Different look after upgrading to TGE 1.5 (Dif-format)
by Klaus Riesterer · in Artist Corner · 10/26/2006 (4:11 am) · 13 replies
First I will say hallo to all of you, because I'm new in this forum and also to Torque.
Using TGE 1.4 for a few days I have no recent probems, but now using 1.5 there is a question about the DIF-format. In 1.4 all looked fine but now there are unwanted shadows (breaks in texture) on my dif-objekt(s):
Can anyone explain where I make my mistake?
The Dif was created wit Quark/ Map2Dif
Thx
Klaus
Using TGE 1.4 for a few days I have no recent probems, but now using 1.5 there is a question about the DIF-format. In 1.4 all looked fine but now there are unwanted shadows (breaks in texture) on my dif-objekt(s):
Can anyone explain where I make my mistake?
The Dif was created wit Quark/ Map2Dif
Thx
Klaus
#2
In TGE 1.4 the same object shows fine, maybe there is any lightning property (TLK)?
Greetings
Klaus
10/26/2006 (4:37 am)
I'll give this a try, but I want to use not DTS for a bridge building ...In TGE 1.4 the same object shows fine, maybe there is any lightning property (TLK)?
Greetings
Klaus
#3
sorry Klaus... i haven't had my first cup of coffee yet... i'm still half asleep...
diregard, as my post is really not of any help at all... again, sorry...
--Mike
10/26/2006 (5:07 am)
Oooops... yeah.... dog gone it, i totally read DIF and responded DTS...sorry Klaus... i haven't had my first cup of coffee yet... i'm still half asleep...
diregard, as my post is really not of any help at all... again, sorry...
--Mike
#4
Try using integer precision or using a snap to grid/vert tool (if available, I'm not too familiar with Quark).
10/26/2006 (5:12 am)
It looks like the interior was modeled in Quark with floating point precision enabled. This can cause surfaces to improperly align (the adjacent surfaces do not meet up properly and do not share verts). TGE 1.5's new lighting system uses adjacent surface information to avoid shadowing from surfaces that are a continuation of the surface being lit.Try using integer precision or using a snap to grid/vert tool (if available, I'm not too familiar with Quark).
#5
I going try more around with this - hopefully to find a solution.
Klaus
10/26/2006 (2:19 pm)
I've looked again and can't find any errors. No double or near vertexes, no holes. Grid-Snap is anabled at gridsize of 4. Maybe there is an error in texture orientation or size but this should not result in this manner.I going try more around with this - hopefully to find a solution.
Klaus
#6
If yes: Good, let me know please
If not: Thank you for helping so far
10/28/2006 (9:56 am)
No solution found till now ... any other ideas?If yes: Good, let me know please
If not: Thank you for helping so far
#7
10/28/2006 (12:47 pm)
Try importing the map into Hammer, fixing any misaligned brush seams, then use the updated map (without opening it in Quark) to build your dif. Those edge shadows are consistent with the Quark floating point issues mentioned earlier.
#8
10/28/2006 (12:58 pm)
You might upload the source file for some of us to take a look?
#9
Here are the files
I hope there is all you need in the zip.
Regards
Klaus
10/28/2006 (1:24 pm)
Sorry I didn't thought anything ...Here are the files
I hope there is all you need in the zip.
Regards
Klaus
#10
1)Go ahead and cut out the inner part (the pathway)
2)Not all of the textures are properly tiled, so that may be part of the problem. Start at one end and tag the face (right-click on the face's widget and hit "tag face"). Then go to each polygon on the path and right-click then texture->wrap texture from tagged (you may have to hunt around for this command).
3) recompile, make sacrifices to to the QuArK gods.
That worked for me in 1.5. Now, my settings may be different than yours. I know I set my floating points decimal pretty high, so if the above doesn't work you may have to go do that (I'm not responsible if it causes new problems ;) ).
Good luck!
-Eric
10/28/2006 (8:46 pm)
I did these two things and it seemed to work fine (using map2dif_plus and QuArK 6.5 alpha 8). Forgive me if you already know how to tag/wrap:1)Go ahead and cut out the inner part (the pathway)
2)Not all of the textures are properly tiled, so that may be part of the problem. Start at one end and tag the face (right-click on the face's widget and hit "tag face"). Then go to each polygon on the path and right-click then texture->wrap texture from tagged (you may have to hunt around for this command).
3) recompile, make sacrifices to to the QuArK gods.
That worked for me in 1.5. Now, my settings may be different than yours. I know I set my floating points decimal pretty high, so if the above doesn't work you may have to go do that (I'm not responsible if it causes new problems ;) ).
Good luck!
-Eric
#11
*Edit:
OK, tested now:
Overall that is not the solution - looks same after doing these steps.
Have you tested it with my source? Could you please send me the dif then?
So we can see where the crux is.
Where did you set the decimal precision?
Sorry, if my questions are a bit silly, but there are mutch new things for me. ;-)
Regards
Klaus
10/29/2006 (2:52 am)
Never tagged anything in Quark - maybe thats the reason. Thank you for the answer.*Edit:
OK, tested now:
Overall that is not the solution - looks same after doing these steps.
Have you tested it with my source? Could you please send me the dif then?
So we can see where the crux is.
Where did you set the decimal precision?
Sorry, if my questions are a bit silly, but there are mutch new things for me. ;-)
Regards
Klaus
#12
Simple answer: In QuArK with your better-textured version, go to Options->Configuration...->map-Options and check Use Integral Vertices as Threepoints.
This seems to make things better, though not perfect.
More complex answer (For John's benefit, as well):
You can download the newer QRK, DIF and MAP files here (temporarily).
-All tries look good in 1.4x version of TLK
-In order to rule out John's suggestion of making sure all the points are on the grid. The "force to grid" option doesn't work well and still leaves small offsets, so I checked each point (as far as I know). There were quite a few that were off a tiny bit, like 0.0002, so I made them whole numbers.
-I made sure all textures were wrapped seamlessly as well as I could.
-Tested various Detail brush arrangements (no luck)
-Tried with and without vertices as threepoints.
-Tested in both map2dif_plus and Synapse Gaming's map2dif. At first, map2dif_plus seemed to work better until I turned off threepoints then they were close to same.
-Turned off floating points which made for odd imperfections, even with the on-grid version.
Bottom line: Texture wrapping, vertices as threepoints, and grid snap all seem to affect quality of the shadowing (my theory is that texture wrapping affects the model because of the way polygons are created in map2dif, though I have nothing to back that up with). The biggest change came from turning off threepoints.
Unfortunately, even after all of this there's still a faint shadow that can be detected on either slope.
10/29/2006 (3:55 pm)
OK, I spent a few hours with this and did a lot of testing. Simple answer: In QuArK with your better-textured version, go to Options->Configuration...->map-Options and check Use Integral Vertices as Threepoints.
This seems to make things better, though not perfect.
More complex answer (For John's benefit, as well):
You can download the newer QRK, DIF and MAP files here (temporarily).
-All tries look good in 1.4x version of TLK
-In order to rule out John's suggestion of making sure all the points are on the grid. The "force to grid" option doesn't work well and still leaves small offsets, so I checked each point (as far as I know). There were quite a few that were off a tiny bit, like 0.0002, so I made them whole numbers.
-I made sure all textures were wrapped seamlessly as well as I could.
-Tested various Detail brush arrangements (no luck)
-Tried with and without vertices as threepoints.
-Tested in both map2dif_plus and Synapse Gaming's map2dif. At first, map2dif_plus seemed to work better until I turned off threepoints then they were close to same.
-Turned off floating points which made for odd imperfections, even with the on-grid version.
Bottom line: Texture wrapping, vertices as threepoints, and grid snap all seem to affect quality of the shadowing (my theory is that texture wrapping affects the model because of the way polygons are created in map2dif, though I have nothing to back that up with). The biggest change came from turning off threepoints.
Unfortunately, even after all of this there's still a faint shadow that can be detected on either slope.
#13
It really looks much better than before! :)
Use Integral Vertices as Threepoints was not set in my Quark - now its default.
So for the moment there is no concluding solution, right? - OK, have to wait if there will be a workaround on Quark or map2diff_plus wich will help to clean up this.
Are there other tools witch will do a better job here?
Regards
Klaus
10/29/2006 (4:25 pm)
Eric, thanks for spending a lot of time and testing this all !It really looks much better than before! :)
Use Integral Vertices as Threepoints was not set in my Quark - now its default.
So for the moment there is no concluding solution, right? - OK, have to wait if there will be a workaround on Quark or map2diff_plus wich will help to clean up this.
Are there other tools witch will do a better job here?
Regards
Klaus
Torque Owner Michael Hense
if you have access to Milkshape, can you import the model into Milkshape and export it to dts from there, and see if the breaks still exist?
i've run into what looks like the same sort thing with models i export from Ultimate Unwrap... when i run em through Milkshape exporter it seems to disappear...
--Mike