These images and videos are from an unreleased project that I had the pleasure of being a Senior Technical Artist for. They represent efforts to stand up process and quality bar for background content in the first biome of our game. This project had a 2.5D orthographic camera with exterior and interior settings that required multiple foreground and background layers moving in unison to create a the feeling of distance via parallaxing motions. Our Art Director Jeramy Cooke had set up a system of actors that we could parent objects to creating explicit layers in the scene that would move as a unit. My job was to generate an aspirational content target with an eye towards process and pipeline.

This included workflows for generating clouds, landscapes and building assets as well as assembly of them in the scene. Because out worlds were procedurally assembled from a set of "room" like building blocks it was necessary to think about how to assemble backgrounds that could function regardless of the size and ratio of the level being generated. We considered procedural assembly for the backgrounds early on, but the desired quality and composition bar was high enough that we would have ended up building new tooling per background problem set along with lengthy troubleshooting for edge cases. Instead we landed on generating a handful of different layers that could be mixed and matched for different levels based on pre-defined data assets. We also looked at simple blueprint tools for speeding up more tedious elements like placing small buildings and trees that would add fidelity, but didn't require hand placement. This allowed us to focus on the larger gestures and then fill in the gaps quickly. We also built the background scenes to be a couple times larger than the standard room sizes so that we could offset them in X/Y space at runtime to get variation and avoid obvious repetition from zone to zone.

One of the challenges with this approach was CPU overhead from attaching many actors to a parent parallaxing blueprint and then offsetting them at runtime on tick. This many static mesh component moving in concert with the player camera caused some big spikes initially. The immediate answer was to reduce overall actors counts which is where the Actor to ISM scripts represented in a separate post came in. This reduced the largest spikes and made this approach more palatable. Longer term if the project had continued I had planned to explore a second approach involving merging assembled scenes into even few static mesh actors via geometry scripts to further reduce draw counts and number of mesh components.

Creative Direction - Duncan Drummond
Art Direction - Jeramy Cooke
Concept Art - Jei Shi