SEAR DevBlog, weeks of 10/5 and 10/12: Playtest just around the corner
We’ve got a lot going on over here.
The Resonator’s initial states are mostly implemented. The hologram flows from one appearance to the next, matches the color of success/failure presented by the wave sensor and transmission lines, and is starting to look like the object we had always envisioned. It’s really cool to see it come together.
We have two level designers tackling the two levels we need to pull together within the next month or so.
We’ve got our color settings backend working so we can accommodate color blindness and custom color presets from the player - still need the frontend implementation of this, but that will be quick.
The entire level loading process has been completely rewritten to take the heavy loading OUT of the 2-second transition window and instead spaced out over the course of the phase-level. This plays well on my machine but we’re running tests on a couple more before the playtest to get a better feel for any potential hitches. In related news we’ll be adding our ray-tracing settings and toggles prior to playtest as well.
Related to the level loading redesign has been a redesign of the RaceManager system that ranked the player - previously a placeholder single-player-only design, now expanded out to a full competitor subsystem which tracks the player as one of 12 and assigns an identity and score to each NPC racer which is tracked across the race, resulting in a far more accurate (and less punitive) final rank score. This system also lays the groundwork for true online multiplayer.
The problem child remains the NPC AI development. This is one place where SEAR’s open-ended, “anything goes” level design philosophy is really biting us. It’s cool for the player (and for the designers) when one level is a strict racetrack, the next is a terrain-informed offroad landscape track, and the next is a wide open landscape field aimed in a general direction. It keeps the experience fresh and varies the subgoals and playstyle by phase. But AI to travel along a single racetrack can be much simpler than AI that has to read an entire landscape field with slowing terrain, hard blocks, ramps, tunnels, pickups, shortcuts, and all the rest of it. Our original AI was a simple spline-based tracker with essentially no decision-making behind it at all - enough to show the game idea, but not much more. For that reason I’ve been very reticent about using splines at all in the AI development… but I think I might need to find a middle ground here soon.
Final announcement on the Steam Playtest run dates will be out this week. If you’re on our mailing list, watch for it!
Comments