Working on rewriting core logic using C++ to make it completely independent from Unity3D and increase performance. Also support of multiply zoom levels are planned (e.g. render Earth as globe on zoom level one and buildings/all details in 3D on highest level).
Just first raw results showing chosen direction - rendered globe at zoom level 1:
No texturing yet, color gradient logic is slightly broken, but I was able to use coastline data (from natural earth shapefiles) and mesh generation is quite fast.
Current plan is start from releasing this as FREE unity asset for desktop and mobile platforms (e.g. Android) when all major features are implemented and it is stable (cannot estimate so far the concrete date, but for sure I need several months). I’m already aware about making it easy to use and considering this as a MUST requirement.
Nice thing, that core part has no dependency on Unity at all as I’m writing it using C++11 without any Unity or platform specific API. You call something like loadTile with quadkey, desired mapcss style + callbacks which are called to return you back generated mesh or raw data (node/way/relation for given region if you want to process it from your code). Then is up to specific engine to build “game object” from this data.
Theoretically it can be reused in some other 3D gaming engine or visualization software.
Here, text and sphere for each place are created on unity3d side from geodata provided by the library. Of course, visual representation can be improved: text from opposite side of globe should not be visible, but this is just example.
I was able to create Android build (armeabi-v7a). It contains so far some hardcoded values (sdcard path for offline map data, start geocoordinate), but shows that it is working and performance quite good even without any additional improvements which I’m pretty sure I can do.