3rd party imports

I took a first look at building in sl, please bear with me as I'm a bit new to this.

There are plenty of great things to say about sl and the work done by Linden Labd but still.

There are software dedicated to the single purpose of 3d modelling, no matter how hard Linden Labs work, it's not realistic of them to keep up and beeing competitive. People are making great things in sl, but given the option of importing 3d meshes, eg 3ds meshes - sort of the standard since more or less everybody else can export in 3ds, user creativity would take a leap forward in my book.

There must be more people that would like that feature.

Cheers
/Alex

Alex,

While I am no expert myself, I have read enough in the forums that "Streaming" 3d images is much more difficult to do than rezzing a 3d image stored locally. In order for residents to view each other's content in real time, as it's created, it must be streamed. Thus the prim limits per parcel of land and the in game 3d modeler.

There is one tool for importing 3d maps in game. It's a script written by a resident, but the overwhelming response is that it creates very prim heavy objects that quickly use up prim allocation and lag servers.

So while your idea would be great were it feasible, my understanding is the nature of the technology makes it impossible to do.

I'm no where near the authority on this, but I'm bored and thought a quick uneducated answer might help. I'm sure you can find out more on the subject by searching the forums.

The general wisdom is that yes, attempting to stream meshes would quickly saturate the end-user's bandwidth and make for a much slower user experience. Prim-based modeling side-steps the bandwidth issue because the prim data is all procedural -- the only data that needs to be sent is object types, scales, deformations and relative position to other objects.

The freedom of mesh-based modeling will arrive, eventually. I think it's just a matter of technological progress on both the service provider side (Linden's research and development) and the end-user side (bandwidth increases while cost decreases).

It will be nice one day in some brand of virtual world software. For LL to implement this now, they'd have to limit it somehow to keep us melting our connections.

I'm going to be positive for a change:

I would like to see even more bugs added to the current building tools. I also want to see the sphere and torus prims removed. Hey, maybe Linden Lab should just leave us a cube to build with. I love a challenge!.

The crappier the building tools the less competition I have.

Oh yes, I'm in a such a positive mood today. I am at one with the universe!


* kicks the cat *

Be careful if you kick the cat, you might hurt your foot or scuff your shoes.

Instead of kicking the cat, you could just gently put your foot under it, then launch the cat, or maybe play hackeysack with it, less danger of harm to yourself or your shoes that way.

I've never bought the bandwidth argument myself. Textures are many, many times bigger storage-wise, than realtime meshes. (we aint talking 3 million poly meshes here - we are talking in the range of 300.)

Now, at a pinch, I could - purely in the theoretical - accept that the procedural meshes are faster to render; however, the number of wasted polygons probably outweighs significantly.

I think the real reason is twofold.

1. Inworld editor is something LL likes;
2. Adding meshes would require a structural overhaul of the client.

I can provide an example of why high-level CAD tools (as opposed to 3D tools like Maya, Max, SoftImage, aso) don't typically use mesh data for modeling - Link .

I did the above render in Maya from Pro/E data. The original CAD file for these parts is about a meg in size. After I exported the originals to a mesh to get them into Maya for rendering, file size went to something like 20Meg (or maybe 40Meg... I forget). It's no comparison.

Consequently I'd rather see more CAD-like tools and let the client handle conversion on the fly. Streaming mesh files just doesn't make sense to me from what I see.

Chances are, by the look of that model, it's several million tris.

Game models, on the other hand, are in the hundreds, or low thousands.

Assuming each vertex stores the following information:
R,G,B tint (1 byte each)
X,Y,Z position (4 bytes each)
U,V (4 bytes each)

And assuming an overhead of 30% (which is far more than it should be), a 5000 triangle model should be .. 18KB before compression (which should compact it down significantly more).

By comparison (since we are speaking without compression), a texture of 256x256x24 is 192KB.

Meshes are tiny. They only bloat out to 20+mb when you are working in Maya/Max/Lightwave and doing incredible amounts of subdivision, resulting in millions of vertices.

Agreed, if meshes are haevy, just put a limit on them. Remember, the more verticies you have in a mesh, the less depenant you are on textures.

If there are technical issues with this now, I understand. But still, I think it would make for an improvement.

Cheers
/Alex

There is so much legacy data to support, changing or adding more ways for objects to be streamed and textured and making sure they don't break would probably be too much to ask.

I would think at one point, the whole grid might have to be wiped clean and we can start with a brand new client that supports mesh imports, nurbs shapes, pretty shaders etc. Its either this, or SL just trundles down the path of obsolescence. Probably best done before a competitor to SL comes up with something that is more advanced and looks prettier.

For the moment, I would be happy enough to ask for an improvement to the parametric objects we currently have. More flexibility like bend modifiers, shapes and also some form of edge or vertex snapping so we don't need to eyeball things all the time.

The SL avatar .obj mesh file at less than 4k is already almost 372k, and an old quake 3 model of mine at only 497 tri's is 88kb. Also a 256x256x24 image in jpeg2k format should be less than 192kb (I have a nice test file now that's only 15kb).

But the point I was making was that you can get the level of detail I'm showing in a small file size. Smaller than any mesh.

[edit: the above numbers match up to that rendered mesh which is under 400,000 tri's iirc]

Agree. I'd rather they improve the current tools. Some simple Extrude and Rotate tools would have a big impact imo.

Aah, but both those examples can be explained somewhat:

OBJ is a ASCII based format, so it's very filesize heavy, the Quake3 model includes animations and textures (but is probably a good base, since chances are your going to need those in SL as well.).

To be honest, I don't buy the "imports would lead to more creativity" arguement. There are very few shapes you can't do with prims now, with some creativity, and adding a handful more parameters to the prims would fix a lot of the remaining ones (Hollow offset X,Y,Z, skew direction, along with another skew field, a few other things)... Creativity isn't a function of tools. Starax' post may have been sarcasm, but he has a good point, IMO... You can be creative with very litttle. Heck, look at "A Tale in the Desert"... The creativity of people limited to bales of hay and pottery and stuff never fails to amaze me.

Not that I'm saying importing meshes would be a bad thing, I just don't buy the creativity arguement. If you can't be creative in SL because you can't import meshes, that's your failing, not SLs.

No, you can be creative with the current tools; but you would not hand a famous painter crayons and recycled paper.

Right now, what we have is the equivilent of crayons - sure, enough time, dedication and skill, and sure - you will be able to make something that looks reasonably good; but they are still crayons - the tools we have are extremely limited.

In SL, it's very easy to run up against prim limits, that partly exist because of the amount of wasted polygons that are being subjected onto your client. With prims, you wont see production quality graphics in SL (Neil is probably the closest person I know of to succeeding there - and even then, you can see it's limiting)

Look at the linden non-prim trees we have; and try make something look half as good with prims. It's extremely difficult because we cant fine tune the appearance to that level without eating half a million prims.

I'm not saying I disagree. But, using your own example - People who want to, who have the skill, will succeed regardless of what limitations SL throws at them. People who don't, won't, regardless of how much SL helps them along.

I don't mean this as a dig to anyone in particular... lord knows I *can* see the uses of importing meshes directly, I'm a vehicle designer, for God's sake. But everytime someone says "We need mesh importing!" I think to myself, "Oh, great. Now prim-wangs will be mesh-wangs, and prim-bling will be mesh-bling."... I guess my fealing is, so few people are taking full advantage of the tools we *already do have* that I don't feel making *new* tools for people to take even less advantage of is any great priority.

I know this isn't going to be too popular, but I think it'd be an actively bad thing.

As others have mentioned, there's a lot you can do with prims. But more importantly, the ability to have your AV stand there in world and assemble things on the spot in the view of others is important. It means you can build socially, it means we can have real live building classes (compared to which scripting classes are doable but awkward, classes in texture drawing? pretty much impossible), it means things can be done on the fly, etc.

Moving the ability to model objects, or the ability to model "higher quality" objects, out to another external program would harm a lot of this. So if we can have a mesh editor within the client, that'd be great, but moving more of it out of world seems like a socially bad thing to me.

I kinda agree with this, too. Right now, you have a whole lot of people building, on more or less even footing, with people who have been using Max or Maya or something for years. If I came here, and the grid was populated with a bajillion professionaly done imports from multi-thousand-dollar software, I probably would never have even tried to compete.

At least with textures or animations, the buy in for a decent program is cheap-ish... Something almost anyone could afford if they actually wanted it - Paint Shop Pro is 100 dollars, same as Poser 4, less if you find either on sale somewhere. The cheapest *good* 3D program I know of is several times that.

GMAX might have been viable, crippled as it is, but that got discontinued recently.

You might check out Blender, it's free.

Notice I said "good".

Blender, IMO, has vast leagues to go before I would put it in the same catagory as a commercial 3D package.

Sure, but this was kinda exactly the argument I wanted to avoid making - "we have to avoid having external imports, because the limits of prims are needed to bergeron the 3D modelling professionals, without which they'd dominate and make it pointless for others to build."

That's an invalid argument, because a) there's a whole bunch of problems with bergeronning to start with; b) if you're going to say that, you'd have to explain why texture designers or scripters are not bergerroned to any such extent; and c) users with 3D tools still have an advantage because they can bake textures.

In point was that building, precisely because it's part of the SL environment, occupies a unique niche at the moment. It's where SL truly becomes a social workshop instead of just a display and marketing area. We should be trying to make more things be this way, not less. So, by all means make models more sophisticated, but not by pushing the actual construction of models off onto external tools. Keep the editing in-world. And give me a script and notecard editor I can share with a group, so others can see what I'm typing or, if I gave them permission, their cursor can fly around and work next to mine. Give us an in-world art package (maybe they could get some kind of license for GIMP or Texture Maker) that everyone can see. Make it more social, not less.

/Ramble On

It isn't easy building anything decent while in a world occupied by lunatics with guns. Building silly things for fun while in the sandbox is great. But if you want to build something detailed with high quality textures then you need to remove yourself from possible distractions. This means going somewhere quiet, switching into 'Busy' mode and writing obnoxious comments in your profile. So we may aswell just shut the Second Life client down and start up a bug free and specialized application like 3DS Max. I know many good builders that when building will probably spend more time in another application creating textures and posting crap on the forums. (Yes, I'm talking about you!!!)

The current building tools make Second Life fun to mess about with for the first few months. It is quite novel to be able to build and chat at the same time. But after a few months then that novelty starts to wear off. I rarely see any good builders in the sandbox. I just see new players having alot of fun. Some of those new players will one day be good builders and I'm sure you wont see them in the sandbox anymore. Although that geezer who makes the heavily armed teddies in the corner of Newcomb is an exception! Oh, and Mr Grommet with his cars that are more detailed than the real thing!

I'd rather see the building tools taken out of Second Life completely. Not just for the benefit of the builders, but for the benefit of the people who like to socialize. I think its becoming difficult for us all to work out who is here for fun and chat and who is here to create or make money. I'd like to be able to fly around and know that the people I see in game are there because they want to have fun and not because they want to be left alone.

Second Life has become similar to a busy office and we aint got a clue who is busy and who isn't. It would be nice to see Second Life become a playground again.


One more thing:

Nobody has questioned what would happen when that annoying bastard L33tVip3r Omega finds out that he can upload and sell 3D models that he found at www.Renderosity.com and www.Turbosquid.com.

/Ramble Off



Disclaimer - Most of the characters in this post are fictitious. But if there really is somebody out there that goes by the name of L33tVip3r, then I am going to smash your face in.

Can't we just keep the building tools as they are and have in-world mesh support or building?
Such as each of the verticies being created on their own, linked after all are in place, then fuse them so the server see's it as your mesh. The client can see the location of each vertice just as it does with prims.

Limit the amount of verticies that can be linked. Limit the distance they can be from eachother. Just as it is with prims for obviouse reasons.

Unless the vector data for each point is too much for bandwith.

I wasn't using it as an arguement per say... Notice that I said I still would like to see imports.

I'm just saying, it is an unavoidable reality.

I wasn't actually saying, by the way, that we need to 'bergeron' the pros... Just that the existing system put everyone on equal footing in terms of ability. The pros still are damn good, and have the advantage of not only their existing skill but the really expensive tools... It's just that other people are able to compete, because they dont have to buy and learn a terribly expensive piece of software to do so. If users were able to import models, there is basicly no way people stuck with the prim-based modeling system could compete with the level of stuff that could be made.

I'm not trying to 'hold back' the people who are good/experienced. Far from it. I just don't see the need to completely destroy the ability of the average user to produce good content, when, as I stated before, I don't feel we are anywhere near having exhausted the limits of the current system.