Displacement autobump

A very powerful feature in Maverick is displacement autobump. Autobump was introduced when Indie was released some months ago, but was improved in our most recent Studio/Indie build; now it is interactively configurable.

In Maverick, displacement works as follows:

• Geometry is diced (i.e., subdivided) by the max edge length thresholds you can configure globally and/or at the object.
• Then the displacement heightmap is used to push vertices along their normals.
• All detail not captured in the previous 2 steps, is automatically added as a normal map, which strength is now interactively editable.

This approach to displacement has great advantages:

• No matter how crude your base mesh or how conservative your subdivision thresholds are, all fine detail is always present.
• This allows for great detail while keeping the memory footprint and warm up time low. i.e., you can afford large subdivision thresholds.
• Autobump is interactively configurable. So you can crank up details or mute them a bit to your liking, in real-time.

In other words, you can achieve incredible detail: automatically, fast, and without bursting the GPU RAM.

This powertool works in combination with all the other advanced properties of our subdivision and displacement system, such as adaptiveness, culling, …

This Youtube video that we just uploaded makes use of autobump in a practical example: a material downloaded from Substance Source:

 

Summary

Thanks to autobump, a practical workflow for displacement would be:

1. Set a subdivision threshold that captures (only) the coarse details of your material. Do not overshoot! Just make sure that features such as ridges and thick wrinkles are captured.
2. Then play with autobump to restore all the detail that is missing. Note that the default autobump value is often ok.
3. As a bonus, make sure that the autobump epsilon value is set to match the resolution of your displacement map (e.g., 4096px).

These slides below illustrate how, thanks to autobump, a crude subdivision threshold produces the exact same result as a very fine subdivision threshold:

 

 

Thanks for watching!