Quote Originally Posted by Redrobes View Post
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.