Texture file

From TrainzOnline
(Difference between revisions)
Jump to: navigation, search
(File format: DXT Compression link added)
 
(27 intermediate revisions by 8 users not shown)
Line 1: Line 1:
[[Texture file]]s are ASCII text format files describing a single texture. Texture files have a ".texture.txt" name extension, however when a texture file is specified within a [[Config.txt file]] or similar, the ".txt" must be omitted, leaving the ".texture" extension.
+
{{ORP-top}}
 +
Trainz '''Texture files''' are ASCII text format specification files describing a single texture. Texture files have a ".texture.txt" name extension, however when a texture file is specified within a [[config.txt file]] or similar, the ".txt" must be omitted (as you are specifying a resource name, not a file path), leaving only the ".texture" extension.
  
'''TODO: Merge [[.texture.txt Files]] page into this page.'''
+
Texture files (*.texture.txt) should be distinguished from [[image file]]s (*.tga, etc.) and [[material]]s. A material may utilize zero or more texture files, and a texture file may utilize one or more image files. While we often refer to the image data as "a texture", this usage isn't strictly correct as any given texture also has a number of metadata attributes which are not present in the source image. Additionally, a texture may be compressed (at installation time) which may actually lose some of the data originally present in the source image.
  
 +
When importing a [[trainzmesh file]] from an [[FBX file]], any image files referenced in the FBX which do not already have corresponding texture files will have generic texture files created automatically. These files may be suitable for basic purposes, but may need to be customized for more advanced functionality. Once the file has been created (and optionally customised) the importer will not recreate/overwrite it on subsequent runs.
  
==File format==
+
This page describes [["Trainz-build" number|content format v5.6]].
 +
 
 +
 
 +
=File format=
  
 
The user-editable *.texture.txt files are ASCII text formatted key-value pairs with the following options. Please note that this format is not the same (despite some similarities) as other text-based formats used by Trainz. Please be aware of the [[filename character restrictions]] when naming texture and image files.
 
The user-editable *.texture.txt files are ASCII text formatted key-value pairs with the following options. Please note that this format is not the same (despite some similarities) as other text-based formats used by Trainz. Please be aware of the [[filename character restrictions]] when naming texture and image files.
  
 +
;Syntax
 +
Each key is specified on a new line. There is no white space anywhere in the line. A empty value is sometimes valid. The syntax is:
 +
 +
<key>=<value>
 +
 +
;Example:
 +
Primary=WayCoolTexture.tga
 +
Alpha=WayCoolTexture.tga
 +
Tile=st
 +
UsageHint=albedo
 +
AlphaHint=semitransparent
 +
MipAlgorithm=heavy-alpha
 +
''(Example only, not recommended settings)''
 +
 +
 +
==Keys and Values==
  
 
  Primary=diffusemap.bmp
 
  Primary=diffusemap.bmp
Line 17: Line 38:
 
  Alpha=diffusemap.bmp
 
  Alpha=diffusemap.bmp
  
This tag combination specifies the filenames of the image file representing the color (R, G, B) channels of the texture and the image file representing the alpha (A) channel of the texture. The alpha channel image should be grayscale. Neither file should contain an alpha channel.
+
This tag combination specifies the filenames of the image file representing the color (R, G, B) channels of the texture (Primary) and the image file representing the transparency of the texture (Alpha). The Alpha image should be grayscale. Neither file should contain an alpha (A) channel (ie, both images should be 24-bit). The two image files must be of identical size.
  
  
Line 23: Line 44:
 
  Alpha=diffusemap.tga
 
  Alpha=diffusemap.tga
  
This tag combination specifies the filename of a single image file representing the color and alpha (R, G, B, A) channels of the texture. In this case both tags must reference the same filename. The image file must contain an alpha channel.
+
This tag combination specifies the filename of a single image file representing the color and transparency (R, G, B, A) channels of the texture. In this case both tags must reference the same filename. The image file must contain an alpha channel.
  
  
  Compression=dxt1
+
  Compression=dxt3
  
This tag forces the usage of a particular texture compression scheme. If omitted, Trainz will choose an appropriate texture compression scheme based on the contents of the supplied image files. If you don't have a specific effect in mind, don't include this tag. Valid options are 'none', 'dxt1', 'dxt3', 'dxt5'. Refer [[DXT Compression]].  
+
This tag hints the usage of a particular texture compression scheme. If you don't have a specific effect in mind, don't include this tag. Valid options are currently 'none', 'dxt1', 'dxt3', 'dxt5'. Refer [[DXT Compression]].
 +
 
 +
The current selection logic for texture compression is as follows, however it should be noted that this may change between Trainz versions. The first matching rule is used.
 +
 
 +
* Special texture formats (such as HDR) which don't support compression are stored uncompressed.
 +
* "Compression=none" forces no compression. This gives the best visual result for a given input image, but typically yields worse visual results per output texture filesize. It is often better to activate compression and double the image dimensions, rather than disabling compression.
 +
* Images with an "opaque" alpha hint use DXT1 compression.
 +
* "Compression=dxt3" will use DXT3 compression.
 +
* DXT5 compression is used.
  
  
Line 34: Line 63:
  
 
The 'NormalMapHint' tag allows the texture to be specified as a [[normal map]]. By default, Trainz will assume that your texture is a [[diffuse map]] and will apply some processing options which are not appropriate for normal maps. You must specify this option for normal maps. Valid options are 'none' and 'normalmap'.
 
The 'NormalMapHint' tag allows the texture to be specified as a [[normal map]]. By default, Trainz will assume that your texture is a [[diffuse map]] and will apply some processing options which are not appropriate for normal maps. You must specify this option for normal maps. Valid options are 'none' and 'normalmap'.
 +
  
 
  ModifyMap=flipgreen
 
  ModifyMap=flipgreen
  
The 'ModifyMap' option allows a texture's green color channel to be flipped.  Some programs such as 3ds Max create [[normal map]]s with the Y axis facing the opposite way to what is expected by the Trainz shaders.  This option can be used to make the bump effect look correct in Trainz. The option is applied by the [[Content Manager]] upon Commit. Valid options are 'none' and 'flipgreen'.
+
The 'ModifyMap' option allows a texture's green color channel to be flipped.  Some programs such as [[Autodesk 3ds Max]] create [[normal map]]s with the Y axis facing the opposite way to what is expected by the Trainz shaders.  This option can be used to make the bump effect look correct in Trainz. Valid options are 'none' and 'flipgreen'.
  
  
 
  AlphaHint=opaque
 
  AlphaHint=opaque
  
The 'AlphaHint' tag allows you to force a particular alpha mode on the texture. Valid options are 'opaque' (meaning that the alpha channel is unused and effectively pure white), 'semitransparent' (meaning that the alpha channel is a grayscale blend) and 'masked' (meaning that the alpha channel is a pure black&white bitmap.) If omitted, Trainz will supply an appropriate value for this tag based on the Alpha channel. Specifying this option prevents slight discoloration in a mask texture from triggering the blended alpha path.
+
The 'AlphaHint' tag allows you to force a particular alpha mode on the texture. Valid options are 'opaque' (meaning that the alpha channel is overwritten with pure white), 'semitransparent' (meaning that the alpha channel is a grayscale blend) and 'masked' (meaning that the alpha channel is clamped to a pure black&white bitmap based on a 50% threshold.) If omitted, Trainz will supply an appropriate value for this tag based on the actual contents of the Alpha channel.
  
  
MagFilter=linear
 
  
Controls the filtering style for texel magnification. This is used when a texel is displayed at larger than real size (eg. a single texel is represented with more than a single pixel.) Valid settings for the magnification filter are 'nearest', 'linear', and 'default'. Trainz currently defaults to linear interpolation. Selecting 'nearest' will use a nearest-neighbor filtering mode which results in a "pixelated" output.
+
  Tile=st
   
+
  
MinFilter=linear
+
Tiling refers to the ability for texture coordinates outside the range of (0.0 .. 1.0) to be treated as valid coordinates on an infinitely tiled texture. With tiling active, coordinates which pass the right/bottom side of the texture effectively wrap back onto the left/top side, and so on. Valid options include 's' (tile horizontally), 't' (tile vertically), 'st' (tile both horizontally and vertically), and 'none' (do not tile the image).
  
Controls the filtering style for texel minification. This is used when a texel is displayed at smaller than real size (eg. a single pixel is represented with more than a single texel.) Valid settings for the minification filter are 'nearest', 'linear', and 'default'. Trainz currently defaults to linear interpolation. Selecting 'nearest' will use a quantized filtering mode which results in a "noisy" output.
 
  
 +
UsageHint=albedo
  
MipFilter=linear
+
Textures are processed slightly differently depending on their intended use in the rendering pipeline. Valid options are 'albedo', 'normal', and 'parameter'. If no usage hint is supplied, the game will attempt to guess the ideal processing steps based on other clues, which may lead to undesirable outcomes if the guesses are wrong.
  
Controls the filtering style for texture mipmap selection. Mipmaps are used during minification to help prevent the moire effect pattern. The cost of using mipmaps is that the resultant image will blur slightly. The benefits of using mipmaps are reduced quantization noise and reduced memory footprint (since the higher detail mip levels can be unloaded when not in use.) Valid settings for the mip filter are 'nearest', 'linear', 'default' and 'none'. Trainz currently defaults to trilinear interpolation. Selecting 'nearest' will cause the hardware to select a single mipmap for each pixel, which results in visible banding at mipmap boundaries. Selecting 'none' will disable mipmaps, resulting in lower performance, higher memory usage, and "noisy" output.
 
  
 +
MipAlgorithm=weight-by-alpha
  
Tile=st
+
Specifies the [[mipmap]] generation algorithm to use for the texture. Valid options are 'weight-by-alpha', 'no-weight-by-alpha', and 'heavy-alpha'. If not specified, the 'UsageHint' for the texture is consulted to determine the default behaviour; for albedo textures the 'weight-by-alpha' mechanism is used, while other textures default to 'no-weight-by-alpha'. The 'weight-by-alpha' mechanism treats the alpha channel as an opacity and prioritizes opaque texels over transparent ones. The "no-weight-by-alpha" hint can be used to ensure that the game will give equal weighting to texels regardless of their alpha channel, which is desirable if the alpha channel does not represent opacity. The 'heavy-alpha' mechanism effectively increases the alpha channel on lower mip levels, which is useful on semitransparent foliage (eg. most trees) to avoid the tree appearing overly transparent in the distance.
  
Tiling refers to the ability for texture coordinates outside the range of (0.0 .. 1.0) to be treated as valid coordinates on an infinitely tiled texture. With tiling active, coordinates which pass the right/bottom side of the texture effectively wrap back onto the left/top side, and so on. Valid options include 's' (tile horizontally), 't' (tile vertically), 'st' (tile both horizontally and vertically), and 'none' (do not tile the image).
+
The value of this key is only relevant where [[Content Manager]] is responsible for producing mipmap images. It is not relevant on any mipmap images provided in the source images.
  
==Image formats==
 
The following image formats are acceptable as source data for the Texture file:
 
  
* Targa (.tga) files. Targa is a lossless image format with optional compression and optional alpha channel.
+
nonpoweroftwo=1
* JPEG (.jpg) files. JPEG is a lossy image compression format with no alpha channel support.
+
* Windows Bitmap (.bmp) files. BMP is a lossless image format with no compression and optional alpha channel.
+
  
These images in association with the appropriate *.texture.txt file are used by [[Content Manager]] to create the N3V binary textures that are used in the simulator. The N3V binary textures are use [[DXT Compression]] and contain image LOD information.
+
By default, all textures must have both their width and height as a power of two (eg. 8, 16, 32, 64, 128, 256, 512, 1024, 2048), though they need not be square. Setting the value '1' for the 'nonpoweroftwo' key allows this restriction to be relaxed. This is currently considered experimental and may have unexpected performance, quality, or functionality outcomes. It is not recommended that end-users utilize this format at the current time.
  
All built-in content has only N3V binary textures that cannot be opened with normal imaging software.  Appropriate tools to view these can be found on the [[Tools]] page.
+
Non-power-of-two (NPOT) textures may not support mipmapping and thus may be noticeably affected by the moire effect in addition to having higher overall GPU bandwidth requirements.
  
==Notes on Image Editors==
+
=Texture Formats=
 +
The following texture formats are supported internally by Trainz on desktop platforms.
 +
* R8 G8 B8 A8
 +
* DXT1
 +
* DXT3
 +
* DXT5
 +
* R32F G32F B32F A32F
 +
 
 +
The selection of internal format depends on the source image format and the value of the 'Compression' key.
 +
 
 +
=Texture Size=
 +
Textures must be no more than 4096 pixels in either dimension, although it is strongly recommended that textures are kept below this maximum to avoid performance problems. Small objects should typically use approximately 512x512. Large, highly detailed objects such as train vehicles or ground textures will typically use 2048x2048.
 +
Uniform-color textures (ie. textures in which all [[texel]]s have the same color) should be 8x8, 2x2, or 1x1.
 +
Textures must be a power-of-two size in each dimension (unless the 'nonpoweroftwo' key is activated), but need not be square. Some materials may give unexpected results for non-square textures; consult the material-specific documentation for details.
 +
 
 +
=Source Image Formats=
 +
The following image formats are acceptable as source data for the texture file. Note that these formats are not used at runtime, and that format-specific benefits such as compression will apply to the DLS download size of the content but not to the runtime memory footprint or GPU bandwidth usage.
 +
 
 +
* Targa (.tga) files. Targa is a lossless image format with optional compression and optional alpha channel. This is a sensible choice where compatibility with older versions of Trainz is desired.
 +
* JPEG (.jpg) files. JPEG is a lossy image compression format with no alpha channel support. This format should generally be avoided since it will typically lead to the image being compressed in a lossy format (JPG), decompressed, recompressed again into another lossy format (DXT1), losing detail at each stage. Worse, if JPEG files are used during the creation of the image, each edit-save-test cycle may result in a loss of image quality.
 +
* Windows Bitmap (.bmp) files. BMP is a lossless image format with no compression and optional alpha channel. This format should generally be avoided since it does not offer compression and has no redeeming benefits.
 +
* HDR (.hdr) files. HDR is a lossless image format with 32 bits of precision per channel. The R32F G32F B32F A32F internal texture format is used, and texture compression is not supported. This is a very expensive format (in terms of memory and performance) and should only be used where absolutely essential. Textures created in this format should be kept as small as possible (eg. 256-512 pixels wide and high) but at least 8x8 pixels wide and high.
 +
* PNG (.png) files. PNG is a lossless compressed image format with alpha channel support. This is the best choice where compatibility with older versions of Trainz is not required.
 +
* DDS (.dds) files. This has very limited support intended for specific internal purposes; it is not recommended that end-users utilize this format at the current time.
 +
 
 +
All images which comprise a given texture must have identical dimensions (width, height).
 +
 
 +
Source images should use the SRGB color space except where material-specific documentation indicates otherwise.
 +
 
 +
=Notes on Image Editors=
 
* The Targa Exporter plugin shipped with some older versions of Adobe Photoshop is broken and silently corrupts or discards alpha channels; a free update is available which resolves this issue.
 
* The Targa Exporter plugin shipped with some older versions of Adobe Photoshop is broken and silently corrupts or discards alpha channels; a free update is available which resolves this issue.
 
* Some image editors do not support alpha channels on BMP files.
 
* Some image editors do not support alpha channels on BMP files.
 +
* By default, Photoshop does not allow for explicit editing of the alpha channel in PNG images. There is a free plugin called [https://www.fnordware.com/superpng/ SuperPNG] which integrates better PNG support to enable saving and editing of PNG alpha masks.
 +
* [http://www.amnoid.de/ddsview/ ddsview ] is an open source image editor that can properly open all of the texture formats used by Trainz. This is an easy tool to quickly convert between PNG parameters and TGA parameters.
  
 
+
=*.txt File Extension=
==.txt file extension==
+
Texture files have a ".texture.txt" name extension, however when a texture file is specified within a [[config.txt file]] or similar, the ".txt" must be omitted, leaving the ".texture" extension.
Texture files have a ".texture.txt" name extension, however when a texture file is specified within a [[Config.txt file]] or similar, the ".txt" must be omitted, leaving the ".texture" extension.
+
 
For example, a texture file named ''grass.texture.txt'' is specified in a [[KIND Groundtexture]] config file using the syntax ''texture "grass.texture"''.
 
For example, a texture file named ''grass.texture.txt'' is specified in a [[KIND Groundtexture]] config file using the syntax ''texture "grass.texture"''.
  
==Use of image files in place of texture files==
+
=Use of Image Files in Place of Texture Files=
Trainz has historically supported using any supported image file in place of a [[Texture file]]. This technique has obvious limitations and is no longer recommended.
+
Trainz previously supported using ''any supported image file'' type in place of a [[Texture file]]. [[TANE]] does not support this technique; a texture.txt file must be used.
  
 
[[Category:File formats]]
 
[[Category:File formats]]
 +
[[Category:Modeling]]
 +
[[Category:Content creation]]
 +
{{ORP-bot}}

Latest revision as of 14:37, 1 October 2024

Trainz Texture files are ASCII text format specification files describing a single texture. Texture files have a ".texture.txt" name extension, however when a texture file is specified within a config.txt file or similar, the ".txt" must be omitted (as you are specifying a resource name, not a file path), leaving only the ".texture" extension.

Texture files (*.texture.txt) should be distinguished from image files (*.tga, etc.) and materials. A material may utilize zero or more texture files, and a texture file may utilize one or more image files. While we often refer to the image data as "a texture", this usage isn't strictly correct as any given texture also has a number of metadata attributes which are not present in the source image. Additionally, a texture may be compressed (at installation time) which may actually lose some of the data originally present in the source image.

When importing a trainzmesh file from an FBX file, any image files referenced in the FBX which do not already have corresponding texture files will have generic texture files created automatically. These files may be suitable for basic purposes, but may need to be customized for more advanced functionality. Once the file has been created (and optionally customised) the importer will not recreate/overwrite it on subsequent runs.

This page describes content format v5.6.


Contents

[edit] File format

The user-editable *.texture.txt files are ASCII text formatted key-value pairs with the following options. Please note that this format is not the same (despite some similarities) as other text-based formats used by Trainz. Please be aware of the filename character restrictions when naming texture and image files.

Syntax

Each key is specified on a new line. There is no white space anywhere in the line. A empty value is sometimes valid. The syntax is:

<key>=<value>
Example
Primary=WayCoolTexture.tga
Alpha=WayCoolTexture.tga
Tile=st
UsageHint=albedo
AlphaHint=semitransparent
MipAlgorithm=heavy-alpha

(Example only, not recommended settings)


[edit] Keys and Values

Primary=diffusemap.bmp

This tag specifies the filename of the image file representing the color (R, G, B) channels of the texture. Any alpha channel in the image file is ignored. The resultant texture will have a pure white (100%) alpha channel.


Primary=diffusemap.jpg
Alpha=diffusemap.bmp

This tag combination specifies the filenames of the image file representing the color (R, G, B) channels of the texture (Primary) and the image file representing the transparency of the texture (Alpha). The Alpha image should be grayscale. Neither file should contain an alpha (A) channel (ie, both images should be 24-bit). The two image files must be of identical size.


Primary=diffusemap.tga
Alpha=diffusemap.tga

This tag combination specifies the filename of a single image file representing the color and transparency (R, G, B, A) channels of the texture. In this case both tags must reference the same filename. The image file must contain an alpha channel.


Compression=dxt3

This tag hints the usage of a particular texture compression scheme. If you don't have a specific effect in mind, don't include this tag. Valid options are currently 'none', 'dxt1', 'dxt3', 'dxt5'. Refer DXT Compression.

The current selection logic for texture compression is as follows, however it should be noted that this may change between Trainz versions. The first matching rule is used.

  • Special texture formats (such as HDR) which don't support compression are stored uncompressed.
  • "Compression=none" forces no compression. This gives the best visual result for a given input image, but typically yields worse visual results per output texture filesize. It is often better to activate compression and double the image dimensions, rather than disabling compression.
  • Images with an "opaque" alpha hint use DXT1 compression.
  • "Compression=dxt3" will use DXT3 compression.
  • DXT5 compression is used.


NormalMapHint=normalmap

The 'NormalMapHint' tag allows the texture to be specified as a normal map. By default, Trainz will assume that your texture is a diffuse map and will apply some processing options which are not appropriate for normal maps. You must specify this option for normal maps. Valid options are 'none' and 'normalmap'.


ModifyMap=flipgreen

The 'ModifyMap' option allows a texture's green color channel to be flipped. Some programs such as Autodesk 3ds Max create normal maps with the Y axis facing the opposite way to what is expected by the Trainz shaders. This option can be used to make the bump effect look correct in Trainz. Valid options are 'none' and 'flipgreen'.


AlphaHint=opaque

The 'AlphaHint' tag allows you to force a particular alpha mode on the texture. Valid options are 'opaque' (meaning that the alpha channel is overwritten with pure white), 'semitransparent' (meaning that the alpha channel is a grayscale blend) and 'masked' (meaning that the alpha channel is clamped to a pure black&white bitmap based on a 50% threshold.) If omitted, Trainz will supply an appropriate value for this tag based on the actual contents of the Alpha channel.


Tile=st

Tiling refers to the ability for texture coordinates outside the range of (0.0 .. 1.0) to be treated as valid coordinates on an infinitely tiled texture. With tiling active, coordinates which pass the right/bottom side of the texture effectively wrap back onto the left/top side, and so on. Valid options include 's' (tile horizontally), 't' (tile vertically), 'st' (tile both horizontally and vertically), and 'none' (do not tile the image).


UsageHint=albedo

Textures are processed slightly differently depending on their intended use in the rendering pipeline. Valid options are 'albedo', 'normal', and 'parameter'. If no usage hint is supplied, the game will attempt to guess the ideal processing steps based on other clues, which may lead to undesirable outcomes if the guesses are wrong.


MipAlgorithm=weight-by-alpha

Specifies the mipmap generation algorithm to use for the texture. Valid options are 'weight-by-alpha', 'no-weight-by-alpha', and 'heavy-alpha'. If not specified, the 'UsageHint' for the texture is consulted to determine the default behaviour; for albedo textures the 'weight-by-alpha' mechanism is used, while other textures default to 'no-weight-by-alpha'. The 'weight-by-alpha' mechanism treats the alpha channel as an opacity and prioritizes opaque texels over transparent ones. The "no-weight-by-alpha" hint can be used to ensure that the game will give equal weighting to texels regardless of their alpha channel, which is desirable if the alpha channel does not represent opacity. The 'heavy-alpha' mechanism effectively increases the alpha channel on lower mip levels, which is useful on semitransparent foliage (eg. most trees) to avoid the tree appearing overly transparent in the distance.

The value of this key is only relevant where Content Manager is responsible for producing mipmap images. It is not relevant on any mipmap images provided in the source images.


nonpoweroftwo=1

By default, all textures must have both their width and height as a power of two (eg. 8, 16, 32, 64, 128, 256, 512, 1024, 2048), though they need not be square. Setting the value '1' for the 'nonpoweroftwo' key allows this restriction to be relaxed. This is currently considered experimental and may have unexpected performance, quality, or functionality outcomes. It is not recommended that end-users utilize this format at the current time.

Non-power-of-two (NPOT) textures may not support mipmapping and thus may be noticeably affected by the moire effect in addition to having higher overall GPU bandwidth requirements.

[edit] Texture Formats

The following texture formats are supported internally by Trainz on desktop platforms.

  • R8 G8 B8 A8
  • DXT1
  • DXT3
  • DXT5
  • R32F G32F B32F A32F

The selection of internal format depends on the source image format and the value of the 'Compression' key.

[edit] Texture Size

Textures must be no more than 4096 pixels in either dimension, although it is strongly recommended that textures are kept below this maximum to avoid performance problems. Small objects should typically use approximately 512x512. Large, highly detailed objects such as train vehicles or ground textures will typically use 2048x2048. Uniform-color textures (ie. textures in which all texels have the same color) should be 8x8, 2x2, or 1x1. Textures must be a power-of-two size in each dimension (unless the 'nonpoweroftwo' key is activated), but need not be square. Some materials may give unexpected results for non-square textures; consult the material-specific documentation for details.

[edit] Source Image Formats

The following image formats are acceptable as source data for the texture file. Note that these formats are not used at runtime, and that format-specific benefits such as compression will apply to the DLS download size of the content but not to the runtime memory footprint or GPU bandwidth usage.

  • Targa (.tga) files. Targa is a lossless image format with optional compression and optional alpha channel. This is a sensible choice where compatibility with older versions of Trainz is desired.
  • JPEG (.jpg) files. JPEG is a lossy image compression format with no alpha channel support. This format should generally be avoided since it will typically lead to the image being compressed in a lossy format (JPG), decompressed, recompressed again into another lossy format (DXT1), losing detail at each stage. Worse, if JPEG files are used during the creation of the image, each edit-save-test cycle may result in a loss of image quality.
  • Windows Bitmap (.bmp) files. BMP is a lossless image format with no compression and optional alpha channel. This format should generally be avoided since it does not offer compression and has no redeeming benefits.
  • HDR (.hdr) files. HDR is a lossless image format with 32 bits of precision per channel. The R32F G32F B32F A32F internal texture format is used, and texture compression is not supported. This is a very expensive format (in terms of memory and performance) and should only be used where absolutely essential. Textures created in this format should be kept as small as possible (eg. 256-512 pixels wide and high) but at least 8x8 pixels wide and high.
  • PNG (.png) files. PNG is a lossless compressed image format with alpha channel support. This is the best choice where compatibility with older versions of Trainz is not required.
  • DDS (.dds) files. This has very limited support intended for specific internal purposes; it is not recommended that end-users utilize this format at the current time.

All images which comprise a given texture must have identical dimensions (width, height).

Source images should use the SRGB color space except where material-specific documentation indicates otherwise.

[edit] Notes on Image Editors

  • The Targa Exporter plugin shipped with some older versions of Adobe Photoshop is broken and silently corrupts or discards alpha channels; a free update is available which resolves this issue.
  • Some image editors do not support alpha channels on BMP files.
  • By default, Photoshop does not allow for explicit editing of the alpha channel in PNG images. There is a free plugin called SuperPNG which integrates better PNG support to enable saving and editing of PNG alpha masks.
  • ddsview is an open source image editor that can properly open all of the texture formats used by Trainz. This is an easy tool to quickly convert between PNG parameters and TGA parameters.

[edit] *.txt File Extension

Texture files have a ".texture.txt" name extension, however when a texture file is specified within a config.txt file or similar, the ".txt" must be omitted, leaving the ".texture" extension. For example, a texture file named grass.texture.txt is specified in a KIND Groundtexture config file using the syntax texture "grass.texture".

[edit] Use of Image Files in Place of Texture Files

Trainz previously supported using any supported image file type in place of a Texture file. TANE does not support this technique; a texture.txt file must be used.

Personal tools