Realistic Mountain Heightmap Making

[Redrobes]

As far as I know there is nothing that creates *really* realistic procedural height maps of eroded mountains. But theres many apps which try and some do quite well though you can usually tell if you look hard at them.

For free try Wilbur and there are some free noise map generators out there which you can get some milage out of.
Of the paid ones I think theres L3DT, Terragen, Fractal Terrains to name a couple tho there are many.
The most realistic that I have seen would be WorldMachine. But some apps are better at some types of rock and mountains than others. I think thats the rub of it. I dont know of an app that could generate the grand canyon and some of those wind eroded pillars for examples. Theres plenty of procedural generators that use real world terrain as a basis and they can do a better job but thats not entirely procedural.

Personally, I like to take a large mountainous region of height map and blend a load together then make it seamless once and keep that on hand. Then when needed, apply it to a graduated mask. Making the mask by hand is a lot easier than making the whole mountain. So not procedural at all but it looks very realistic since it came from real world height map data in the first place. You can keep several types of seamless real world data sets for different types of terrain that way.

Found a link to a list of them:
http://vterrain.org/Packages/Artificial/

[waldronate]

For free try Wilbur

https://cartographersguild.com/showthread.php?t=29412 is a good starting point for this.

There are several major problems with existing terrain generators, in my opinion. The terrain that we see around us is an emergent property of several forces, all acting at microscopic levels. What software generates is an approximation of those processes (that is the software uses a model or set of models with very specific assumptions). The output of most software packages has a range of plausibility for the outputs.

Wilbur's erosion models, for example, mostly generate outputs that are plausible for the scale range of about 1 meter per pixel to about 100 meters per pixel and it has quite a few limits on the range of altitude variance over which it will generate plausible results. This range of plausibility comes from the models that Wilbur employs, which are basically the simplest thing that I could code for the kinds of maps that I was making (precipiton erosion, for example, is constant-strength materials and only nearest-neighbor sediment transport). Wilbur can generate fairly plausible results if its limitations are kept in mind. It can generate some marginally plausible mid-scale mountains if you're willing to phrase the problem in a certain way ( https://www.cartographersguild.com/a...chmentid=80877 from the Muna Workshop thread is an example ).

As Redrobes suggested, cloning existing terrain patches (especially with a texture synthesis algorithm behind it) is a good way to go if you're not concerned too much with river flow and just want to get results that look plausible at first glance.