Posted: Wed Apr 28, 2010 8:29 pm
Cmd. Cheyd quietly waits and prays for the Windows version... Or a Photoshop script... Or Plug-in... Or...
For information and discussion about Oolite.
https://bb.oolite.space/
What I said above is absolutely true: I wrote the graphical version because it was easier than writing a comprehensible guide to installing and using command a line tool. It took a few hours of coding on Sunday evening, and another hour or so of polishing spread out over several days. The Windows and Linux programmers in the audience are welcome to take this as a challenge. I’ve already made it easier for them by refactoring the core code (which is vanilla C) to be more embeddable. ;-)Cmd. Cheyd wrote:Cmd. Cheyd quietly waits and prays for the Windows version... Or a Photoshop script... Or Plug-in... Or...
Each face of the cube must be a power-of-two square, so the height will always be 1.5 times a power of two.Griff wrote:does the texture size power of two rule apply to cube maps? i've used this program to convert some 1024x512 'latlong' maps and the resulting cube map comes out as 512 x 3072. Do i need to resize this now to 512x2048 or will Oolite/openGl deal with this map as 6 512x512 tiles?
<chortles>London sits in a steamy jungle straddling the equator, with a climate generally resembling Manila's. The food is still bland, the Thames is full of piranha, and it's the only place on Earth where tigers apologize as they attack you.
The architecture isn’t very well suited to that. You could quite easily make a filter (a “source”, in planettool terms, that reads from another source, like MatrixTransformer does), but the renderers (“sinks”) perform Gaussian weighted filtering that doesn’t necessarily make sense for normals.submersible wrote:I've been meaning to look at the source for planettool to see if I can devise a normal map generator for cube maps. I noticed that the Oolite planet texture generator builds a normal map by sampling a noise map (lat-long) . For me - building a useful normal map for use on a spherical object from latlong or cube maps has been devilishly hard with current tools. The GIMP normal map extension is OK - but it is planar centric.
kFPMGammaLinear
instead of kFPMGammaSRGB
to FPMWritePNG
.Code: Select all
FPMColor HeightMapToNormalMapFilter(Coordinates where, RenderFlags flags, void *context)
{
HeightMapToNormalMapFilterContext *cx = context;
// Read input coordinates, possibly converting from vector to latitude/longitude.
float lat, lon;
CoordsGetLatLongRad(where, &lat, &lon);
// Offset coordinates northward by some angle, presumably specified at command
// line and defaulted to something sensible based on size.
Coordinates northOffsetCoords = MakeCoordsLatLongRad(lat + cx->offset, lon);
FPColor northSample = cx->source(northOffsetCoords, flags, cx->sourceContext);
// Three more samples here...
// ...Then calculate normal from gradients.
}
Which one would that be? There aren’t any tickets for planettool in the bug tracker.Cmd. Cheyd wrote:Any chance of getting the bug I reported last year (or was it the year before that?) fixed?
I'm noticing this going from a cube to a cubexJensAyton wrote:While it can read from cube maps, it fails at the edges of faces and produces artefacts. This does not affect writing cube maps.