While Dwindling Heights Keep is designed as an open-world environment, the game starts in its jail, through which Tilo travels in a linear sequence. This let us control the introduction of core game mechanics, which the player learns through text prompts and gameplay set-ups.
We first created a brief for the level, describing what we wanted to achieve in the level, including: -A detailed walkthrough of gameplay -A flow diagram of connections between required spaces -Sketches of gameplay/tutorial set-ups.
Reference material was gathered from nearby sites, to help inform both the visual style and the spatial design of the level.
In a testbed we'd already set rules for Tilo's movement. Eg: -How high & far Tilo jumps -How steep a slope can be before Tilo slides down -Height of stairs etc. These rules have tolerances (eg. jump height is 1.5m +/- 0.2m). This lets the final environment seem more natural.
With those rules in mind Seith started blocking out the keep using primitives in Unity, running Tilo through it and adjusting the space until it meets needs of the brief. At the same time he built vignettes of corridors and rooms, to define and test the art style.
The primitive layout was then exported to Autodesk Maya, and replaced with detailed and textured geometry. Because we use tessellation, the base geometry doesn't need to be to highly detailed, making it easy to make changes.
Visibility zones were defined, so only Tilo's immediate surroundings are rendered, improving performance. Seith also wrote a custom Maya>Unity exporter, that allows him to quickly edit and replace assets while the game is running.
Unity prefabs are used to create common and reusable interactive assets like doorways, chests, cupboards, windows etc. This helps Seith quickly iterate, and establishes visual rules for interaction (the player will learn to recognize items that can be interacted with).
Fine details (lighting, set dressing etc.) are introduced as late as possible in the process, in order that Seith can quickly iterate without having to move hundreds of tiny items).
This project is unusual in that we don't create concept art. Since all art (including level art) is created by one person we've been able to skip that stage. Seith is effectively doing concept design in 3D while laying out the level!