertices and such. The rest of the map, though, is open air.
Exclude Delkfutt’s Tower - it’s a massive white tower you enter from Qufim and I wouldn’t be concerned, initially, about modeling that - though it would be placed eventually, so space would be left for it.
I’m not asking for someone to tell me how to do it or give me step by step. But I’m sure I can learn some valuable tips from those more experienced in such a task.
I’d copy the map into a paint program, make it gray scale, then paint solid black over the oceans, and increase the contrast a little bit so that the higher ground areas were whiter. Then, I’d add the map as an image texture to a plane that was sub-divided about six times, then use the Noise function to make a height-field out of it, then make any adjustments from there.
That is certainly a way to get the topography in place a bit faster. I’ve taken that approach in the past for real-time terrain and have never been happy with the results (edges aren’t always aligned to follow the contour of a hill or cliff and you get that ugly jagged thing going on). But I think I can see it working as a basis to at least start on something that would be tweaked anyway. Would certainly save the time of having to select all the faces, etc.
Grayscale and noise was my first thought too, cap’n, but that’s a shadow map rather than relief… the dark parts aren’t higher, they’re shaded from the “sun.” It’s still a good idea in that it’ll make the reference image easier to reference.
Otherwise, I wouldn’t worry so much about matching the contour (i.e. moving verts x, y), but rather leave the subdivided mesh as a grid (as viewed from top) and move the verts in the z direction.
I’ve found the decimation modifier works fairly well to get rid of the “grid” look and cut down the vert count. A subdivision modifier is also a good tool for this.
I’m not familiar with Final Fantasy, but if it’s a 3D game you may be able to extract the map data from the game files. You might Google around for instruction along those lines.
Hmm, well in this case - and I know I didn’t explain this in my first post - the purpose of this exercise o’ mine is three-fold:
Improve my skills at modeling in Blender
Improve modeling terrains for real-time 3D in Blender
Wind up with a final product that can be exported, loaded into a 3D game engine (which ever) and explored.
I’m very inspired by the way terrain is created in FFXI and, I believe, in Everquest II. In each game, terrain is “hand-crafted”, not “height-mapped” with 3D props placed for cliffs and such, as in World of Warcraft or Unreal Tournament. The FFXI approach just seems to have a much more “real” and “solid” feel to me. Hard to explain, really.
I don’t knock the heightmap/prop method as it certainly has its benefits. But, to me, relying on reusable props leads to repetition in geometry and many times things won’t look “unified” - you can see the creases where the terrain mesh meets the prop, etc. Either way I’m going to have to model all these assets I want to use so may as well model them all in as a contiguous, unified model.
So that’s the approach I’m going for. Might have to break it down into segments to make it more efficient in a real-time engine, but I can worry about that later.
So for me, relying on a straight-up square mesh with vertices draped along the Z values of a height-map wouldn’t get the effect I’m after by itself - though it might serve as a good starting point.
