WIPs: CTZn

[kklors]

Niiiiice! Wish that stuff would be faster.
You clever coding people should totally get more volumetrics stuff running for Indigo. Every time I give stuff like this a chance I fail ultra harddd haha.

[CTZn]

Wish that stuff would be faster.

Aye. Big stuff anyway. A planetary wide layer seems out of reach unless I messed something up, even just the clouds with no atmosphere. Just spent my first hours on this. A kilometric volume is renderable in atmosphere though.

The atmosphere medium is highly performant, could there be Mie and Raylegh shader offsets that would not impact performances crazingly ? Would be awesome, if anyone is to use an Earth simulation for rendering anyway. Oh, and spherical space coordinates out of the box !

Side note, my mother's flat I am occupying is being adjudicated tomorrow, I'll go black for about a couple weeks anytime. I'm alright, was warned but had mislead hopes until today. Darn.

I may add dirty pics (render wise) the next hours (or weeks for the matter ^^).

[yonosoy]

Hope this could be usefull. If this is a meter diameter sphere, the shader is actually trying to render something like "shave foam" (it's mean the density of a cloud is not so high). The impact must be dramatic even out of scale.
In other point, if it's possible, try to differentiate the shape from the material itself and "play" with phase functions in order to obtain some chroma (not sure of what this could mean).
A way could be to manipulate the atmosphere medium (actually not possible). So try to cut the shape through the concept of precence (three spheres: atmosphere precedence 2, clouds precedence 3, cutout whit atmosphere medium 4 (and the same definition)). (maybe...)

Encouragement, you are a spark !!

Just realizing that there's the same problem with sphere 4 than 2... (is not possible the cutout in this way since code is for the medium itself), neither blending whith a vacuum... Then encouragement again.

[CTZn]

Hi yonosoy, I've been through this; my last tests where at a real scale, with adapted clouds parameters.

Planetary wide cloud sphere is a no go. Boxed scattering media included into the atmosphere medium will start rendering nicely and drop to a halt around 1spp (after a few minutes), showing a good sampling rate while the actual spp would stay fixed (image updates showing no sampling progress). Removing the planet sphere seems to give better results but I didn't dig in this direction for the matter. Indigo would apparently need some enhancements in compatibility between atmosphere and other scattering media. The best method is still a box (may it be huge) over a plane with a simulated sunsky. This works good.

I had tabulated data that should make clouds produce rims etc, from a downloaded app that generates all sort of optical data at will, app wich name I can not remember right now (nor can I find it or the data on my disk :/). Also I'm busy as I'm moving place the next days (back into that small room featured somewhere in this forum ).

I've been continuously interrupting renders (killing indigo.exe) in order to save ressources for the next ones, and I kind of became discouraged after quite many tries and tweaks. Hours each day since I started, believe me. No pics atm guys, sorry.

To be continued

[CTZn]

Just realizing that there's the same problem with sphere 4 than 2... (is not possible the cutout in this way since code is for the medium itself), neither blending whith a vacuum...

The ideal setup is 3 spheres (+ planet globe) for one clouds layer: Low precedence atmosphere medium on top, mid precedence clouds sphere, high precedence atmosphere below. This does not render efficiently atm though (if at all).

I was finally able to spot the application that would generate atmospheric tabulated data: http://www.philiplaven.com/mieplot.htm

[CTZn]

Well, that was when I may have put the right data in the right place, because the clouds are not green or pink. I may be wrong but I think that only the phase function is relevant to Indigo from miePlot. So that's a tabulated phase function with a eyeballed scattering (too low ?) and no absorbtion.

The plane and volume have a width of 20km; there's about 16km of them to the opposite corner. The camera is at a height of 33m. The volume is ranging from 1000m to 2000m of altitude.

[CTZn]

don't think I'll be letting it render much more. Some details peaking. This is PT-MLT, I think I should have used PT alone.

Edit: I'm sharing the scene file for them in need Possibly my little parsing script messed the data up, I was expecting 1800 values for the tabulated phase function, but found less than 1200... edit: indeed the tabulated data string is seemingly trimmed, Maya's string buffer limit was hit probably.

[OnoSendai]

nice.

[CTZn]

thanks, not too bad actually ! The shader has some mistake, try one of the other lines for the curve variable (I've edited the previous post with the PIGS). By the way, I'm using Indigo 3.6.25.

PS: many values are rigged within Maya, using MEL in ISL. Also, the volume has still a spherical deformation matching the Earth radius, and it is the radial version indeed.

[yonosoy]

HI. Please inspect every change before to incorporate them. So many thanks.
Good news about speed (x7.5, from 3k to 22k samples per second here), but in part due to the increament of the step size from 20 to 60 (maybe too much). The "tweaks" (sorry for that, but seems that you are able to understand child's words and translate it to an adult language) are not neccesaries but almost twice speed. Some absorption, Henyey-Greenstein phase... Explore it. Some low-level detail could be great, and the intention codifying around PI seems awesome.

Edit: the loss of weight of the file probably means that the tabulated data information was lost in the process of copy and paste of the scene ... and there's a way for a possible cutout through the blend parameter shader (a tree of the three materials described above (an only sphere / box ) , still not sure about this, it's only in order to free the material shader from the shape itself (a negative for the cutout shape) ...
The advantages of a shader instead voxels or a voxel density map definition is a jump (and no kind of fireflies for post-pro AD (since there are no geometry involved)).

clouds_tabulatedPF3.6.pigs

(16.9 KiB) Downloaded 386 times

Happy Christmas to all.

[CTZn]

A: I despise adults, specially the self-proclamed ones.
B: the tabulated data is incomplete; the last number is wrong and should be removed. There are 616 other values missing.
C: Thinking about it, I know for sure that the data needs a conversion that was not done. The script I wrote to extract the data from the miePlot output needs enhancements than I can not provide at will. This is the second reason why the current Indigo output is incorrect. There is no copy/paste step involved.
D: this may be a job for you, but until further notice it is a hobby for me and I currently do not have the time for a hobby (frankly those are bad times for me).

Happy Holidays, see you soon.

[pixie]

Merry Christmas CTZn!

[CTZn]

Meh thank you dear, next year so You enjoy these moments for me, and for them in need above all !

[yonosoy]

Dear good spirit, a return.
Some details (some of them in order to be free soon (bidireccional) of was is necessary at the moment:

Code: Select all

        <bidirectional>false</bidirectional>

	<max_depth>10000</max_depth>
	<max_num_consec_rejections>1000</max_num_consec_rejections>
	<super_sample_factor>2</super_sample_factor>

	<max_num_subdivisions>15</max_num_subdivisions>
	<subdivide_pixel_threshold>4</subdivide_pixel_threshold>

	<precedence>3</precedence>
	<ior>1.00000003</ior>
	<step_size>48</step_size>
	            <absorption_coefficient>
                <constant>
                    <uniform>
                        <value>0.000005</value>
                    </uniform>
                </constant>
            </absorption_coefficient>
                <phase_function>
                    <henyey_greenstein>
                        <g_spectrum>
                            <uniform>
                                <value>0.353</value>
                            </uniform>
                        </g_spectrum>
                    </henyey_greenstein>
                </phase_function>

	noize =  abs(fbm(noisePos, 1))
	paramShapeShift = fbm01(noisePos * 6.0, 4)

No necessary kill indigo.

Is the best try; render this ( or other) maiden voyage to the convergence (maybe near 2000 samples/pixel: fast) and congrat again. Here a vote for a next procedural scene from your hand in the distribution of indigo.

a natural situation_00.pigs

(16.96 KiB) Downloaded 390 times

[Zom-B]

I took the freedom to start to render that beauty on my home network, reaches a max of
380 samples/sec. I'll upload the results later