Red

[efflux]

I found a little time for TG2 so did some work:

http://www.renderosity.com/mod/gallery/media/folder_154/file_1537971.jpg

Amazingly this planet only took a few hours to build but that's because TG2 has such an easy workflow. I mainly worked on atmosphere and terrain, experimenting a bit more with some of TG2's more realistic capabilities rather than the Mojoesque stuff I did previously but I did end up messing the terrain around a lot. This render looks quite detailed yet it is the simplest planet I have created. A lot of the work is done with the strata and outcrops shader which is really a lot more versatile than I first thought. In the TG2 documentation section, Piliph posted a link to some more info with pictures which are worth looking at if you haven't seen it yet. Unfortunately I got my head around this shader before seeing these pictures, otherwise I'd have learnt a lot faster:

http://flore-de-bourgogne.chez-alice.fr/tuto1/strata.html

I set the smooth surface patch size quite high in compute terrain. This seems to greatly aid the result when using the strata and outcrops.

There is also one perlin displacement function going into a surface layer but that's not doing a whole lot anyway.

The rest of the displacement work is tweaking one terrain fractal which is redirected. Noise settings and warping are tweaked around a lot but crucially, the displacement spike limit is set to a very low value to smooth out the terrain. This means you have to up the displacement amplitude value a lot to compensate for the lost spikes. Generally, I do not like those peaks or billowy hilltops that you get with a perlin fractal unless it's specifically to create some kind of tower structures to displace in weird ways.

I may post another earlier black and white WIP render of this planet where slightly more mesa like structures were obtained. It was just an intermediate test but I like it so I'll render again with higher settings.

Once again, I did a little post adjustment in Lightzone. It's worth checking this app out. It's a totally awesome tool. I've been using the free Linux version but now I'll probably buy the Mac version. It seems to function really well on my Mac and sadly the Linux version has not been upgraded to V3. The basic version of this app is fully functional in it's editing power but doesn't have a few things in it's image browser compared to the full version. It's worth every penny of the cost.

[rcallicotte]

Oh, oh.  I want some!  I want some!

Nice work.

[dhavalmistry]

oh I want a piece of that baby too...

[old_blaggard]

Very nice!  This is a truly fantastic render!

[efflux]

Thanks guys.

A couple of other points I should mention here.

The original render was 3000 x 1500 so downsizing it makes it very smooth but I have upped the samples in clouds to 64 and 32 in haze. Higher than previous renders. Haze can get away with a bit less because some slight graininess in the haze can be OK. Slight grain is still apparent in the clouds at full size and more in the haze. Downsized and it's very smooth. I'm testing rendering bigger because I will print some stuff in future. Looking at my sequence of previous work it's getting smoother and smoother. I use detail 0.8 and AA 6. I may go further with the quality in future but this planet has no reflective shaders like some of my others so render times weren't as bad.

Another point is that I started experimenting with a dot product graph. Get ray origin and get normal to dot product but I'm not certain of the best way to use this for blending colours yet. I've had it sent to abs scalar and to any input of a build colour. It works but not exactly as I'd like. I need more control of the colours. In the render above (where I subtly blended this function in) it may be having more of an effect than I realize. Very subtly more bluish colours on some surfaces away from the ray origin or at least at some angle from the ray origin but it's not right yet.

[old_blaggard]

I would be interested in hearing more about the dot product graph - what purpose are you trying to achieve with it?

[efflux]

OK. This is what I want to achieve. I want a value that can change the colour of a surface depending on the angle of it's normal from another point. By sending two inputs into a dot product, it outputs the cosine of the angle between the two input vectors. It's just what to then send it to and what value system is being output. Then how to get this to a value that can blend or create colours. I've only got it working in a kind of hack fashion.

[old_blaggard]

Ok, that makes sense, although I never would have thought to play with that .  I hope you keep on working on this.

[Athesdan]

Splendid work efflux !

[efflux]

Here's a really simple little function graph I often use to get colour changes with altitude. I did use it subtly on the render linked above. I plan on extending this idea further:



If you divide a get altitude node by say the highest altitude value of your terrain (enter this into the constant scalar attached to divide scalar), it will get you a suitable value to blend two colours. Send this to a mix controller on a mix colour node. Select two colours in constant colour nodes for inputting into the mix colour node then plug this to somewhere that expects to receive colour. It's dead easy but there are further doors open here to extend this idea. You may wish to change the dividing value to change the rate of colour gradation with altitude.

I plan on extending all these simple little ideas to more complexity and eventually put all this info somewhere. Probably on a my website (not built yet).

As for the dot product idea. I have not solved that yet. What does dot product output? Is it radians?

[old_blaggard]

All the dot products I've worked with have output radians, which means you'll probably have to convert to employ the cosine function you'll want to use.

[efflux]

That's what I thought. I'll experiment further.

[mogn]

If you transform your input to a 0..1 variable you can use this 9 colours equdistant gradient color:

[efflux]

Thanks mogn. Yes, these are the kind of ways I'm thinking to extend this. I haven't checked your file yet but will do so. We have no colour gradient so some limitations without creating these graphs.

[Volker Harun]

This is marvelous!