Topic awaiting preservation: Max Slop is back

After my last server name change, I didn't bother with Max Slop until I had some more material ready. Well, I finally got some more material ready. And, with a little luck, this will be a good home.

Max Slop

Three new little ditties at the bottom.

Pretty soon I'll be leaning more towards combining 3d and 2d.
I'm even thinking about a series about normals. I've already got some normal junk in the wings, but I'm thinking about getting into the nitty gritty. You know me.

If something is broken, please let me know.
And other comments welcome.

Woo! I've always loved Max Slop. For one, that projected painting map in the last tutorial was one little trick I didn't know, but should make painting textures a hell of a lot easier for me (I can model and naimate fine, but my textureing and charactre rigging really need work, as do my lighting skills).

Yeah, I love that projection painting trick. Seems a bit unweildly at first, but not so bad once you get the flow down.

Time to ramble for a bit.

One thing I'm a fan of is using reflection maps over raytracing reflections.
One trick I use on occassion is to use a reflection map to mask itself.

Do up your Standard material.
In the Reflection slot, use a Mask map.
In the map slot of that, toss in your reflection map.
In the mask slot, copy the same reflection map.
Then, in each one, you can use Output roll-out to tweak.

What that basically does is fake intensity by manipulating opacity. Kind of like highlights from HDRI.
See?

Of course, not without it's problems. For example, wanting to reflecting other objects in the scene.

How to reflect objects and keep the fake opacity based highlights?
I've tried quite a few things and I didn't like any of them. Although, Shellacing was pretty close to what I want.

Then I got to thinking about the vampire effect as mentioned in the other thread.
Ah! That was the last piece that I needed to get what I want.
Actually, the vampire effect and it's reverse.

Start with a reflection map. Rather, an environment map.
Rather than put this in the scene environment, put it on a sphere with inverted normals.
Make sure that self-illumination is 100%.
Right Click > Properties, turn off shadow junk and make it invisible to reflections/refractions.
This is the vampire effect.

Copy the sphere and scale it down just a tad.
For this material, slap on the same reflection/environment map in the colour slot.
In the opacity slot, copy the map.
Again, self-illumination 100%.
On this smaller sphere, Right Click > Properties, and make it invisible to the camera.
This will be the opposite of the vampire effect.

Toss in a plane and a teapot and some other junk.
Toss in a light or two.

On one of the objects, give it a raytrace material and make it reflective.
Also a good idea to make the reflection additive.

What you should end up with is an object that reflects other objects.
Shows highlights from the smaller environment sphere, but doesn't reflect the bigger environment sphere.

In the material of the smaller sphere, tweak Output on the colour and/or mask to taste.

Tada.
Nice and lovely reflections and hightlights without all of the rendering overhead common with HDRI.

Although, still one last tweak to add.
Shouldn't be too difficult.

Gotta fly.

A very quick babble of normal things to come.

On the left is some some very basic planes and chopped boxes for house trim.
On the right is the same but with normal maps applied.

The difference is subtle, but can make a big difference.
Also, I need to tweak the vertical corner piece.

The normals maps:
normaltrim2.jpg - for the corner piece. Like I said, I need to rebuild this one and beef it up a bit more.
normaltrim3.jpg - for those square pieces. A whopping two planes chopped from a box in the ratio 4:1.

I'll see about getting some serious chunks up for ya'll to play with.

(Edited by warjournal on 09-14-2004 04:10)

(Edited by warjournal on 09-14-2004 04:11)

Interesting.

*Skaarjj gives warjournal a hearty slap on the back*

Keep up the good work my friend. I think you have much to teach us all. And all lessons we would do well to learn.

Just for you, Skaarjj: Looping Noise.

Taking a break from the normal material.
Before I get back to it, I really want to get the fundamental procedural stuff out of the way.

Holy bite my ass!

My sub-conscious had the answer to another problem. Just spend around 3 hours of fucking around with paper and pencil trying to figure out what it was trying to tell me. Doodles of donuts, 3d space that has been polar coorded, symbol for infinity, and a graph of r(t)=1+cos(t*2). Even some doodles of magnets with several variations of lines of force.

Seems like everything that I was doodling kept going back to the infinity symbol and the polar graph in one way or another. Rho and theta were somehow the winning combo, but I just couldn't see the connection to XYZ.

Then it hit me - the answer is in 5 dimensions! The whole time, the answer was in combining XYZ with rho and theta! Not connecting, but combining!

XYZ are the primary inputs to describe the space (procedural), then modified with RT.
Since RT is polar, it can be made to loop in time in a massive variety of ways.

Fuck me I'm a genius.

Time for another epiphany.

Remember when I was messing with AI and colours?
Well, toss in some some trig in the right places and you end up with a procedural that tiles all directions - even in 5 dimensions if you want.

Perfect.

Now, I did mess with using AI and trig, but I was using the trig for UI reasons. In other words, the trig was in the wrong places.

*my drops his jaw*

Bloody 'ell...this is definately something I'm going to have to play around with later on today...

Not sure when I'll get to playing around. We got family coming in from out of state, so that will be taking up a decent chunk of my time.

A few quick notes:
- When randomizing the nodes, I find weights with a range of -8 to 8 to be fairly decent. Bigger range didn't seem to make that much of a difference, but don't discount it as something to try.
- Random threshholds I prefer range -4 to 4.
- Expect gradients and nothing terribly fancy.

Got to thinking obout 2d time and realized that not really necessary given how AI works. Instead of t1 and t2, just plug
cos(t)
sin(t)
for two time inputs based on the same number.

The same has to be done for XYZ. If you just go
cos(x)
you will get ping pong and not the right kind of tiling.
So,
cos(x)
sin(x)
cos(y)
sin(y)
cos(z)
sin(z)

That's a total of 8 inputs to start fiddling with.
I think 5 hidden should be good.
Of course, 1 output for B&W.

Few other things to figure out.
For example, I'm not exactly sure how I'm going to test this stuff. Hopefully loading a model and shading it with code won't be too hard to figure out.
But the heavy thing on my mind is the gradient-ish output of AI. I'm sure there is a way to mix in some of the ideas from other procedural junk to mix things up a bit. I do have an idea based on Worley's work, but no idea if it will pan out.

All of these good ideas and still so much more to go.

If you do any fiddling, Skaarj, I would like to hear it. You never know when the tiniest thing will set off another avalanche.

Holy crap #1
Took a mere hour to get the majority of the code to test.

Holy crap #2
It works.
I'm going to say that again.
It freaking works.

And here we have 4 random examples:



The range for the tiling is -pi to +pi in all dimension. This means that the inputs have to be scaled appropriately. Did those examples with divide by 15. Since my testing area is 200x200, that will give a tile approximately of 2x2.

I would say that is definitely proof of theory.

Now it's just a matter of getting beefy.

Man, I'm good.
Why, it almost hurts.

Added ability to load and save weights and threshholds. This means that they can be modified. Granted, by hand, but it's something.
And another dinky thing here and there.

First is a loop through z:



Looks like it ping pongs, but it doesn't. Pay close attention to the black dot that grows and shrinks, and compare it's life cycle to the rest.

Now a loop through t:



I'm sure with some practice I can get some better examples.
Oh, I actually bothered with some real math so that all dimensions tile in 100. That is the range of these simple animations. However, the doc sizes are 150x150 to show that x and y still tile.

Between those two, you should be able to see where they intersect. For one frame, both are exactly the same. Neat, huh?

I'm going to take a break.
In a day or two I'll give RGB output a shot. Always more interesting.

(Edited by warjournal on 09-22-2004 04:44)

Colour:



Not sure what it's good for, though.

And now, the generic secret to making any texture tileable in 2d:



Same idea applies to more dimensions. Mostly likely, 1d at a time.

(Edited by warjournal on 09-23-2004 04:23)