Jump to content

olli_

Member
  • Content Count

    102
  • Joined

  • Last visited

  • Medals

  • Medals

Posts posted by olli_


  1. Thats just how the cockpit looks.. im guessing its not much of a problem with the real version with the DAS so you can just see through the whole cockpit.

    I did some ambient occlusion tweaking in an effort to make add some depth to the cockpit, the changes are very subtle but i think its going in the right direction:

    http://i.imgur.com/gnR4d0I.jpg (155 kB)

    also most of the album pictures were taken with the zoomed out view (i was holding -) so you can see more of the cockpit with less pictures. This picture is taken with the default FOV.


  2. alright i did some testing and ive outlined my results on polycount.. might as well copy the wall of text here if anyone finds any help in it:

    alrighty. I decided to break out the "insane graph editing stuff" after all since nothing else was working. I created a control mesh which was basically a flattened cylinder but with the normals preserved (so a flat plane with a round normal surface)

    I baked normals on to this control mesh from a flat plane, so in max the normals were working as intended and the control mesh was rendering as a flat plane with the normals.

    Next I moved on to oxygen where i realized i cant export explicit vertex normals with the FBX importer (im pretty sure its oxygens fault since i double checked that the normals are indeed exported with the fbx) BUT i managed to import vertex normals with the .obj importer, although even then i had to manually set all the normals to explicit (edit normals>parameters>make explicit).

    Now the only problem is that the .obj importer is a piece of shit compared to the FBX importer, but thats not relevant to the tangent space topic so ill leave that rant for another day.

    now that i had a control mesh, and a normal map which should make the mesh flat, i moved on to photoshop and started a long process of trial and error with the curve editor.. Here's what i came up with:

    PNk6l6j.png

    With this curve, the normal map correctly flattened my control mesh normals. (at least within the available spectrum)

    as you can see almost half of the spectrum is completely unused. If I hadnt spent the last month bashing my head against the wall with this engine I'd have discarded these results as absurd but quite frankly im not the least bit surprised.

    (all values below are approximate)

    To explain the curve a little more, basically what it does is instead of the usually -90 to 90 degree range of normal maps, the range is a lot less. Around -50 to 50 for 0-255. Additionally the "gradient" isnt linear, its slightly curved at the ends. The scale seems linear between around 95-160 and from there it starts to loosen, so one grayscale unit equals a smaller amount of degrees than it does within 95-160. My curve corrects that for the most part, but I made it completely by hand so its obviously not 100% accurate.

    Now if you'll excuse me i have a date with the obj importer.

    edit: I had a lot of trouble getting the obj importer to work properly (invertex faces, mirrored normals, all kinds of creepy stuff) so i might as well post solutions i found here. in order to import a model properly with the .obj importer, first i had to rotate my mesh 180 degrees around the X axis (so its upside down) then I reset xform and exported the .obj with all optimization options off, flip YZ on and all the normal stuff ticked. (also export as quads, its easier to triangulate in o2 since results may vary so you dont have to re-export)

    In o2 you just import the obj with FLIP Y ticked. I also had to triangulate \ (NOT triangulate / as a lot of people in this thread claim to be the correct way)

    the reason you have to rotate the mesh 180 degrees before you export is because if you dont have flipY ticked in the importer, your faces will be flipped, and if you dont have the mesh rotated 180 degrees, flipY will import your mesh upside down, and if you try to rotate it in o2 the mesh will rotate but the normals wont, so itll FUBAR.

    i would give anything to meet whoever implemented this pipeline so i could give them a real good slap on the face.


  3. Im trying to get my normal maps to look good in in buldozer but the normal space is really weird. My normals look fine in 3ds max, but in bulldozer it doesnt render out properly.

    http://i.imgur.com/AKTcLnH.png

    (left is 3ds max and right is buldozer)

    some info:

    model is exported as fbx 2010

    normal map is 32 bit tga (tried 24bit as well) with _NOHQ suffix

    ive tried all output options in handplane with an object space normal but none of them worked in arma. source engine was closest i think but still noticably off)

    have tried triangulate / and triangulate \ in O2. neither worked.

    have triangulated my fbx before i exported but for some reason in the O2 fbx importer makes it back in to quads

    edit: is there some way to import explicit vertex normals? that might help


  4. Hey are there any rules in RV4 with regards to splitting a model in to materials? I read somewhere that each moving part should have its own material, but im able to apply a single material to my whole model with lots of moving parts just fine.

    also is there a connection between selections and materials? something like 1 selection = 1 material?


  5. What im trying to do is get the camera direction vector and use it in various HUD elements in class MFD.

    the existing HUD functionality is extremely limited considering modern technology (not to mention FUTURE technology in 2035) and the only way i can see of expanding that is scripting, and even that wont be possible if i cant use scripted variables in the MFD class..


  6. Hey everyone. finally made an entry on the MANW website this week and figured id also make a thread here.

    http://makearmanotwar.com/entry/Za7p8ntdue

    So yeah im making an F-35C for the contest, additional info can be found in the entry description above ^

    Im completely new to Arma modding, and the engine isnt the easiest to use (to say the least), but im making slow and steady progress like a snail riding on a turtle.

    Im primarly a 3d artist and have zero coding experience (which is cool because pretty much everything in this engine is made with code) :D

    The plane model itself is completed, but i havent baked normals or textured yet (saving that for last because its the easy part)

    however, I have completed some missile models for your viewing pleasure.

    http://i.imgur.com/xMsBA2p.jpg (110 kB)

    also heres a panorama of the High poly cockpit. The image is like 10000x5000 or something, so it may take about forever to load, and due to my nonexistent js knowledge, my panorama viewer might not be very stable or compatible with all browsers. I take no responsibility for anything.

    http://triangulatemesh.com/envmap_viewer.html


  7. So the rules seem to say that any addon submission must work 100% with just a basic Arma 3 installation (im interpreting that as it cant require other 3rd party addons)

    Does this mean that i cannot have any connection to 3rd party mods without being disqualified, or can i have "additional" features that require a 3rd party addon but that are not integral to my addon.

    for example i have a midair refueling nozzle on my plane, and it would be cool to make it compatible with some 3rd party flying tanker if people wanted to use it.


  8. alright thanks for all the help guys, ive managed to solve the problem.

    for any future generations struggling with this:

    it seems that the ground contact points in LandContact LOD cannot be named the same as the wheels themsleves (despite me reading from MULTIPLE sources on these forums that the LandContact LOD does NOTHING and has NO EFFECT on ANYTHING), so if your wheels are wheel_1, wheel_2, and wheel_3, the landcontact points have to be something else. I named mine wheel_1_stabil_1, wheel_2_stabil and wheel_3_stabil. Thats what they are named in the sample plane but im not sure if they absolutely must be named that, and some other community mods seem have the landcontact points named differently so go figure.

    I also moved all the LODs in my p3d so the plane's center of gravity is closer to the 0,0,0 point of the scene. I dont know if this has any effect but the sample plane is set up like this so im just going with the flow. Originally i had the plane's wheels aligned with the Z = 0 plane (which seems only logical for a vehicle that moves on the ground, but i should already have learned that logic is worth jack shit when dealing with this engine.)

    my theory on why this works:

    I have absolutely no idea


  9. ok so a few more things ive figured out,

    the game seems to think my airplane is in the air even when its on the ground. the wheels dont roll when i move (theyre setup correctly ive tested in buldozer, and they used to roll before i renamed my wheel geometry proxy as componentXX) and I can retract the landing gear even while on the ground, so the plane just lowers down on to its belly. I can also re-extend them, which makes the plane stand up back on to its legs (it looks pretty funny). Basically the plane spawns in the "on the ground" mode, which means i can get out and cant retract landing gear, but once i move even slightly, the plane enters "flight mode".

    still got no idea what is causing this :/

    edit: just to clarify, there are 3 wheels, each has its own geometry LOD mesh (cylindrical, plenty of sides, like 18 i think) and each is named wheel_1, wheel_2, wheel_3 and also Component19, Component20, Component21

    edit2: i removed the Component19-21 selections for the wheel Geom and this does fix steering, and the plane is correctly simulated as "on the ground" but now the wheels have no collision and just sink in to the ground... im able to replicate all the functionality i need, just not all at the same time


  10. is there a way to define fresnel using a texture instead of just having a single value for the whole material?

    im trying to make an object which has a few chromed parts, but theyre quite small and integrated in to the general topology so i cant just split the selection and use a second rvmat

×