New World Notes

Why It Will Take Awhile for Bakes on Mesh to Make a Difference in SL

Commenting on Cassie's video comparing SL mesh bodies, and her note that Bakes on Mesh doesn't seem to improve performance [Ed. Not exactly, see update below], Patchouli W provides this analysis:

BoM's true value contribution will only surface once onion-skin geometry is stripped away from BoM-optimized versions of the same body and people stop wearing more than one copy of the body geometry because of the removal of said onion-skins. The other optimization that will need to happen - and this is trickier because of limitations on how BoM works currently as well as so many legacy mesh clothing items being reliant on them - is the removal of legacy alpha cutting in favor of a BoM model that can handle 'cuts' made by using system layer alphas - this would reduce the amount of duplicated geometry required to fake smooth transitions between disparate pieces of geometry as well.

Only when these two changes are pushed through in full will the full value of BoM in cutting the impact of mesh bodies on the SL grid be fully realized. Then there's the issue of trying to port this work over to cope with normal and specular/env channels baking, which is an entirely new kettle of fish since the way normals are blended is a little different compared to diffuse and specular texture layering.

This may be true or not true, but what's striking to me is all the 3D graphics technical jargon one must know to understand the state of Second Life mesh. Or as it was expressed in the immortal line from The Limey:

Cassie Demos a Dozen Top SL Mesh Bodies & Documents Their Performance

In case you missed it, here's the massive review of top mesh bodies from master YouTuber Cassie Middle -- which is how she created this super valuable spreadsheet listing them all by avatar complexity, triangle count, and so on.

Some of the bodies she demos here are Bakes on Mesh, but Cassie tells me she saw no difference in performance for BoM bodies, as compared to traditional mesh. (Probably because creators are still figuring out how to use and optimize BoM.) However, she adds, "It's very fun to use; I can use a ton of older tattoos and skins, I can stack them on top of each other, and it still looks amazing on the bodies."

The video itself is a labor of love and an insane amount of time: 

Highly Detailed SL Mesh Bodies Don't Require Excessive Triangles - Look!

Pictured: A well-optimized body from Niramyth Productions

Some readers have complained about posts like this one last week pointing out how poorly optimized and resource heavy the leading SL mesh bodies are, basically saying, "We want a lot of avatar detail!" 

This, however, is a major misunderstanding: It's perfectly possible for a mesh body to be well-optimized and have high detail. By way of demonstration, take a look at the mesh body above, a picture sent over to me by Black Dragon developer NiranV. 

"It looks HIGHLY detailed and is super low poly with just roughly 24,000. Of course head, hands and feet add a bit more too but overall it seems to be around the 40,000 mark." By contrast, most of the popular SL mesh bodies have triangles in the mid to high six figures, and don't look any more detailed than this one.

"I mean just look at the details," NiranV goes on. What a difference good textures and a bit of normal mapping can do. That body looks right out of Unreal Engine and it doesn't use any shitty onion layers. But don't tell the Maitreya/Belleza fanboys, they will screech."

Maybe they will -- or maybe they will demand more from the owners of the Maitreya/Belleza brands.

Oddly enough, some readers (especially on Facebook) have claimed that calls for better optimized Second Life content are an attack on Second Life itself. They're really about calling for a better virtual world. As reader Tonia Thraw explains:

An Optimized Second Life Looks & Perform (Almost) As Well as a AAA Game

This is a jaw-dropping sneak preview to an upcoming roleplaying sci-fi game, and while you might assume otherwise on first glance, it's not a AAA game for consoles: It's based in Second Life. Called New Frontiers (Discord channel here). Lead developed by JDB Bing, it's expected to officially launch in a few months, but you can sneak preview it in person: Click here to teleport.  

Shot in the Black Dragon viewer at about 30 FPS (perfectly decent for an RPG), Mr. Bing tells me the sim is being optimized for high performance: "There isn't a bunch of a 1024 textures everywhere," as he puts it. "Mostly we are trying to stay away from resource-wasting items and sticking with creations from Isil/Tokeli/Skye. They use materials aka speculars/normals along with limiting their polycount and texture usage."

As such, it's a clearcut case study on the importance of optimization, and what Second Life users, creators, and Linden Lab itself are missing out on, when they refuse to put a premium on optimizing:

Bakes on Mesh Will Greatly Improve Second Life Performance -- If SLers, Viewers & Mesh Brands Adopt It

Left and right: Bakes on mesh when using the updated SL viewer... and a current 3rd party viewer (photos by Rhonda Huntress via this thread) 

Linden Lab just announced the official release of Bakes on Mesh [aka BOM], which aims to optimize mesh-based attachments. (Lack of mesh optimization has utterly crippled Second Life performance in recent years.) And while actual SL users are struggling to get it to work with their existing mesh and current viewers (see above), it does have great promise. NiranV, creator of the Black Dragon SL viewer and Second Life's sharpest critic of poor performance, is cautiously optimistic. 

"Bakes on Mesh idea is nice and it's a much needed thing, especially with onion layers," as he puts it to me. "This could potentially reduce texture memory usage by up to 75%, and polygon counts too." He's planning to merge BOM to his own viewer soon, and Firestorm, Second LIfe's most popular viewer, is already doing so.

One key challenge, however, is BOM's compatibility with existing mesh-based avatar content. It remains to be seen if the leading brands will update existing content for their customers -- or require them to buy new BOM-based mesh.

Linden Lab Should Show More Interest in Optimized SL Content... and Less Favoritism to Resource-Heavy Creation (Comment of the Week)

Massive discussion over last week's post about Linden Lab promoting Catwa, culminating in this rant by Soda Sullivan.

First, shame on you Wagner for simply dismissing in your lead paragraph the fact that Linden Lab though it was okay to put its finger on the scales of the market and promote a single creator of mesh heads over all the others.

The irony that they chose one of the baddest actors in polygon resource designs is ironic. It is basically a nice kick in the teeth for any content creator trying to make a less resource heavy product and get into the market.

For the record, I didn't "dismiss" that fact, but instead wrote "that's a very worthwhile debate for another time".  And by all means, I'm totally happy if "another time" is now!

Soda goes on to dismiss SLers who resist comparing Second Life to AAA games, which have better graphics and much better optimized mesh content -- and spares her chief ire for Linden Lab CEO Ebbe Altberg:

Linden Lab Officially Promoting Resource-Heavy Second Life Content That Drastically Hurts SL User Experience

The official Second Life Facebook page recently featured this post promoting a user-made video tutorial for the Catwa Bento Mesh head, one of the most popular and bestselling user brands in Second Life. Some SLers have called this company favoritism and free promotion for content that has competition in the virtual world, and that's a very worthwhile debate for another time.

Perhaps even more notable, however, is this question: Why is Linden Lab promoting mesh-based content that is so resource-heavy, it substantially degrades the user experience for nearly every Second Life user? As reported last year, the top-selling mesh bodies in Second Life -- Maitreya, Belleza and SLINK -- are extremely resource-heavy, and tend to degrade performance of everyone around them, including the users wearing these bodies.

They're so resource-heavy, in fact, a former Linden Lab engineer once told me that using them is effectively like launching a DDOS attack inside Second Life. And now, Linden Lab is effectively promoting those attacks.

But wait, this is just a Facebook post about the Catwa head, not a whole body -- just the head can't be that bad, can it? I asked SL graphics expert NiranV, developer of the acclaimed, graphics-optimized Black Dragon SL viewer, to take a look in the Catwa shop. This is what he found:

 

Poorly Optimized Avatars Are the Pollution-Spewing SUVs of The Virtual World - Comment of the Week

The debate over poorly optimized mesh bodies which threaten Second Life's existence as a shared, virtual world, continues -- this time, with an apt metaphor from Soda Sullivan:

The official avatar needs a complete overhaul. It has been the subject of many debates fro years and years and let's face it, the verdict it in and this avatar was built wrong. It is too tall, it is ill-proportioned, and the improved mesh avis are complicated as hell to try to get a new user comfortable with customizing... This overhaul should have been done over a decade ago and now it is crucial.

Linden Lab wanting to keep "hands off" makes perfect sense in most cases, but not here.

Think of cars. Currently our roads are full of gas-powered mammoth vehicles that are literally killing us with their exhaust. So , the government wants people in electric cars. They [encourage] manufacturers in that direction with both carrot and stick. Governor Linden should act the same way with the avatar. Make a new basic avatar, and then set down some rules of the road for anyone who wants to make their own heads, bodies, etc.

Or to put it another way: Wearing a poorly optimized mesh body in public spaces is like driving a massive SUV with a busted catalytic converter around town. Enough people do that, and everyone just stays indoors.

Unless and until Linden Lab does do anything about this, there are alternatives. Reader Lelani Carver reminds readers there is a kind of Prius Hybrid of mesh bodies on the market:

Watch: How to Improve SL Land Impact By Optimizing Mesh in Blender

Poorly optimized content continues to be a bane of Second Life, so here's a useful tutorial for optimizing SL buildings by exporting them in Blender. Presented by longtime creator "Deep Elements", who tells me this process reduces overall land impact by quite a lot:

"It's an EXTREME difference," as DE tells me. "My current build (three floor Building) would have been 271 Land Impact - now it is 94, with a functional elevator, too! All rooms/floors have realistic trimming on ceiling, floor, windows, etc. There is a lot in this build, and to be under 100 is simply marvelous."  

Deep Elements explains how his method reduces Land Impact so significantly:

Can a Multiple Texture Upload Feature Improve Second Life Performance?

Here's a new feature request on the Second Life JIRA worth following, submitted by Firestorm developer Beq Janus, who explains her reasoning for it in a New World Notes comment:

As was raised in the past in this blog, one of the problems we face in Second Life is inefficient content design, and a move that encourages the creation of high resolution (1024) textures as part of a default workflow is going to lead to a worsening of this situation. To this end I have proposed an amendment to the viewer (and server) that would allow for the simplified creation of multiple resolutions in a single transaction. Ultimately I would like to see this generate a range of resolutions from a single (high resolution) source image, for a single upload fee. I think the feature has value without the single fee change, but the take up would be far lower. The hope is that with such a feature, artists and creators can work with high-resolution images and the viewer will produce multiple resolution uploads allowing the creators to test the appearance and make better decisions about when a 1024x1024 image is really needed.

The UI mock-up pictured here is part of Beq's proposal in the JIRA, a way to subtly encourage mesh creators to avoid uploading unnecessarily resource-heavy, poorly optimized textures which are now choking SL performance: