waldronate's wilbur is a good tool
BUT-- yes there is always one
north and south of +60 -60 wilbur dose not work great on a planet map
you need to remap the poles to polar stereographic ( +45 to +90 and -45 to -90 )
erode then and remap back to simple cylindrical and BLEND!!! it back into the 16 bit or 32 bit float heightmap
it really IS HARD to do a map 90 north to 90 south latitude and -180 to +180 longitude ,and have it look good .
most of the time i only run 3 or 4 erode passes in wilbur, more than that it starts looking a bit obvious
also if you look at my examples you will notice i do a gama point move to make the lowlands HIGHER
-- looks better when rendering most of the time BUT is not realistic
recently i have been using Blender nodes to make matching sets of
texture
heightmap
and city lights - this is iffy??? still
http://celestiaproject.net/forum/vie...135033#p135033
and
http://celestiaproject.net/forum/vie...hp?f=5&t=17466
and
http://forum.celestialmatters.org/vi...4&t=830#p13877
also the OLD "fracplanet" dose a decent base heightmap that will NEED a lot of work to render well
-- the codebase needs updating
wilbur has some good spherical noise maps that can be made and with time mixing them into something that loos well enough
there is also the VERY OLD "libnoise" tutorials in coding
it makes a ok base map to start with
as for REALISTIC , that takes TIME a LOT of CPU or GPU cycles and tools used by geologists and researchers
there are 2 GRASS plugins that after a bit can be ran in Qgis but running GRASS is a lot easier
and you WILL need all night to run the simulation while you sleep
( work on SMALL areas while testing then use a big map )
r.terraflow and r.landscape.evol
https://grass.osgeo.org/grass72/manu...terraflow.html
and
https://grass.osgeo.org/grass70/manu...cape.evol.html
a good web site for these
http://isaacullah.github.io/GRASS/
or the " topoToolBox" for Matlab ( VERY EXPENSIVE AND VERY SLOW single threaded )