"Kenneth" <kdw### [at] gmailcom> wrote:

Hi Kenneth,
Thanks very much for writing out the code to demonstrate those media effects!
:)

That's quite interesting - I've only run a few variants, and it already shows
effects that I might not have predicted on my own, like the interaction between
the reds and the blues with the green cloud.

I think that if you look over the scene, and think about why someone might want
to use it, then a few things come to my mind.

(I'm just gonna spew the raw contents of my overly caffeinated brain...)




First, I'd suggest that the scene be broken up into a grid of smaller clouds so
that a few different variants could be compared, and decrease the need (and
time) to render 2, 3, or 9 different images.
(Less overall media coverage of the visible scene would also likely result in
faster rendering.)

I found this video just yesterday, and I think this guy had some good ideas as
well:
https://www.youtube.com/watch?v=KOJFM-FOJVg
There's nothing inherently in the image that lets the user know exactly what
settings were used - so maybe a [encoded] caption of some kind could be included
under each cloud.
(perhaps have a long vector, or an array, or a macro take that encoded list and
generate the scene that way, to immediately reproduce it)
I'd also include a _grayscale_ long-box.
And maybe a light source or a white emitting cylinder or long-box.

This type of experimental scene unfortunately suffers from its reliance on the
end user to come up with the values and combinations to be used, and hopefully
home in on some desired result.

As useful but simple and immediate enhancement - I'd just include some sample
values in the code to be rendered to show off media effects that you are
learning are possible.  You greater experience and knowledge of your own code
would allow you to provide examples that others might not come up with.


I think that after reading Ansel Adams' photography trilogy, where he explains
the basis of the Zone System, that a truly excellent scene should be a tool
whereby a user can have some idea of what they want, and iteratively home in on
it.
It should provide a logical, methodical way of reliably establishing in no
uncertain terms how to get (very close to) what the user wants.
Like a manually-guided octree search.  Just keep picking branches and updating
until you arrive at a leaf node.

If there is an equation of the resulting color between a solid or a light or a
media[um] and the cloud media - that could be graphed in rgb color space, then
that would be a useful tool to look at as a guide, and perhaps provide
inspiration for further enhancements of the scene.
Somehow declaring the visible scene as a pigment and using eval_pigment() to
give a readout of the perceived color at certain points in the image might be a
useful tool to get a better grasp of how media [numerically] interacts with
other scene elements.
Constructing an animation to show the progress of a certain value from one
extreme to another would be a great addition as well.

The problem here is that there are so many variables that can affect how the
media will look, and the effects it will have on other scene elements.
lighting, reflection, shadows, self-shadowing... RADIOSITY.  :|

Perhaps a number of different "environments" to test the media clouds in.
A model of the Cornell Box might appeal to an orderly scientific investigation.
https://en.wikipedia.org/wiki/Cornell_box

Or at least a few regular, predictable (and more boring) scene objects with
known dimensions, so that the effect of different "densities" (overall
path-length  https://en.wikipedia.org/wiki/Beer%E2%80%93Lambert_law ) can be
seen.
Perhaps a nice way to restructure the test scene would be to put a cloud in the
center, surrounded by a ring of wedge-shaped media containers in front of 50%
reflectance (gray) boxes with an HSV strip down the middle or an HSV color
hexagon in the center.

Probably with even ambient value and orthographic camera.



Thanks again - I really love what you've done with the clouds and the
fire/smoke, and I hope I've given you some fun an interesting ideas.
I appreciate that you've taken the time to work a lot of this out, and I hope I
get some long-uninterrupted round-tuits soon, so that I can experiment with some
media scenes, and have them work without too much frustration.

Post a reply to this message