However Sansar Handles "Sub-Optimal" User-Generated Content, VR Worlds With UGC Face Massive Challenge

Last week's post on Ebbe Altberg's presentation about Sansar provoked some backlash from Linden Lab, who disputed my interpretation that the CEO was suggesting his company would "filter" sub-optimal user-generated content created for Sansar. I've made a correction/update in good faith, but for the record, here's exactly what Ebbe says at about minute 51 of his talk (video below the break) when speaking about a key challenge for launching Sansar:

"One challenge we have is that we don't have full control over the content being created. Users don't necessarily know how to create extremely optimized content that a gaming studio would. So we have to understand how to deal with sub-optimal content, and find ways to automatically optimize content."

One could argue that the very act of identifying and distinguishing "sub-optimal" content versus optimal content by definition involves some variety of filtering, but let that go. Because however Linden Lab decides to deal with sub-optimal content, it brings up a larger, thornier problem. I'm not a 3D graphics expert, but the challenge as I see it basically breaks down like this:

Unless you're creating a user-generated world with procedural/building block-type content creation tools, ala Minecraft or Second Life's prim-based system, where the primitive nature of the graphics is part of the world's charm, you're going to want to build that world with high-resolution 3D models. Which basically gives you three options, when it comes to user-generated content:

• Pre-fabricated building blocks and templates with which users can "snap together" parts of already optimized 3D content, ala Sims 3.
• A consumer-friendly, "dumbed down" 3D modeling system that creates optimal content out of the box -- a platform to be either created or purchased by Linden Lab.
• Alternately, require users to create their own content in industry standard 3D modeling platforms -- as difficult as they are for 99% of average users to master.

If I had to guess, option one is the most preferable one, in that it would encourage the most users to create the most content, with the lowest barrier to entry. Trouble is, while template-ized platforms like Sims 3 are really great for a mass market audience, they can also be pretty limiting.

In any case, we still don't know what solution Linden Lab has in mind -- and to judge by this NWN comment from Linden spokesman Peter Gray, Linden Lab is still trying to figure it out themselves:

He was commenting on the challenges of providing a platform for user-created virtual experiences, which go beyond what pro studios creating VR content face. Whereas those pros may have a great deal of experience in creating optimized content, our platform needs to also empower non-pros to create high-quality and highly performant virtual experiences across a wide variety of use cases. That could involve providing tools to help users create experiences that are optimized for performance with HMDs, for example. It's still early in the development of our new platform, and how it will handle this particular challenge is a topic we'll address in the future...

And to be fair, I can't blame him for saying this -- if virtual reality-based worlds are really going to be mass market and also enable mass user-generated content, this is probably the biggest hurdle to overcome.

Please share this post:

Tweet

 

Comments

[Cube Republic]

They should just drop this land impact nonsense and charge by the triangle, and give a region or whatever they will have a triangle limit. You can bet content creators will soon get a handle on optimizing their content if this is the case. It's exactly what cloud party did, so it can be done. I feel a hybrid prim come mesh generator in-world gizmo would be great for novice users. Also technology is a million light years now from what we had when SL first opened, so really the sky is the limit as far as what can be done. As far as I'm concerned they just need to take the charm and magic of second life and just upgrade it.

Personally I don't really think a virtual world is a virtual world unless it's completely continuous ( like the world we live in).

[Nalates]

Since content is such a basic part of a virtual world and alpha-beta testing is only months away, I find it hard to believe they have not decided how to handle user created content. They may not have FINALIZED their thinking and plans, but content in a VW shapes how everything in the system works.

I am pretty sure a Blue Mars approach limiting who can make content and be approved before importing to the live grid is out.

The automatic LOD and mesh simplification that SL mesh import uses is pretty bad. So, automated optimizing is probably not a workable solution.

I am guessing we will have something like SL with primitives and mesh model import. In this area the 'better' Ebbe talks about is likely to mean something more like Cloud Party where the in world building tools were better.

We are hoping to see SL get better Render Cost feedback. Oz says they are in the works. The server side is already in place. Presumably we will see similar feedback in SANSAR. So, the optimizing may be more a matter of social engineering than programming.

Also, avatar attachments in SL have no Land Impact cost. So, people use ridiculous numbers of polygons when making mesh clothes. I expect there to be some change in that respect in SANSAR. Economic engineering is usually very effective.

Whatever they are thinking about sub-optimal content and however they handle it, I believe they see including it as a basic system requirement.

[irihapeti]

the first two you mention kinda go together

a simpilfied inworld building tool (similar to prims) which can copy bake a construction into a single object. And a fairly comprehensive library of starter component blocks

i think the simplist way to control this is to put a LI on everything, including textures, animations, sounds, particles, lights, etc

the LI of a baked construction should be at least under half of a non-baked made out of components. Just to encourage people to bake

[zz bottom]

Funny to hear about cloud party and Blue mars as we know how those ended.
In world tools to build is the option to go.
Second Life, despite all, is still an amazing tool to build and much better user friendly then any other.
Part of it's success is just that.
Remove that and Sansar will be doomed as an alternative.

[Masami Kuramoto]

@ zz bottom

Apparently you don't know how Cloud Party ended.

[Ezra]

Cloud Party had this issue sorted, and Lindens like Philip and Kelly came around to experience it. If Linden Lab was smart, they'd use it.

The system was basically:

1. Have a real material system where normal maps were first class and always on for everyone (as opposed to SL where anyone can, and a lot of people do turn off Basic Shader), and then creators wouldn't be compelled to use vertices for small details.

2. Give the default and any custom avatar skeletons a resource budget, analog to regions in Second Life having a resource budget in the form of Land Impact.

With Cloud Party, creators of skeletons could create "attachment points" and budget out how many triangles could occupy the point.

Combined with similar budgeting for the island/sim itself, and control of how many avatars could be in the island/sim, there was a hard cap on how many triangles would ever be in the scene, not unlike a single player game.

A proper material and rendering system along with resource capping at avatar and sim level is all that's necessary.

[Metacam Oh]

Let users host their own land/sims and this wouldnt be an issue, let people assign their own hardware resources etc

[Iggy]

Educators want ordinary student to build collaboratively and in-world. That would be ideal with building-blocks and prefab, easily modded scripts.

I'll leave "industry standard" tools for the 1% who have time and professional incentives to learn them. The best sort of world for me would permit both sorts of building. I could support more talented builders by buying stuff my students and I can't / don't want to make.

[Snookums San]

I've only built a little in SL over the years, and mainly been a user of content.

Layers definitely need to gone, so everything pure Mesh.

One thing i have found having just upgraded to 100% Slink body, feet, hands, head, because almost all mesh clothing is made for non mesh bodies the work tends to be very sloppy, a lot is not rigged correctly, as the creators know they can just wipe out the default body with an alpha to hide the bad work.

So there does have to be a very strict standard in SL2, then the user not paying good money after bad, and on the plus side, if content creators are forced to create to a very high standard that is guaranteed to fit correctly then they will have much more time to create as they won't need to spend so much time dealing with customer support.

I do agree that a lot of the software for 3d creation is really hard to learn if it not something your used to using, I know I have tried a few times to get my head around blender and other 3d software and still can't figure it out, hence I am more a user than a creator as of yet.

But I feel if there was a strict standard so everyone knew to get content in to the world it had to be say a minimum specification, in long run I think it be easier to learn as it would be like a preset to work to.

Just my opinion on the subject as a user for what it worth.

[Troy McConaghy]

There are (software) tools out there which can automatically optimize 3D meshes while retaining visual detail. Game developers already use them. Linden Lab could build one into the content import / upload tool.

[Shockwave Yareach]

It doesn't really matter. If they don't make a unified grid where they land I bought is still there in the new version, then it will not succeed. Just like other more capable and less expensive VWs have not succeeded.

As for quality issues, there should simply be a license to sell in marketplace. Before you can have your stuff there, you have to prove you know how to build properly by showing Quality Linden your stuff. But since equally simple solutions to stolen ip exists (judge linden) and griefing exists (cop linden), and aren't in place because the lab wishes to disengage with its customers, don't hold breath that Quality Linden will be logging in anytime soon.

With the lab's refusal to provide any service and with history showing people refuse to leave SL because of their financial investment, their talk about a new world with linden guarantee of quality is laughable. If we can't take the stuff we own - even the shoddy stuff - we won't go.

[Prometheus Creations]

They should just allow collada uploads and set limits on triangles/polygons on uploaded meshes, depending on their size and LOD. A complete PBR DX11 material system allows for great detail with much lower polygon usage by adaptive tessellation, and we are soon entering DX12 which will show extreme performance boosts (~200% better performance)in all supporting graphics cards.

I believe the current LI system is a good way to "charge" for area usage, however a limited polygon budget on avatar attachments is necessary. I have seen simple clothes in excess of 120,000 polygons, the budget a game would use for about 20 NPC characters appearing on the screen at the same time.