REAL modeling in SL??

Just realized, if 3D Modeling software such as Maya & 3D STudio Max are using OpenGL for their Interface and so on, then WHY can't we have better build tools in SL which is also OpenGL??

Because Maya and 3D Studio are advanced 3D modeling programs, something SL's not meant to be. Not to mention the additional archetechture that would need to be built into the grid to support the creation of advanced objects. It'd be eaiser to have the ability to upload Maya or 3D Studio models like you can textures.

Probably because LL has to be able to stream it efficiently, because LL needs a way to do LOD on all the arbitrary meshes so SL doesn't become a slide-show, because they'd need a way to approximate the shape for physics on the server, and probably plenty more I have no idea about .

If you have a solution, you can always apply for a job and get paid to implement it for all of us .

Try running a Maya scene animation. Maya often draws one from per HOUR, if the scene is complicated enough.

For movies such as Ice Age I think they would have been ecstatic with render times of one frame per hour.

Yeah, even with a big renderfarm like the 3D animation studios have, it still takes anywhere from 6-48 hours to render a single frame.

Or more.
From an article about Monsters Inc:

"The next step is rendering, which is translating all of the information in the files that make up the shot -- sets, colors, character movement, etc -- into a single frame of film. Each frame represents 1/24 of a second of screen time and takes about six hours to render. Some frames have taken as long as 90 hours."

90 hours is a long time.
(And I think the article is off, movies should be at 29.97 frames per second, not 24, though it might be different for animation, just not in my experience and I've made 3D shorts (not good ones, but rendered none the less))

Most anything shot on film is shot at 24fps, so it would make sense that they decided to continue the standard in digital animation. TV is usually shot at 29.97fps for NTSC or 25fps for PAL, unless the show is shot on film in which case they shoot at 24fps and telecine it up to 29.97fps for broadcast. Traditional animation whether for TV or film is almost always done at 24fps. And stop motion animation is usually done at either 6fps or 12fps.

Off-topic:
http://en.wikipedia.org/wiki/24p explains how they go from 24 to 29.97 for NTSC and 25 for PAL.

http://www.doom9.org/synch.htm has pictures/diagrams .

Touche. I did have some instruction into the Realm Of Video, but it by no means covered EVERYTHING. The course actually included some sound and editing because it was Shooting and Lighting for non-film majors and we wouldn't get the experience from classes that focused on sound or editing.

Wow! Didnt think people would consider "Ice Age"-quality stuff in SL. Twould be nice, but those with 'bare minimum specs' computers would grind to a halt. Its nice to dream tho!!!

What I'm thinking of is standard things like REAL UVW Mapping, Vertex & Face Editing, Deformations, Spline Modeling, extrusions, and so on.

ANd the reason it takes Pixar so long to render a single frame is they use ALOT of hair systems. Perfect Hair is one of the goals of CG. Each hair can currently be defined as either a spline or a particle system, with either also being computationally intensive. Same goes for realistic, physically accurate water & fire.

Realistic skin is also quite difficult to pull off. Ever see a movie like the 3rd Matrix movie, during the "Fight With 1000 Smiths"? Your brain is saying "HEY? They look fake!" Although the demo Nvidia did rendering/shading an actor's head in REAL-TIME on their GForce 8800 Ultra was pretty impressive, it still wasnt close to the real thing.

Nvidia Video Link: http://www.techeblog.com/index.php/tech-gadget/video-nvidias-amazing-human-head-demo

Now if only SL looked kinda like that!!

Subsurface scatter. It's currently near impossible in 3D rendering as the physics of it not only need raytracing (OMG CPU intensive) but are exceedingly complex.
Supposedly the PS3 has this for ALL it's games, but I highly doubt it. The demo they showed had visible bones (with a strong enough backlight) and I seriously doubt that game modelers are going to go to that level of detail for a $50 game.

But still, if you take a sim's contents and duplicate it in Maya and don't even texture it and tell it to render it'll take forever. There's just that many polys.

And now people want things that are either:
a) adding polys (deformations add polys!)
b) render intensive ANYWAY.

Also, custom UV maps mean more data to transfer, which means more lag and load times.

The evolutionary (rather than revolutionary) way forward for this is to add new modes to the existing sculpties system. Those fields in the texture map could quite easily be employed as something else beyond NURBS control points.

Morg.

SL is a study in contradictions. sculpties are a relatively sophisticated bit of modeling but it was easier for LL to add them then to create a real joint system,something that ANY 3D animation program needs to be able to do in the real world.

Aye! Wouldnt say near impossible, as Maya & 3D Max already do it, but it IS time consuming.

I was explaining to someone in-world WHY we cant use Polygons for modeling. Wish I used Voice for that one!! My wrists hurt!!

One thing I would like to see added:

** Custom Bump Maps **

I cant' think it'd be that hard. Just an 8 bit Black & White Bitmap or whatever it is they're using now. Cant think it'd be too strenuous on a modern computer?

Actually, custom UVW Maps wouldnt be that difficult. Make it even simpler by using Spherical, Cylindrical & Cube Mapping. We've already been given Planar.

So far Sculpty Prims are a step in the right direction with modeling.

I'll rephrase, "impossible/near impossible to do in real time."



Transfer of the data of the vertex point locations. Sculpies solved this problem, but have inherint limitations due to the meathod.



I don't know why we don't have them either, but FYI a "bump map" is not the same as distortion. A bump map only adds shadows, even if the surface is heavily "bumped" the edge-on view would show the surface as being flat.

What I think would be a good idea for SL, and that would get them closer to true modeling would be shape maps with a black and white displacement map incorporated into it. For those of you who don't know what a shape map would be, I don't blame you, I think I just made it up. I don't mean sculpties, I mean a shape that has actually been unfolded to fit into a 2d space, much like a uv map, except, along it's border is a color coded system that tells the shape where to connect reconnect in 3d space. Unlike sculpties, you can have acutual complex inorganic shapes, as well as curved surfaces (which does pose a bit a problem with mapping it to a 2d surface, but can be worked out using the b/w displacement map i mentioned) And you don't get that swirlyness of death everytime you don't do a sculptie quite right, or I could just be a n00b. You could use a variety of modeling methods and texturing would be easy, all you would have to do is UV unwrap it and you get a perfect fit. Sliders could be incorporated so that one could adjust the heightmap distance aswell. I don't know if this Idea would take more processor time than current sculpties, but I am pretty sure if it is greater it can't exceed it by much. I really think for now image maps are the way to go because of the current support.

It really is just simpler to use a 3rd part program to make a sculpt map of that.

You won't get your deformations, but those require more mesh verts than SL can draw per prim.

So just how many LOD levels does SL use ? Most games use 3 some use up to 5. If LL were to set a max poly count for meshes at each LOD I can't see why it should cause any significan drop to frame rates over the current prim heavy creations we already use.

I think the problem is that SL doesn't have a way of on-the-fly calculating LOD levels. All of them are hard coded (exept sculpies because they are JPG2000 format which loads in the blurry, less-blurry, almost, done fassion which causes a LOD change just because of the way the data is interpreted).

A custom UV map would essentially mean uploading multiple textures for one object. So instead of just one for a prim, you've got two. And for sculpted prims instead of two you've got three. Also for sculpted prims it would make little sense considering the resolution you're limited to. If they were expanded to 256x256 I could see it being more plausible though.




What you're describing reads the same as sculpted prims (to my eyes), except what you're suggesting seems to be unwrapping a complex object into multiple parts on the same texture. While that would seem to make sense and UV mapping probably would be easier, sculpted prims as a single 64x64 texture is really a better way to go.

Yeah, I think s/he is looking for some way to do this:

http://www.jawa9000.com/gallery/Trippe/Detail/Texture-map.jpg

well now comes my time to contribute my two cents...

in applications like Maya and 3DS Max the process of RENDERING takes hours mostly because the concept is intended to archive high quality images with hardware which cant produce them in real time. if you have a scene with 1Million plus polies or extreamly detailed and complicated models[in various formats, Nurbs Sub-Divs polies ect ect the list goes on], yes it will take a lot of time to render however the concept of game development is to produce content which can be rendered in real time such as animations which are only meant to be viewed not archived (saved to disk)

these features [RT rendering] are prevolent in the applications in the viewports themselves otherwise you couldnt tweek aspects of a project without first rendering out sometimes massive frames and even in the situations where you do render frames the process can and should be broken down into pieces to reduce render times overall.

the greatest diffrence between SL and high end 3d apps is the support and performance of geometry and the expected level of detail assosiated with it based off the machine running them.

where maya can handel a lot more information in real time it would be nearly impossible to stream it efficently to multiple clients without explicit localy loaded content and a quality casheing system.

That said I do belive that the variaty of tools and objects you can create in Maya or the like dwarfs those of SL which is unforturnet however LL currently dosent have the infastructure in place to incoperate many of the features applications like Maya can use or the accuracy of them for that matter.

I think the engine could use a more complex and robust modeling toolset as well as a process of animation[and rigging, we cant for get the rigging] we currently dont have any way to apply these tools to objects on the grid...

Yes, but at what point can something be implemented without bogging down the network with data, yet be rendered in real time, while also being easy to use without the need of an external application (hopefully), while maintaining the power of the original [Maya] tool?

For the sake of argument: subdivs.

SL could handle subdiv objects relatively easily rendering wise, but...they have hierarchy with SL currently lacks (there's object -> prim and that's about it). Though if given hierarchy, one could exploit that to create intensly complicated sub divisions that could both bog the network (raw data) and the client (rendering it). So it's limited in power--only so many levels of hierarchy, only so many divisions.

For the sake of argument: deformations (i.e. Bend and Twist)

If we add these, they need to stack (hierarchy) so I can bend something two different ways and get funny shaps. Then twist it and bend it some more. Maybe add a squash and a melt.
Again, network bogging (raw data) -> limit the number -> lose power.

And in either case we need hierarchy to implement, at which point people will want bones and Inverse Kinematics arguing that "we have hierarchies now, but why not bone structures?".

well i'm not making an argument for SubDivs just useing that as an example. the point i'm making is that building in SecondLife is basic for a reason. the more complex something is the longer its going to take to stream.

it is true that we really dont have a benchmark for SL as we usualy know it. mostly the published data is on the number of prims not on the performance of visable polygons.

I do think we can use a hierarchy though i just dont think it should stop with that. we can do without IK considering how animation is handeled. BVH files animate by frame samples and Iks would be something that is used in the parent producing program.

the biggest thing to consider in this argument is that in secondlife to have good performance ACCURACY is what is taking a hit. limiting us to a budget will do only so much to slow down creation of negitive explotation.