Slight delay + Previews + Devlog

Hey everyone, I won’t be able meet the end of June release I’d promised. But I’m aiming for 10th July. I’ll share more details below but wanted to share these awesome previews first:

All in-game footage. Few older captures at the bottom of page too. Now for the devlog.

Lights & Shadows

Spent the last whole week working on this environment, its lighting, shadows, optimisation & more. I initially was thinking of skipping this but I feel like atmosphere matters here. It’s turned out even better than I’d hoped thanks to recent perf improvements allowing for 2 light sources each with their own shadows (lamp and moon). Previously, you might’ve have noticed that the game had just 1 shadow and it wasn’t “accurate” in a physical sense.

Similar to the reflections improvement I shared in my last post, I’ve implemented the same algorithm for a more efficient point light shadowing system that lets me cast shadows for in-room lights from lamps more cheaply. Previously, this was too expensive as point lights are way more perf intensive than directional lights like moon or sun light (6 times to be exact). Having both the moon and lamp causing shadows has allowed for way more dynamic lighting for this room.

Not just that, I’ve overhauled both my environment/stage optimisation process. The models are not only more memory efficient, now, I’ve implemented draw call batching which is another key component in keeping CPU time fast.

And I’ve also nearly perfected my baked lightmap generation for the stages themselves too as part of this work. The lightmaps are much higher quality now with no visible flaws (like you might’ve noticed especially in the last chapter). And more efficient as well.

Most of this is just last week or so. I’ve spent most of the working on…

Player animations!

This is the most complex feature that I’ve spent most of this month on, and been slowly working towards from April. And it works flawlessly from VR to desktop to mobile.

Until now, you might’ve noticed our player was always in a fixed pose with not only limited freedom but you would never see the player move/animate to change postions. You would always see the “eyes close” transition and just show up in a new place in a new pose. And you’d never see the player avatar move on their own.

This solves that. Now, I can show the player animating smoothly from: player walking to a place -> player in a certain pose with freedom -> smoothly to another animation and so on. I cannot understate how complex something like this is to implement in VR. You’re basically ping ponging between letting player’s movements be the source of truth to the animation being the source of truth. And doing this smoothly is an even bigger challenge.

When animating, for comfort in VR, your head movement and rotations are respected to a certain degree. But if you over shoot, the scene re-centers around your new position automatically. You get a tiny bit of similar looking around freedom in “animation” mode in non-VR too. And when animation ends and you’re back in control, you are able to smoothly control the player again with no black screens and smooth transitions. This applies to hands as well in VR, you will see them smoothly follow your hands when you’re in control and fade out to animating pose. There is so much nuance and customization I’ve added to this to support all sorts of animation cases I envision. I think if anybody has played VR games extensively, they’ll be able to appreciate this quite a lot.

The full locomotion system isn’t done but the hardest part is. I will be working on the full locomotion after this chapter as the bedroom scene really doesn’t require moving around that much!

New physics engine

I really wasn’t planning to do this until much later but here we are. The old physics engine was not only inflexible, it would seriously hog the CPU during heavy collisions. We’ve now moved over to Jolt, which is a physics engine used by many AAA games like Death Stranding & Horizon Zero Dawn. It is multi threaded and just more performant in general for larger scenes.

This required me to improve my collision strategy a lot. But you can expect much much smoother tongue gameplay. And higher quality collisions smoothing. For example, if you push against the boot starting this version, there is no danger of going inside but also the movement doesn’t stop jarringly or retract the tongue, your character’s head slides naturally to the side where it can (non-VR only).

I’ve also optimized the spring bones (breasts and hair) separately and made them multi threaded so they don’t block the CPU as much. But Jolt is so fast, I’m considering getting rid of that separate cheap spring bone physics completely and just using it. Not to mention, it has true soft body physics support. So may be one day we can actually do soft deformation against breasts or soles of the feet. This is a super long term goal but who knows.

What’s left?

Well, a tiny bit of shadow improvements. But then it’s just finishing up the dialogue and actually animating everything. Already started on the basics of this. So, not expecting to run into any problems. I might even release a few minutes of game play early. But I see the natural end for this chapter being when Artoria reverse the summoning and takes you with her. And the next part of this chapter will pick up from there.

Thank you for reading!

Bonus: If you read this far, below are some older captures from smartphones with in-game post processing turned off. You can expect 30fps-60fps smooth gameplay even with these graphics on most smartphones!