|
From: Fernando G. L. <fer...@gm...> - 2020-08-14 11:30:56
|
On Thu, Aug 13, 2020 at 11:11 PM merspieler <mer...@ai...> wrote: > Basically, we'd render to a cube map and project from there to screen. > Ofc. this would be a performance hit. To minimize this, we should add > a mask to only render parts of the cubemap, that get projected will be > rendered in the first place. > > As well, it would be nice, if it is possible to add that pipeline > independent. > So we don't have to maintain a seconds als-panini pipeline. > Not sure tho, if that's currently possible. > > I've found implementations for Quake > https://github.com/shaunlebron/blinky > and Minecraft > https://github.com/shaunlebron/flex-fov > which might help. > > So, is that possible to implement with compositor? > > And if so, how do we (= I + helper) get started? This should indeed be possible with the compositor. As James said, it provides a lot of the tools required to do this, it even allows rendering to specific cubemap faces. However, the implementation can be tricky. As you said, you'd like it to be pipeline independent. It can't be pipeline independent though as what you are describing inherently requires changes to the pipeline. :) My proposal is to build some kind of pipeline "addon": an independent XML file with the necessary buffers/passes which can be included into any pipeline. The main issue is that FG's XML parser doesn't allow multiple include="", so I'm not sure what would be the most elegant way to do this. Fernando |