To get the best results from Knald we suggest that you follow a small number of recommendations to ensure that the quality of the processed maps is as high as possible.
The bit depth of the images that you use in Knald have a large impact on the quality of the processed maps. Unfortunately, 8bit images have extremely low precision due to the relatively small number of values that can be represented per channel.
As a result of this limited precision, it can be difficult to reproduce variations of angle and slope with acceptable accuracy. This limitation of 8bit height maps may become apparent in the form of banding. These artifacts may not be immediately noticeable on Normal maps, however they are still present and may reduce the accuracy and fidelity of Knald’s output.
The maximum amount of unique values per channel for each bit depth is listed below:
We recommend that you use the highest bit depth possible in whatever source images are to processed by Knald. If your current pipeline is restricted to 8bit, you can export the final map as 8bit from within Knald once the processing has been completed. You will always get more accurate results by authoring source images in higher bit depths and only converting to your target format (eg: 8bit) during Knald’s export.
For production-quality results we recommend using at least 16bit source images whenever possible.
While Knald can be used on any OpenCL compliant graphics card, you will get faster (though identical) results if you use a more powerful card.
Using an Alpha channel on imported Normal/Derivative may substantially increase the quality of your results in cases where the image UVs are anything other than simple planar unwrap to a quad.
All areas within the black area of an Alpha channel will be pinned and interpreted as flat. (128, 128, 255) Knald will assume these areas have no variance in normal information and will be processed as a flat plane. Typically, using the Alpha channel that is generated automatically when baking will be sufficient. However, useful and interesting results may be achieved using gradients and other variants aside from hard opacity.
Using an incorrect background color on imported Normal/Derivative maps will always significantly decrease the quality of your results in cases where the source image’s Alpha channel is either unused or not present.
Never use White or Black as your background color for these maps!
The correct colurs that should be used are as follows: