OTOY Forums

pixym wrote:

radiance wrote:

pixym wrote:This is what I get in 2mn with the Glass sample scene in Thea render beta

for a fair comparison algorithm wise, you should disable MLT and use pathtracing, or await MLT in octane 2.2.
These kind of scenes are a nightmare to render without MLT.

Radiance

I am not at office right now, but it seemed to be the unbiased engine...

yeah, but just try running it with pathtracing (not bidir path tracing, plain pathtracing), and no MLT.
It should take hours.

Radiance

justix wrote:Complete render of my previous post.

1600x1200 - 1600 maxsamples render time 3:26.14

it looks ok to me except some noise area in the liquid that I immediately identify after few minutes but I though that it would clear off.

These look to me like they were modeled with the interpenetrating geometry technique (VRay style). See earlier in the thread here -- that method generates incorrect results *and* takes much longer. It's faster and more accurate with the separate surface technique (PhilBo's linked version).

I can confirm that the Thea Render water glass is not modelled with the interpenetrating geometry technique. Its modelled with the last technique you mentioned. Except the water and the top water surface wich has the same surface (same object).

Cheers.

Chris

Voidmonster wrote:

justix wrote:Complete render of my previous post.

1600x1200 - 1600 maxsamples render time 3:26.14

it looks ok to me except some noise area in the liquid that I immediately identify after few minutes but I though that it would clear off.

These look to me like they were modeled with the interpenetrating geometry technique (VRay style). See earlier in the thread here -- that method generates incorrect results *and* takes much longer. It's faster and more accurate with the separate surface technique (PhilBo's linked version).

it has been modelled with Rhino, I should maybe check the Normals but I will follow the separate technique to see how this last can improve and eliminate the inside noise

I adjusted a model to match (I think anyway) the required method for intersecting dielectrics, but I'm not getting the correct refractions. See the attached images - one shows the wire frame, with the liquid, liquid surface, and the glass all as separate objects - with normals point inward on the liquid, up on the surface, and out on the glass. But if you look at the quick render you can see the liquid is not refracting out to the edge of the glass (glass IOR is currently at 1.55).

The first model I tried had intersecting geometry (because originally the scene was in Vray). Has something changed in the way Octane handles intersections?

Thanks /b

Seems something very specific to the angle of the camera. If I move it even a bit in any direction the liquid refracts properly. Maybe just an oddity of this file. Seems cool now anyway.

b