June/July Entry: Lost Ichenjanda

[gilgamec]

Ah! Another contest to try out some new procedural techniques.

I'll make a map of the lost city of Ichenjanda, as excavated by a team of archaeologists researching the old Challipai empire. Ichenjanda is situated on a hill overlooking the Bujapur River, and had a commanding position in the control of military and commercial movements on the river. It was abandoned nearly a thousand years ago, but had been declining before then. In legendary times, it was the capital of a mighty empire; this is the Ichenjanda the archaological team is seeking.

This one looks like a good opportunity to make a more schematic map, with vector and not raster graphics. I've been playing around with some ideas on how to get common effects procedurally. As with last month's Challenge, I start with a sketch:
sketch.png
This shows the location of the river, forest, and hill crest. Add some noise to loosen up the boundaries, create a height map, and render with contour lines and some other effects, and we get my first
### Latest WIP ###
Ichenjanda-01.png
(This is just an overview inset, showing the situation of Ichenjanda in the surrounding area. The main map will be focused on the city, on the hill in the middle.)

[msa]

Oooo! I'm looking forward to another one of your series! I might recommend this time that, at some point, you stop with the autogen and finish off by hand. I think what you are doing procedurally is fantastic, but I think the touch of an artist's hand might be good for the final product.

[gilgamec]

I might recommend this time that, at some point, you stop with the autogen and finish off by hand. I think what you are doing procedurally is fantastic, but I think the touch of an artist's hand might be good for the final product.

*gasp*! Anathema!

I know what you mean, though. Doing things by hand makes them so much more ... controllable. The problem is that I don't really have a clear picture in mind of what I'm after, so I have to experiment to see what I want, and it's much easier to mix-and-match experiments when they're procedural -- when the forest shading depends on the forest outline depends on the contours depends on the initial sketch, I can fiddle with the sketch and immediately see the change in the forest shading. (To some extent, it's also the fact that I have no real artistic talent.)

This pretty much pushes me to the realization that I'm only going to use a manual change if

1. it's much (like, orders of magnitude) easier than doing the same (or similar) thing procedurally, or
2. it only comes at the very end of the production chain.

For instance, I probably could have done the current map in Illustrator in an hour or so (and I'm by no means talented with Illustrator), as opposed to the 18+ hours it took to code. However, the procedural version gives me

• a height map I can use later, when placing the Lost City's streets;
• guaranteed consistency between this overview and the later detail maps;
• tighter control over stylistic consistency (this could probably also be done in Illustrator, but it's much more work); and
• the ability to change the sketch (which I did a couple of times during development) and have the changes filter down easily.

The only things that are manual are the text, vertical scale (which I picked to give "interesting" contour lines), and contour line label placement (which took the longest time of any of these things; re-positioning all of them now, however, would probably take less than ten minutes). The text is probably going to always be manual. The vertical scale I chose to maximize interesting contours, so it's not really automatizable. The contour line label placement is the only one that could conceivably be done procedurally, but I don't have a clear idea how. If I decide to come back to contour maps later, it'll be the first thing I look at.

Even given all of that, the only things I'm really not happy about are some strange digit alignments in the contour labels, and the distribution of streamlines in the river. (And the rendering of the streamlines, though they look good in the original PDF; it seems to be a conversion issue.) If you can give some examples of what I could twiddle manually to substantially improve things, I'd love to hear them!

I've done some more work this morning, and I think I'm pretty much done with this inset image; I might put some of the larger building outlines on it, once I've placed them, depending on how cluttered it looks.

So, I guess this is my ### Latest WIP ###:
Ichenjanda-02.png

[msa]

Doing things by hand makes them so much more ... controllable. The problem is that I don't really have a clear picture in mind of what I'm after, so I have to experiment to see what I want, and it's much easier to mix-and-match experiments when they're procedural

You should definitely keep doing the procedural stuff! Its amazing!

I'm more talking about the last week or so, for the finishing touches. Like... last month with your river widths. What you did procedurally was fantastic, but you might could have gone in by hand and touched that up and taken it up a notch.

I'm still jealous of your map generation. Its a problem I've thought about a lot in my other hobby as a roguelike programmer.

[ravells]

I know what you mean, though. Doing things by hand makes them so much more ... controllable. The problem is that I don't really have a clear picture in mind of what I'm after, so I have to experiment to see what I want, and it's much easier to mix-and-match experiments when they're procedural -- when the forest shading depends on the forest outline depends on the contours depends on the initial sketch, I can fiddle with the sketch and immediately see the change in the forest shading. (To some extent, it's also the fact that I have no real artistic talent.)

That pretty much describes me and my methods to a tee!