REAL modeling in SL??

So was I.



IK is how animations are created. It's what keeps a bone from growing longer when you move the end around.



This is the problem and the point. You can't implement anything into SL that will cause a performance issue on slower machines. It won't be long before my laptop can't run SL anymore, and when that happens I won't be a happy dragon. I no longer own a desktop machine (though if I buy these parts from a friend and get some DDR2 ram I will).

We can look to the future and say "this is the best way to do this for SL, but the internet and the average person's computer aren't up to it yet and won't be for 10 years" on some things, but if that's the case admit to it being out of reach for now and come back in 5 years.

But for today we need to find ways to increase performance, stability and scalability. Feature wise we can do things that don't add too much bandwith or require too much CPU time.

not entierly true, if you boil down a joint hierarchy to its most minimum states its just offset and rotation (from origin, which would be parent bone)

That's like saying "If tractors and airplanes both use gas, how come tractors can't fly"
3dsmax lets you choose whether you want OpenGL or DirectX7/8/9, and you can switch between them at anytime, so its completely irrelevant.


Bones sound like a great idea actually, but those would probably come with Havoc 2 (or 3 or whatever), since the current physics engine doesn't support "joint constraints" like most modern engines do. Adding a simple feature such as that will make lots of crazy stuff such as ragdolls possible And there will be no reason to import any IK from 3rd party programs, it would be doable in the viewer.

yes, true. also let it be known that just because the maya exporter can export maps at the resolution of a 4096 square you shouldnt use it to upload. it doent help anything...

I've already been attacked by someone who rezed 36 unique MICRO prims with huge sculpt maps.

in reference to how objects are assembled, SL does use Polys. Not entierly sure if its produced by a formula or what but I know that you can grab the 3d info with Glintercept. the objects aren't created as normal 3d applications usualy do (really low divisions) they always have at least 1024 geometry on them to acomidate SL deforms (twists skews and basic scale) I belive its all basic poser blendshapes/morphtargets.

Old post, but I am going to comment on this one issue anyway. Subsurface scatter is ***not*** done via raytracing, it uses a cheat. The physics of it is in fact damn near impossible to do at all. "Real" subsurface scattering would involve a media, under the surface, with variable levels of color, reflection, density, etc. POV-Ray, which is an attempt to do a 100% physics based engine for 3D, has trouble doing this, and it **has** media.

In actuality, SSS is done far more simply. One uses two surfaces. The top surface is semi-transparent, the under surface is generally one without transparency, though one could make it semi-transparent too, to allow you to see stuff "below" that too. The trick is simply this. When a ray hits the surface, it normally bounces at a "predictable" angle. This angle with always be a set angle, depending one the properties of the surface it hit, and any things like IOR (internal refraction), which might "bend" the ray as it passes through the surface. SSS works by slightly randomizing where the ray gets bent. So, if you have an IOR of .8, the *actual* numbers generated, semi-randomly, across the surface may range between .7 and .9, or some such variation. Since the ray doesn't *enter* at the angle it would have normally, it picks up colors from the subsurface that do not match what it would have with "normal" IOR. The result looks like real skin.

Real "physics based" SSS would, as I said, use media. The problem being, *that* is mathematically intensive, nearly impossible to get right, and also not perfect, since its kind of hard to generate media with blood vessels, etc. running through it, at least not without a *huge* amount of extra meshes, multiple media types, etc. Not even Pixar or ILM would be insane enough to use that method. POV-Ray BTW doesn't support SSS at all, its not a "physics based" solution, so its not included as an option. Maya and the rest are not so picky, so they use SSS *instead* of media.

Now, just to be clear, this means that the difference in the math between "normal" traces and SSS is almost nothing:

Normal:

Did ray contact an object?
Yes - Get surface color. Is that color partly transparent?
Yes - Does it have IOR?
Yes - Bend the ray path by the IOR, then continue to the next surface.
Repeat 1-4, until you hit the light source instead.

SSS:
Did ray contact an object?
Yes - Get surface color. Is that color partly transparent?
Yes - Does it have IOR?
Yes - Bend the ray path by the IOR +- random scatter, then continue to the next surface.
Repeat 1-4, until you hit the light source instead.

Yeah, that single random number is going to cause **huge** lag. lol

Media, if someone uses that... Now that is going to take a scene that is done in a pure physics engine like POV-Ray, and renders in a half second, and make it render in anything from 5 minutes to 5 hours, depending on how much media there is, and other factors. Its also why IOR is "only" calculated based on when you hit the first and last side of an object, not throughout it, like the real world, and thus why you can't have objects with "variable" IOR, like exist in the real world. Figuring *that* would make calculating SSS via media look like building a 747 from rice grains, and probably take about as long to finish. lol