Incorrect image output
This page lists technical issues in Substance 3D Designer resulting in an image output which is incorrect of unexpected , and offers troubleshooting steps for each.
Visible stepping/banding
Check the Output format (i.e. bitdepth) of the node and all nodes upstream and make sure these node use at least 16-bit Integer precision.
The Output format parameter is often set to the Relative to input inheritance method, which can propagate the low precision throughout the graph. Ideally, by going upstream in the graph you will find the root cause of the issue.
You can quickly identify the precision of a node’s output by taking a look at the text information displayed below the node:
- L/C refer to the image being Grayscale (i.e. Luminance) or Color
- 8/16 means integer encoding
- 16F/32F means floating point encoding
For instance:
- L8: grayscale 8-bit integer
- C16: color 16-bit integer
- C32F: color 32-bit floating point (HDR)
Quality loss in published SBSAR
 Recommended steps
Make sure the Output size property of all Bitmap nodes are set to the Absolute inheritance method.
If that is not the case, their referenced Bitmap resource will be saved at the default 256*256 resolution in the published Substance 3D archive, which will impact the quality of one or more outputs.
Image is blurry
When rearranging pixels in an image, e.g. when resizing a shape or changing the resolution of an image, there are two ways to determine how pixels from the source should be mapped to the destination:
- Nearest: The pixel will be mapped to the target as-is at the matching coordinate. If the target is of lower resolution the pixel may be entirely ignored. If the target is of higher resolution; it will be mapped to all pixels covering its span. The output is crisper and will look slightly aliased.
- Bilinear filtering: A filtering process is applied to the source image so its pixels are mapped to the target resolution in a way that smooths out the transitions between pixels. The output is smoother and will look slightly blurred.
The Transformation 2D node provide a Filtering method option to select which of these two mapping methods should be used.
Most nodes - e.g. Blend - default to bilinear filtering when sampling an input texture of different resolution, which may introduce undesired blurring.
Because the Transformation 2D node is atomic - thus very lightweight - it may be used even if no transformations are needed to change a texture resolution using its Output size property before sending the texture to another node, so you can control the impact of this resizing.
In the Pixel processor node’s function graph, the Sample nodes include the same option to control how the sampled texture should be mapped to the node’s resolution.