Textures

A summarised guide on optimal settings and best practices used in the creation of textures.

Naming Convention

Using an appropriate naming convention is useful in grouping/parsing assets within a project.
Prefix texture assets with a T_ (i.e T_Car_Texture_00).
Append texture assets with the suffix that corresponds to the category of the texture (i.e T_Car_Texture_00_N corresponds to the Texture Normal).

Texture axes should always be a power of 2 (i.e 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192). This is required due to the manner in which the textures are loaded into memory as well as to adhere to Unreal’s optimisation techniques that include mip mapping and texture streaming.
Textures are not required to be square, but rather to adhere to the power rule (i.e 512 x 64, 8192 x 1024, etc).
Following this process allows for LODs to become functional on the texture as a correctly created texture will feature a number of MIP maps which facilitates the scaling requirement for LODs.

There are two ways of storing alpha information in a texture - Embedded and Separate.
Making use of the Separate alpha is ideal as it will be compressed when imported into Unreal versus the Embedded alpha which will be imported in an uncompressed manner.
Making use of the Separate alpha also allows for size control which is independent to the Base color size.
Should you wish to retain the high resolution of the alpha information then the Embedded alpha is more useful. Note that this will have double the cost of the Separate alpha.

This is a method used to reduce the number of textures and amount of memory used throughout the project. It is generally used within VFX and character model applications. The method involves packing multiple masks together into the R, G, B, and A channels of a single image which is then exported.
When using this method, only black and white information will be accessible from the texture, so some form of color parameter multiplication will be required to obtain the respective color information.
Uncheck the sRGB parameter under the Texture section of the Details pane. This then supports the channel extraction.
When using the texture on a material, set the Sampler Type to Masks under the Material Expression Texture Base section of the Details pane. This is required as disabling sRGB on the texture will disable gamma correction on the texture. Mask textures do not require gamma correction as they simply inform the renderer of whether a pixel should be visible or not.

This is a key feature in facilitating good performance within a project. It can be thought of as LODs for textures.
MIP map generation occurs when a texture is imported into Unreal. In most cases, no interaction is required unless for a very specific use case. Such a case is where shimmering or aliasing is noticed on an object. This can be adjusted by selecting one of the Sharpen configurations from the Mip Gen Settings option under the Level Of Detail section in the Details pane of the texture.
This is used to apply a resolution of the texture which is better suited to the render size of the object, rather than incurred the memory cost of rendering the full resolution texture where the full detail cannot be told apart from the scaled down variant.

There are two ways of importing textures into Unreal, namely the Drag and Drop and Import methods.
The Drag and Drop method simply requires a texture to be dragged into the Content Browser from a Windows Explorer Window. The import is automatically handled.
The Import method requires the Import button in the Content Browser to be selected. This opens a Windows Explorer Window to navigate to and select the required texture.
Unreal will assess the file name of the texture to decide the appropriate type of texture being imported (refer to this Naming Convention Guide).

These groups serve to manage how textures are used and displayed within a project. Each group determines the size at which the texture will be drawn within the project, how the texture is minified or magnified and what type of filtering is applied to the texture.
Thus it is imperative that correct group assignment of each texture is performed.
Group assignment is done by changing the Texture Group property of the Level Of Detail section in the Details pane of the texture.

Such an example of a defined texture group configuration:

  TEXTUREGROUP_World=(MinLODSize=1,MaxLODSize=8192,LODBias=1,MinMagFilter=aniso,MipFilter=point)

Texture compression can be adjusted at any time in the editor.
Open the target texture from the Content Browser.
Select the desired Compression Settings type under the Compression section in the Details pane of the texture.
Refer to this Texture Compression Document for information on each type of texture compression.
Certain types of compression will require the Compress Without Alpha property to be set. Consult the above document to determine the required configuration.