Well, the way I imagine it, a planet creator would be a large enough feature
Currently we are simply compiling the existing Earth datasets into a format suitable for use in OT. Custom artificial planets that would
reach the quality level of Earth are hard to make though. You can ask the
ME-DEM guys, who, after running procedural generators and erosion algorithms for one week on a patch of the size of Europe ended up making their world by blending real-world mountains into their maps because even a week of processing couldn't get them anywhere near the quality and detail of the real world.
The reason for it lies both in the existing tools and approaches and in the fact that it's really much more complex than simply using a fractal algorithm to make an artificial ridged-perlin worlds. These fractal worlds can be nice, in an artificial style, but they come with a fault that is also sold as their main strength - self-similarity. Fractals are really build on the concept that things
tend to be self-similar at multiple scales, but in reality that's a huge simplification and a huge loss of information. Following the concept you get worlds that are just that - similar to each other, with similar structures everywhere. Might be interesting for a while, but it simply cannot compare to the real world.
Since the large terrain features are formed by different processes than the smaller ones, our long-term goal is to use a set of procedural techniques each built specifically for given detail level. And also to have means of artistic input into the process, where parts of each stage could be provided in a coarse and to-be-refined form externally.
So, all in all, it can be a pretty complex thing. But like I said, our focus is on the Earth and real-world locations, also because the simulation industry is interested only in the real worlds, so the artificial world maker will come only after that.