|
|
|
[
Permlink
| « Hide
]
Qarl Linden added a comment - 23/Aug/08 10:02 AM
see MISC-1495 to vote for the opposite.
I'd prefer working towards importing meshes. I feel that more resources and time would be necessary to incorporate and learn new modeling tools than to use exisitng RL tools that everyone (who uses such tools) is already comfortable with. This might allow for "Turbo Squidding" but it would mean a more even "playing field" and a more enjoyable user experience.
Tough choice. I'm voting for this one because I believe that from a high-level perspective it'll benefit SL more to have this capability than to improve the current tools. However, I also believe potential adverse effects - on the inworld building community, on streaming, on 3D model content control, aso - can be regulated by placing a relatively high cost on mesh uploads. In addition, meshes could and should make use of a content identification system (something we were once told was in the making).
Some of us have given a fair amount of thought to this issue and would welcome a chance to participate in a discussion on how meshes might be integrated into SL ... whether or not meshes receive priority over prim modeling tools. add mesh support! ADD MESH SUPPORT! sculpties are really an imperfect solution for bringing meshes to sl. i mean i use them constantly, but full mesh support would be such and improvement, and lead to much better content in sl. and the thing that makes sl so much better than most other mmo's is the user created content, something that linden lab seems to forget at times
Given the maturity and availability of mesh creation tools, I think development time would definitely be better spent on mesh import and optimization initially. The ultimately goal should also be to improve the inworld build tools as well, but it should not be the priority.
Both are very useful but I give my vote to mesh import for the reasons stated by Crisitano. The prim building tools are extensive as is and improvement might be incremental at this point. The impact of mesh import would be massive by comparison..
The building community can help in having those with no mesh-tool backgrounds learn how to create them. I'd rather have meshes. They can do things that single regular prims can't; sculpted prims, in addition to being a bit less flexible than meshes, have a vertex count that's – for a fair amount of stuff – needlessly high. I'd much, much rather be able to make my own meshes than fiddle around with prims or sculpties.
Oh dear god yes! Not a tough choice at all for me.
Do not lower the bar for building any more. It's wonderful that everyone can build within minutes of entering SL, but they can already do that. The lower the building bar the lower quality and higher quantity of crap that comes out. If you improve the existing building tools, all you'll do is encourage more prim spam. Allow mesh import and the people who have the ability and have taken the time to hone their skills, will leap at the chance to use actual professional grade tools. Plus, you (LL) seem so keen on SL becoming a viable software/business/commercial/media development platform. This advances all of those and adds design to the list. Some people have already made SL their lives. For those talented people who haven't, the number one reason is that SL is a waste of time. Even after having created something amazing in SL, they can not turn around and apply the technical skills they've learned to the real world. Using the prim modeling tools is not a marketable skill outside of SL, and only marginally marketable in SL. From a business point of view it's all about ROI. Improving the prim tools has reached the point of diminishing returns, but the potential of mesh import has only been gently flirted with in the form of sculpties. My Recomendations: This is, without a doubt, one of the functions that content creators have been craving since they learned to create prims. Please do this.
One caveat, however - no doubt you'll have to limit polygon/vertex counts, please try to strike a balance between technical considerations and effectiveness. If the mesh import tool limits creators due to excessively restrictive polygon limits, then it will be more of a frustration than an enhancement. Also, please consider using .OBJ import - including texture coordinates, so we can render out our texture maps and apply them efficiently, instead of breaking them up and applying them singly to groups of objects. The resulting creativity you'd see from this will be mind-blowing, I can tell you. I've got too many things on my hard drive that I've made that can't be put into Second Life easily because of import/creation constraints. Please implement mesh import! I'm certainly voting for this. Meshes will allow so much more to be done, and allow it to be done in a much more optimized fashion.
On the other hand, I'm leaning towards the opinion that I'd like to see prims continue alongside mesh modeling. New residents should always have an easy way to begin building with no previous experience. It's a large part of what makes SL as approachable as it is. Of course, if a relatively simple mesh-based in-world building tool could be managed, that would certainly be ideal, even if outside applications remained necessary for more advanced techniques. Mesh support with OBJ support.
I personally can't understand why we didn't do this in the first place. We could have even just used a method like a list of verts inside a notecard in a style like OBJ.
We should definitely go towards actual models as the sheer difficulties in expressing detail selectively in a perfectly square mesh just makes it a frustrating and long task of trial and error, hoping that we might actually succeed. I can't stress enough that we should have a variable number of verts instead of the fixed dimension. Also, please consider allowing custom bump/tangent-space normal mapping so if we can not at least have high detail meshes, we can have high detail bump mapping so as to increase that detail. I'll even make a patch exemplifying that if you want. Whatever you do, PLEASE do not use COLLADA.... the world does not need to use XML to transmit model data. Try just using OBJ or at least a format that can be directly converted from OBJ with a converter instead of just providing an in-maya tool this time... I know you like community involvement for these tools but a reference implementation would be nice. I agree with everyone else here, we need more professional ways to create more appealing and useful content. Actual editing tools are enough fair for creators looking for (not so) simple content, so I think it's a waste of resources trying to lower the bar.
Mesh import would be ideal, but bandwidth is an issue that needs consideration. People are going to try downloading a massive 10,000 triangle dragon from the net and import it, then everyone who goes into that sim is going to have to stream it. I propose limiting each imported obj size up to maybe 512 triangles, or whatever seems fair balanced with streaming.
+1 to variable mesh sizes, sometimes an object just needs 12 triangles, sometimes it needs many more. What i mean is that you would be able to upload an obj file up to 512 triangles and have that count as a prim. If mesh import doesn't happen please take a look at openCroquet for ideas on inworld mesh creation, especially extruding from a curve. Support for vector graphics would be great also. If these means turning off NURBS smoothing, I'm all for it. I'm sick and tired of my perfectly symmetrical prims getting lumpy willy-nilly just because I didn't focus enough vertices in the area, or mirroring the same prim resulting in ridiculous puckering effects. Aaaaah how I long for smooth, crisp edges! And this would finally allow that without doing a ridiculous dance of uploading lossless textures. Voted.
Yes this would be ideal, do not know why this wasn't done in the first place. I would love to be able to import meshes to add just that much more detail, realism, and professionalism to my builds. And I do agree with many of the comments here, allowing custom bump/tangent-space normal mapping would be ideal, although how to optimally be able to do that with the varying amounts of bandwidth SL users have would be a little problem. But it would be a awesome improvement in the quality of content
I'd say that's a very bad idea because people would abuse the marketing ground in SL by selling bunch of freebie models. Plus, this would cause more lag in the viewer and bandwidth. Very bad idea.
Vincent has some valid concerns there. I think he's wrong at least about causing lag in the viewer. I think the resale of freebie models is not a new issue to SL, and is separate from this issue. I don't know how bandwidth would be effected, but I'm betting not as severly as one might think.
To address some of those concerns, I would like to suggest that if you do implement model imports, part of their properties window be a count of triangles. transfer size of mesh could also be useful. Similar displays on groups of primitives would allow for comparison. You also might consider giving meshes a weighted prim count. For instance. for every X triangles, this mesh counts as another prim. So you might have very very complex meshes that count as up to 255 prims. All the rules for parcel return, linking, and being physical could then apply in proportion. I believe the argument against weighting before, has been that it might be too hard a concept to follow. However, since it applies to people who have the brains to use a modeling tool to begin with, I don't think it's unreasonable. The math is elementary (school). I think that compelling content is generally not made by people new to 3d modelling. That being said, I wish this was not presented as a "either or" ... I think there is room for both improvement of existing tools and to allow mesh import. I personally think that there is room to improve the parametric prims as well.
However, mesh import will allow me to do FAR more quality content than I can now. As my texturing is tightly integrated with my modelling pipeline, and that SL doesn't allow me to download items I have made in the viewer, I end up taking far too many extra steps just to finally texture something I've made with parametric prims, by literally recreating those prims in the modeller, even though I know I will be making it as a prim in world. Another thing I want to have looked at is animation and custom uvmaps. I'd really like to create items with distortion free uvs, and currently, I cannot do this easily with sculpties. I'd like to see bones/skinning so we can have items animate about realistically. There is far more room to run I believe if we create the framework in SL to work with external tools. Voted. OBJ files do not have animation information - I strongly recommend a format that does, such as COLLADA. and I also strongly urge a polygon limit.
I'm all for true mesh import - given that the open grid people got it working themselves, there's really little excuse for SL to not have it.
Prims still have their place though...they're there, they're accessable to most. A lot of improvements could come just by a UI overhaul of the build tools, which would fall to a different team and could be worked on without excluding other things. The once-promised "fuse prims together as a mesh" feature could also be more possible as a result of doing mesh support. Also, official support for prims >10m plz! "the world does not need to use XML to transmit model data. "
Too late, as the world does already. Your avatar shape is transmitted that way. The avatar skeleton is sitting on your drive as an XML document. COLLADA fits right in, and there is already some support for it in the viewer. My first reaction was based on my own experience with modeling and it was "What existing modeling tools?" Oh I'm sorry that really sounds rather negative doesn't it? No ability to do remedial Boolean, or lathing, or lofting. LL calls that a modeling tool? Pardon me while I giggle a bit.
It's not an either or question either. Why does LL need to have one or the other? Why not do both? Furthermore, if it can be done, why hasn't it been done? Oh, I know rather useless to ask considering the promises we've had about other issues, like Havok. Well if it has to be one I'd vote for mesh import despite my terribly negative sarcastic attitudes. PLEASE please please, .obj import! Sculpts were a step in the right direction, but have way too many glitches and picky limitations.
Aaron wrote: He said it better than I could. This is an excellent idea for how to reconcile .obj meshes with the current prim system. I can grovel if it will help...obj....pretty pretty pleeeeeease? What else can I say that hasn't been said already in other comments?
There is nothing I would like to see more in SL than this. While I have thrilled at the ability to product organic shapes, I have always felt Sculpties were too limited. gIVe uS THe MeSH IMPoRT oR ThE BuNNy GETs iT! Having been introduced to 3D content development first in Second Life and grown to understand and use tools and techniques outside of SL I am excited to hear of mesh support. I really like Aaron Edelweis' recommendations and think they would actually greatly simplify current workflows to accomplish the same thing with several sculpties. Like many have expressed, I think promoting adoption of industry tools and techniques benefits the most people in the long term and moves us closer to a real 3D internet. The separation between creative fun builds and true works of art will increase, but those most committed to SL creativity will be unfettered in their expression using the tools they know. Some may have already mentioned RealXtend's modification to OpenSim to import mesh already so really it is inevitable. Eventually some world will allow it and that world will likely win out over the others because the content quality will be that much better. For LL it is to be ahead of the curve than behind it.
One clear advantage for potential content creators is that they can create, even sell meshes completely outside of the SL economy, in the RL economy. This is also really inevitable and, for better or worse, put content ownership and distribution responsibility more on the shoulders of the creators themselves. I don't exactly know where it would fit, but we should include a discussion about COLLADA. I wonder how many of the attribution and upload issues would relate to COLLADA adoption, but I know little about it specifically. One of the greatest challenges to bringing meshes into SL will be managing level of detail (LOD). This is already an issue with sculpties and tools have been built (Blender plugins) to help plan for this. The LOD specification would be a key component to making external tools work well with SL. I would personally also like to see more REST and other interfaces supported by the libsl or other APIs that allow direct upload of assets, with charge or whatever, and the ability to directly cloth them onto one's avatar. I have been hinting about creating Photoshop macros to preview skins and clothing someday with a single click on a plugin or macro rather than going through the several step process today's workflows require. Thanks Quarl for looking ahead. It keeps us motivated to stay engaged and informed. Surprise surprise. This is far in the lead of 1495. But I'd rather we didn't have to choose between the two and we could simply have them both improved.
DEFINITELY mesh import. But please don't make us pay to upload some obscure mesh file; a dialog to paste in the contents of a .OBJ is all that's needed.
Mesh import will allow for far more accurate builds with less polygons and waaaaaay more flexibility. Oh wow!! do it BUT!!!!!! and it's a big one: SL stands for user-based content therefore keep the humble prim. So: make it so that the people who know how to use meshes can import them but make sure the non-pro user can have his/her cake and eat it too!
 MESH MESH MESH MESH MESH
in-world tools are fine, and should be improved in the general course of things. Lack of mesh support is a deal-killer on lots of potential use cases. It also kicks sand in the face of a huge community of 3D artists who have honed their skills over the years, and would be an incredible fan base if the situation were to change. Right now they are looking at Qwaq, Forterra, X3D etc., anything they can to get into the game. Linden is a non-starter with inworld-only tools. I won't repeat the other points here that I mostly agree with. An important thing to consider is that there's little question that mesh support will happen, I believe it's inevitable. There's also little question of demand. So why not start the process moving sooner rather than later.
Before I'd vote for this or its opposite number, I'd like to see a VAG (Viewpoint Advocacy Group) get together with current and former LIndens, realXtend, openviewer, libopenmv, etc developers and discuss the issues as part of a general OGP/AWG thing on graphics protocols. Avi Bar-Zeev, the guy who wrote the original primitive code way back when, posted this a while back:
http://www.realityprime.com/articles/volumes-of-reading I gotta think we can do better than what LL does right now, but doing it without raising the bar on the system requirements for the SL viewer is an important consideration also. I'm voting for the other side of this issue. One thing came to mind since my original comment over there - how are these meshes going to be handled in the physics engine?
How about a compromise? A lot of people want to see avatars improved - why not limit it to uploading a mesh for an avatar? Cap the vertex/triangles limit to that of the current avatar mesh, or maybe a bit higher, and make note that the bounding box for collisions WILL stay the same as current, so that the physics engine doesn't care. @TaraLi who said "how are thes | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Meta Issue
Open
Normal
SURVEY: vote here if you'd prefer for LL to work on mesh import over improving existing prim modeling tools
