Thomas de Groot <tho### [at] degrootorg> wrote:
> Crystal replaced by a mesh2 version. Less grainy by not using
> stochastic. Render time about 27 minutes. Collected photons, about 1 GB.
>
> Additional scene developments pending.

It's looking better and a little less noisy now. It also seems a bit more
bluish. Looking online, this mineral has a range of hues spanning between
colorless, pale yellow, and pale/bluish green. All seem to be valid.

>> How long did this take to render? The file name says 004d, so I really hope it
>> wasn't 4 days ':/
>
> Oh no! That is the version number ;-) This render was pretty fast: less
> than an hour iirc.

Thank goodness, lol. That makes more sense.

>> I'd like to discover a cheap and realistic way to add internal fractures to
>> mineral renders. Apophyllite is one of those minerals prone to being found in a
>> fractured state. I have two ideas in mind, but both are rather expensive: 1)
>> height fields intersecting not only each other, but also the crystal shape
>> (which is itself an intersection); or 2) isosurfaces. Both can be very, very
>> slow. Sometimes I wish media had an ior block.
>
> Ah... yes indeed. I have not considered your first method, but
> considered the isosurface one. However, I also cringe at the implied
> render time.

Both methods are going to be slow, yeah. And care must be taken when choosing
the fracture planes for a given mineral. Apophyllite appears to exhibit basal
cleavage, so its fractures tend to be horizontal more often than not.

And then there are clouds... Similar to fractures, the clouds you see in
minerals are often reflective/refractive gas or liquid pockets, and so they
might not easily be replicated with scattering media. (It seems to me that there
was an experimental version of POV-Ray a long time ago which allowed one to
specify the reflectivity of participating media. I wonder what happened to
that?)

> Another thing I would like to do is model more asymmetric crystals, like
> they occur in nature. The KrystalShaper models are too perfect for our
> grubby little hands. :-) Maybe by carefully manipulating the set of
> planes in the intersection, or working directly on a mesh2 model
> converted back to .obj for instance, and load it up in our favourite
> modeller...

Personally, before going that far, I'd just change the distance values of each
crystal face in KrystalShaper, and not even mess with any other program save
POV-Ray. KS is great like that; you can make entirely new minerals if you wish.
And there's always doing what I did: copy the HKL+distance data to POV-readable
arrays. Then you could take it a step further and add some randomization before
the HKL+D data is used to make the intersection.

Besides the symmetry, another thing I noticed about the crystal libraries in KS
is that the secondary and tertiary faces (modifications) tend to be somewhat
exaggerated. My alum render, for instance, was not what you'd consider a typical
specimen; you'd usually only see eight main faces in real life. And the
apophyllite model is the same way: in nature you usually only see cuboids
truncated with a steep pyramid. (The bevel between pyramid faces is atypical.)
None of this is a bad thing, of course, and it helps to show us just how a
mineral can form secondary & tertiary faces.

Sam

Post a reply to this message