On dit que cet endroit est le bout du monde, que comme la vie, il n’y a rien après. Que seul les désemparés venaient se perdre ici à tout jamais. #indiedev #gamedev #indiegames #screenshotsaturday

Edgar contemple les tombes et les mausolées. On dit que les dieux sont enterrés ici. Bien loin de la cacophonie des batailles à leurs noms. Jadis, lui aussi se battait pour l’un d’eux. Les dieux sont partis depuis fort longtemps, seules les pierres survivent à la folie des hommes.

Des dédales à n’en plus finir et le temps qui semble s’être arrêté depuis une éternité. Edgar poursuit son chemin, cela fait des jours qu’il navigue à l’aveugle la région du Roi Fou...

The main issue is how I blend probes. It’s too simple right now, and the grid becomes visible when nearby probes fluctuate between cells. It’s worse with sparse probes. I’ll try something new soon! But I want to avoids building a tetrahedral volume like big engines do...

To easily get the surrounding probes from shaders, all probes are spatialized in a 5x5x5 grid. Still, I have to sort them in the shader to find the nearest four probes among the surrounding ones, and that’s not great and bloats performance right now.

How it works: I place probes in the scene, which are just entities from my editor’s point of view. The baking system is pretty simple: I raycast a few times from each probe to get the irradiance of surrounding objects, then aggregate the results and store them as 9 spherical harmonic coefficients.

Also, the SSAO target texture is now half the size of the framebuffer, then upscaled later. This resulted in a significant performance gain with almost no visual cost.

Thanks; it's a symbol, a totem, a direction even, of the game I'm trying to make.

That was a super interesting talk, thank you very much! Indeed, probably a bit of overkill for now, but there's some clever tricks there, like reusing/cached ray projections, how they blend probes...

And the more I read about lighting probes, the more I’m thinking that I could scratch a basic implementation in a week, but I don’t want to fall into some kind of rabbit hole...

I’m thinking about basic irradiance probes, it could be a good enough solution (at least for now). Each probe would be an individual entity, and could be placed manually at first. The results of the computed irradiance are easily serialized too.

The game is largely populated by static entities, so baking lightmaps would probably give the best result. But it’s also pretty complex to implement in the current engine right now...

And I judge how good a programing language is, through this exact metric: Its ability to let me stay in that flow, in that process.