Creating LOD Models - Revision history

Nbohr1more: /* The material definition */

2017-05-14T16:00:50Z

The material definition

← Older revision		Revision as of 16:00, 14 May 2017
Line 113:		Line 113:
	blue global4		blue global4
	}		}
	{
			}
			</pre>

			Note: This section:

			<pre>

			{
	if (global5 == 2)		if (global5 == 2)
	blend add		blend add
Line 126:		Line 134:
	fragmentMap 3 _black // Specular		fragmentMap 3 _black // Specular
	}		}
	}
	</pre>		</pre>

			has been removed as it is no longer needed as of TDM 2.0 (Standalone)

	'''Note:''' If you make the material two-sided, then you can save creating the "backside" patch. However, D3 will do so when it loads any model using this material, and the "backside" is created as an extra surface. This means it will use up one extra [[drawcalls\|drawcall]]. To avoid this, we create a one-sided material, and create the backside ourselves:		'''Note:''' If you make the material two-sided, then you can save creating the "backside" patch. However, D3 will do so when it loads any model using this material, and the "backside" is created as an extra surface. This means it will use up one extra [[drawcalls\|drawcall]]. To avoid this, we create a one-sided material, and create the backside ourselves:


	=== Create the patches, export as ASE ===		=== Create the patches, export as ASE ===

Tels: add links

2011-07-01T08:37:54Z

add links

← Older revision		Revision as of 08:37, 1 July 2011
Line 203:		Line 203:

	* Without a texture atlas, each patch uses its own material, thus creating two surfaces, using one [[drawcalls\|drawcall]] each.		* Without a texture atlas, each patch uses its own material, thus creating two surfaces, using one [[drawcalls\|drawcall]] each.
	* With a two-sided material, D3 will upon load duplicate all surfaces and invert their side. This means you end up with 2 drawcalls instead of one.		* With a [[twosided\|two-sided]] material, D3 will upon load duplicate all surfaces and invert their side. This means you end up with 2 drawcalls instead of one.

	Both issues together will create '''four drawcalls per model per light''', instead '''one per model per light'''. Since each model is drawn multiple times (once for each light), this can add up very fast.		Both issues together will create '''four drawcalls per model per light''', instead '''one per model per light'''. Since each model is drawn multiple times (once for each light), this can add up very fast.
Line 251:		Line 251:
	* The [[LOD]] system		* The [[LOD]] system
	* [[List of LOD Models]]		* [[List of LOD Models]]
			* [[SEED]]
			* [[twosided]] - Visual explanations on how twosided surfaces work
	* http://en.wikipedia.org/wiki/Texture_atlas		* http://en.wikipedia.org/wiki/Texture_atlas
	* [http://bloodgate.com/mirrors/tdm/pub/lod_construction.zip Helper maps with black room and fixed camera pos]		* [http://bloodgate.com/mirrors/tdm/pub/lod_construction.zip Helper maps with black room and fixed camera pos]

	{{editing}} {{scripting}}		{{editing}} {{scripting}}

Tels: fix material

2011-02-16T22:45:27Z

fix material

← Older revision		Revision as of 22:45, 16 February 2011
Line 122:		Line 122:
	vertexParm 3 0		vertexParm 3 0
	fragmentMap 0 cubeMap env/gen1		fragmentMap 0 cubeMap env/gen1
	fragmentMap 1 ~~_black~~ // Bump		fragmentMap 1 _flat // Bump
	fragmentMap 2 textures/darkmod/decals/vegetation/cattail_flat_side // Diffuse		fragmentMap 2 textures/darkmod/decals/vegetation/cattail_flat_side // Diffuse
	fragmentMap 3 _black // Specular		fragmentMap 3 _black // Specular
	}		}
	}		}

Tels: add links to drawcall

2011-01-30T16:19:51Z

add links to drawcall

← Older revision		Revision as of 16:19, 30 January 2011
Line 1:		Line 1:
	== Introduction ==		== Introduction ==

	~~TDN~~ v1.03 ~~will have~~ a LOD system, where the engine can replace models that are far away with simpler versions. The same technique can also be applied manually, by placing decals at some distance to create the illusion of more details.		TDM from v1.03 onwards has a LOD system, where the engine can replace models that are far away with simpler versions. The same technique can also be applied manually, by placing decals at some distance to create the illusion of more details.

	This article details how to create models that can be used for such purposes and what the performance gains are.		This article details how to create models that can be used for such purposes and what the performance gains are.
Line 129:		Line 129:
	</pre>		</pre>

	'''Note:''' If you make the material two-sided, then you can save creating the "backside" patch. However, D3 will do so when it loads any model using this material, and the "backside" is created as an extra surface. This means it will use up one extra drawcall. To avoid this, we create a one-sided material, and create the backside ourselves:		'''Note:''' If you make the material two-sided, then you can save creating the "backside" patch. However, D3 will do so when it loads any model using this material, and the "backside" is created as an extra surface. This means it will use up one extra [[drawcalls\|drawcall]]. To avoid this, we create a one-sided material, and create the backside ourselves:


Line 198:		Line 198:
	== Optimizing ==		== Optimizing ==

	If you haven't followed the steps about a ''texture atlas'' and created a two-sided material, you will notice that the model uses 4 drawcalls in game, compared to the original model. That means placing many of the new models is actually slower than placing the original models, despite the greatly reduced number of triangles.		If you haven't followed the steps about a ''texture atlas'' and created a two-sided material, you will notice that the model uses 4 [[drawcalls]] in game, compared to the original model. That means placing many of the new models is actually slower than placing the original models, despite the greatly reduced number of triangles.

	The reasons are:		The reasons are:

	* Without a texture atlas, each patch uses its own material, thus creating two surfaces, using one drawcall each.		* Without a texture atlas, each patch uses its own material, thus creating two surfaces, using one [[drawcalls\|drawcall]] each.
	* With a two-sided material, D3 will upon load duplicate all surfaces and invert their side. This means you end up with 2 drawcalls instead of one.		* With a two-sided material, D3 will upon load duplicate all surfaces and invert their side. This means you end up with 2 drawcalls instead of one.

Line 243:		Line 243:
	In my first test, I also did a "top-side" view, and placed this as patch at the bottom, horizontally. The theory was that there is something the player can see if he looks straight from the top.		In my first test, I also did a "top-side" view, and placed this as patch at the bottom, horizontally. The theory was that there is something the player can see if he looks straight from the top.

	However, in game tests showed that it only looked funky (often z-fighting with the floor), and since the player is far away from the model, anyway, he can't see it from straight top. If you view it high above, it will still look quite good due to the patch being vertical and thus partly visible. So, just having a front and side few is sufficient.		However, in game tests showed that it only looked funky (often z-fighting with the floor), and since the player is far away from the model, anyway, he can't see it from straight top. If you view it high above, it will still look quite good due to the patch being vertical and thus partly visible. So, just having a front and side few is sufficient for high models.

			However, if the model is very small (like only 20cm high) and the player can walk over the model and view it straight from the top, you should consider adding a horizontal plane, too. Make sure you lift if 0.0125 units above the floor, tho, or it will z-fight with the floor!

	== See also ==		== See also ==

Nbohr1more at 01:02, 26 October 2010

2010-10-26T01:02:28Z

← Older revision		Revision as of 01:02, 26 October 2010
Line 248:		Line 248:

	* The [[LOD]] system		* The [[LOD]] system
			* [[List of LOD Models]]
	* http://en.wikipedia.org/wiki/Texture_atlas		* http://en.wikipedia.org/wiki/Texture_atlas
	* [http://bloodgate.com/mirrors/tdm/pub/lod_construction.zip Helper maps with black room and fixed camera pos]		* [http://bloodgate.com/mirrors/tdm/pub/lod_construction.zip Helper maps with black room and fixed camera pos]

	{{editing}} {{scripting}}		{{editing}} {{scripting}}

Tels: add link to special maps

2010-09-30T12:46:30Z

add link to special maps

← Older revision		Revision as of 12:46, 30 September 2010
Line 28:		Line 28:
	[[Image:Cattails green 4 with patch.png\|right\|thumb\|240px]]		[[Image:Cattails green 4 with patch.png\|right\|thumb\|240px]]

	I put the model into a map, removing everything that was visible behind it.		I put the model into a map (see at the [[#See also\|bottom]] for an archive with pre-made maps with black rooms), removing everything that was visible behind it.

	Then either:		Then either:

	* just use the grey background from the DR camera window if your model doesn't contain this sort of grey		* just use the grey background from the DR camera window if your model doesn't contain this sort of grey
	* '''OR''' then create a simple patch mesh behind it, and texture it with '''textures/sfx/white'''. When you select this patch in DR, it gives a nice uniform brown color		* '''OR''' then create a simple patch mesh behind it, and texture it with '''textures/sfx/white''' or '''textures/darkmod/sfx/black_matt'''. When you select this patch in DR, it gives a nice uniform brown color, or you can use the black version. Use the version where the background clashes the least with your model's colors.

	The idea here is that which we can later separate the background from the model itself in Gimp easily be just selecting the background color.		The idea here is that which we can later separate the background from the model itself in Gimp easily be just selecting the background color.
Line 249:		Line 249:
	* The [[LOD]] system		* The [[LOD]] system
	* http://en.wikipedia.org/wiki/Texture_atlas		* http://en.wikipedia.org/wiki/Texture_atlas
			* [http://bloodgate.com/mirrors/tdm/pub/lod_construction.zip Helper maps with black room and fixed camera pos]

	{{editing}} {{scripting}}		{{editing}} {{scripting}}

Tels: more spaces and typo fix

2010-09-25T12:24:05Z

more spaces and typo fix

← Older revision		Revision as of 12:24, 25 September 2010
Line 190:		Line 190:
	[[Image:Lilies color correct.png\|420px\|Collor coretion with the Curves tool]]		[[Image:Lilies color correct.png\|420px\|Collor coretion with the Curves tool]]

	~~:'''Result:'''~~

	~~[[Image~~:~~Lily before colorcorrect.png\|360px\|Lily model in game before color correction]]  ~~ [[Image:Lily after colorcorrect.png\|360px\|Lily model in game after color correction]]		'''Result:'''

			[[Image:Lily after colorcorrect.png\|360px\|Lily model in game after color correction]]


	== Optimizing ==		== Optimizing ==
Line 239:		Line 241:
	== Remarks ==		== Remarks ==

	In my first test, I also did a "top-side" view, and ~~places~~ this as patch at the bottom, horizontally. The theory was that there is something the player can see if he looks straight from the top.		In my first test, I also did a "top-side" view, and placed this as patch at the bottom, horizontally. The theory was that there is something the player can see if he looks straight from the top.

	However, in game tests showed that it only looked funky (often z-fighting with the floor), and since the player is far away from the model, anyway, he can't see it from straight top. If you view it high above, it will still look quite good due to the patch being vertical and thus partly visible. So, just having a front and side few is sufficient.		However, in game tests showed that it only looked funky (often z-fighting with the floor), and since the player is far away from the model, anyway, he can't see it from straight top. If you view it high above, it will still look quite good due to the patch being vertical and thus partly visible. So, just having a front and side few is sufficient.

Tels: typo

2010-09-25T12:19:41Z

typo

← Older revision		Revision as of 12:19, 25 September 2010
Line 37:		Line 37:
	The idea here is that which we can later separate the background from the model itself in Gimp easily be just selecting the background color.		The idea here is that which we can later separate the background from the model itself in Gimp easily be just selecting the background color.

	'''~~Tips~~:'''		'''Notes:'''

	* Draw the camera view from DR as big as possible to capture the highest quality possible.		* Draw the camera view from DR as big as possible to capture the highest quality possible.
Line 46:		Line 46:
	Then I used the screen capture program from KDE to create screenshots and open them up in Gimp. One shot from the front, one from the side. Try to align the camera as level and at the center of the model as possible. The exact alignment isn't nec, but it simplifies things somewhat.		Then I used the screen capture program from KDE to create screenshots and open them up in Gimp. One shot from the front, one from the side. Try to align the camera as level and at the center of the model as possible. The exact alignment isn't nec, but it simplifies things somewhat.

	''''Tips:''''		'''Tips:'''

	* The KDE screenshot applet has a handy "Open with ... Gimp" button for this so you can take all screenshots and load them in Gimp without saving them first to disk).		* The KDE screenshot applet has a handy "Open with ... Gimp" button for this so you can take all screenshots and load them in Gimp without saving them first to disk).

Tels: remarks fixed

2010-09-25T12:17:49Z

remarks fixed

← Older revision		Revision as of 12:17, 25 September 2010
Line 13:		Line 13:
	=== Start with a suitable model ===		=== Start with a suitable model ===

	I ~~used~~ the cattails4.low model, as I am using it in a map of my own~~, and it is a very detailed model that is still compact and thus promises a good speedup with a simple process:~~		Select a model that:

			* has sufficiently high number of polys (anything > 100) OR has a many surfaces (> 2)
			* will be used in places that will be both far away and near the player in one mission
			* Is compact and possible symmetric from all sides

			I started with the ''cattails4.low'' model as example, as I am using it in a map of my own.

	{{clear}}		{{clear}}
Line 34:		Line 40:

	* Draw the camera view from DR as big as possible to capture the highest quality possible.		* Draw the camera view from DR as big as possible to capture the highest quality possible.
	* Use the "~~Level camera~~" option from the menu (or use {{key\|END}}) to level the camera after moving it at the center line of the model.		* Use the "View => Camera => Center" option from the DR menu (or use {{key\|END}}) to level the camera after moving it at the center line of the model.
	* If you use a patch, make sure you put the patch completely behind the model.		* If you use a patch, make sure you put the patch completely behind the model.
	* If your model has skins, it is best to switch skins and make one screenshot for each skin without adjusting the camera. This way all the skin screenshots will align perfectly without any extra work, and you only need to position the camera once for each shot.		* If your model has skins, it is best to switch skins and make one screenshot for each skin without adjusting the camera. This way all the skin screenshots will align perfectly without any extra work, and you only need to position the camera once for each shot.

Tels: add camera level trick

2010-09-25T12:14:56Z

add camera level trick

← Older revision		Revision as of 12:14, 25 September 2010
Line 34:		Line 34:

	* Draw the camera view from DR as big as possible to capture the highest quality possible.		* Draw the camera view from DR as big as possible to capture the highest quality possible.
			* Use the "Level camera" option from the menu (or use {{key\|END}}) to level the camera after moving it at the center line of the model.
	* If you use a patch, make sure you put the patch completely behind the model.		* If you use a patch, make sure you put the patch completely behind the model.
	* If your model has skins, it is best to switch skins and make one screenshot for each skin without adjusting the camera. This way all the skin screenshots will align perfectly without any extra work, and you only need to position the camera once for each shot.		* If your model has skins, it is best to switch skins and make one screenshot for each skin without adjusting the camera. This way all the skin screenshots will align perfectly without any extra work, and you only need to position the camera once for each shot.