I’ve been on the voxel bandwagon lately, and hoping for partially transparent blocks. This means large objects with transparency z-depth issues.
Looking around the internet, I stumbled upon this technique that a couple of guys invented at NVIDIA for Weighted, Blended Order-Independent Transparency. They have a couple of youtube clips that show very promising results. I would be very happy to achieve that level of quality from an order-independent rendering technique.
Unfortunately, I’m not that fantastic with shaders yet and I’m unable to understand how easy it would be to implement with unity, but they say it’s all made with fairly basic shaders. They mention using two render targets, but also mention a fallback technique in their paper, linked in the article.
What do you think guys? Is this something worth getting excited about?
It looks like the demo scene uses Camera.targetTexture, a pro only feature. Original articles states there are “fallback paths for GPUs or APIs that do not support multiple render targets or per-render target blending functions”.
Still trying to figure out if I can make this work in unity free. Help from anyone with a knack for shaders would be pretty nice…
Bump. I wonder if after three years passing anybody had any progress with this issue?
I downloaded and experimented with that BlendedOIT, but apart from serving as a tech demo just proving that it’s technically possible, it’s not good for anything else. My FPS drops from 30 to ~5 when I add that script to my camera and render with it just two simple semi-transparent umbrellas, and that’s in a very simple scene.
So after so many years passing we still don’t have anything better for rendering semi-transparent objects than the basic camera z-sorting?