Any recommendations for exporting heightmaps out of World Machine… specifically into 16bit .tga, .tif, or .png formats?
Current export options are not working with PhotoshopCS2 or Lightwave. I must output as .raw FP32 into HDRShop, save high precision .tif, open in photoshop and convert to 16bit .tif and then use it in my 3D software…
Hardly a streamlined process. I’ve tried Wilbur as a conversion utility, but I’ve only been able to open .ter or .bt formats out of WM (and I think they open as 24bit rgb files that must be converted to grayscale heightmaps and I lose the high precision output).
The .png and .tif 16 bit options you plan for the future would be most welcome in the present.
Otherwise… I’m having a great time with a very interesting piece of software!
I have been working with WM and CS2 and it seems to work fairly seamlessly.
I output .r16 files which are then opened in photoshop as a raw image.
When opening them as raw images in PS, make sure that you use the following settings:
Width and height set to the resolution you rendered your WM files out at
Channels count = 1
Depth = 16bit
Byte order is IBM PC
Header is 0 bytes.
when writing out files to go back into WM, more or less reverse the process. Make sure you have a single channel image… and you have to use the .raw extension when saving, but you must rename it .r16 prior to opening it in WM.
Quite frankly, I would be surprised if you need more detail than you can get out of 16bits. As Oshyan can attest to, I made the mistake of thinking I needed more at the beginning myself but he was able to clarify things quite quickly…
16bits = 64K steps of resolution. Assuming you have some kind of monster machine which can actually generate the holy grail of 8K maps, that means you will still have 8 times more vertical resolution than you have horizontal res. If you stick to the more achievable 4K maps, then you will have 16 times more vertical resolution than horizontal. In terms of normal usability, that seems to be adequate, especially at the detail levels you would expect from mountains.
Then again, I am quite often wrong, so don’t let me dissuade you from trying. There might well be a very real reason to go to a higher bit depth - but I am not sure how WM works internally (if it is all 16 bit internally then you gain nothing. If it is float internally then it would be nice to actually have the extra detail).
It would be helpful to know what 3D package your using, along with what rendering software.
Thanks for your interest. For the present, I am using lightwave 8.5 for building my initial landscapes… with rendering in maxwell render. Once modo201 comes out, I am sure I will be making less use of lightwave…
So far, the responses supplied have taken care of most of my concerns… now I need to figure out how to “tile” my heightfields as I have run into memory limitations with some of my landscapes in WM (I’ve a 3.8GHz P4, 3Gb ram, nVidia1400, winXPpro).
I’m out of town now but will be back at my studio by Wednesday…
One technique we have found that seems to work fairly well is to create a library of mountains at about 1K. We then place, rotate, and scale them in photoshop where we want them to be (which allows us a fair bit of control as far as art direction is concerned). Once we have them all placed, we then add a little terrain around the base in WM, erode the whole thing a tiny bit to merge them, and render out a 4K tile (can’t seem to get above that without crashing). Rendering out tiles then works reasonably well as long as we take the edges back into PS and fix them up (combine multiple tiles into an 8K image, fix the edges, and then cut back into 4K sections). Then we can render the whole thing through Renderman.