Category: Simulator User Group

The following notes are taken from the Content Creation User Group meeting, held on  Thursday February 16th, 2017 at 1:00pm SLT at the the Hippotropolis Campfire Circle. The meeting is chaired by Vir Linden, and agenda notes, etc, are available on the Content Creation User Group wiki page.

Core Topics

• HTTP asset fetching
• Potential project: animated objects

HTTP Asset Fetching

In 2013 / 2014, the Lab made a huge change to how avatar appearance information and texture and mesh assets are delivered to users, shifting them away from UDP (User Datagram Protocol) delivery through the simulators, to HTTP via Content Delivery Networks (CDNs) – see my past reports on the HTTP updates. and CDN work.

As was indicated at several TPV Developer meetings recently (see here for an example), the Lab has been looking to move more asset types for delivery over the CDN, and this work has now started, with a focus on animations and sounds. This should see improvements in both the speed and reliability of assets, which should be particularly beneficial to animations.

The work is in the early stages, and progress will be tracked through my SL project updates.

Potential Project: Animated Objects

A topic of common conversation at various user group meetings is that of animated objects – e.g. objects which can be animated but which are not necessarily part of the base avatar mesh, and / or things like non-player characters (NPCs).

While it is still very speculative, the Lab is considering how this might be done and what sort of applications people would use such a capability for. One idea has already been extensively documented – “created agents”, which are avatars which do not necessarily have a connection to a viewer in order to operate – see feature request BUG-11368.

The main aim would be to use the same base avatar skeleton for this work, as well as it being compatible with existing rigged objects, rather than introducing something like custom skeletons (seen as a much bigger project). A lot would also depend up things like performance impact (if the simulator is operating a certain volume of NPCs or ridable objects, for example, then these could impact on resources which might otherwise be used by avatars, etc).

One potential way of achieving desired results would be to animate rigged meshes using the avatar skeleton, but without necessarily having the actual avatar base mesh underpinning it. For example, when we use a mesh body for our avatars, we use the base avatar, but hide it with an alpha mask, with the avatar skeleton animating the worn mesh. With an animated object utilising the skeleton, there is no real need to have the underpinning base avatar, as it would in theory never be seen.

One issue is that many mesh models are multiple parts, therefore some means would be required to control them, and this could be lost without the base avatar, together with the ability to attach static objects to something like an NPC. Hence the idea put forward in BUG-11368; the “created agent” would effectively be a special object class, providing the means for multiple animated meshes to operate in concert.

It is unlikely that the bone limit for a given object would be raised to accommodate animated objects, as this is pretty much a limit imposed by people’s graphics cards. During testing, the Lab found that if too many joints are defined for a single object, some graphics cards are unable to render the object correctly. This impact has actually already been seen with some Bento content (FIRE-20763).

Other aspects which would have to be considered are things like Land Impact. Avatars don’t have a land impact, but that may have to change in the case of animated, avatar-like objects – again, seen the performance concerns above. There are also some concerns over possible griefing vectors.

Performance-wise a potential benefit would be animated objects would not require alpha swapping, which requires a fairly hefty performance hit – but this could be countered to a degree (and depending on where you are and how animated objects are used) but the volume of animated objects around you.

Right now the idea is still being discussed internally at the Lab – there is no defined project. However, if you have views on things, attending the Content Creation meetings would be a good place to get them heard.

Other Items

Applying Baked Textures to Mesh Avatars

Still under consideration is a project to allow baked textures to be applied directly to mesh avatars (see here for more). This is still under consideration, but has yet to be formally adopted by the Lab as a project.

Modelling for Efficient Rendering

The subject of efficiency and LODs was the focus of an extended conversation. As I reported in my last Content Creation UG meeting report, Medhue Simoni has been producing a series on the use of Level of Detail (LOD) to help with generating rendering efficient models in Second Life. All three parts of the series are now available on his YouTube channel, and he and I will be discussing them in this blog in the very near future.

The following notes are taken from the Content Creation User Group meeting, held on  Thursday January 26th, 2017 at 1:00pm SLT at the the Hippotropolis Campfire Circle. The meeting is chaired by Vir Linden, and agenda notes, etc, are available on the Content Creation User Group wiki page.

Core Topics

• Potential follow-on projects
• Modelling for efficient rendering
• Animations
• Outfits

Follow-up projects

There has been no further progression on the potential follow-on projects from Bento (see my week #2 Content Creation Group meeting report for details of follow-ups).

• In terms of the baked texture on mesh idea, the Lab has asked for specific content there “onion” meshes are used – and it has been reiterated this is most of the common mesh bodies and heads
• There is still some confusion as to what may be causing the animation conflict issue. While it may be that changes will be made to the animation system in the future, as per the suggestions Vir forward previously (see the link above), the solution for now is to try to address the issue at a scripting level to prevent conflicts.

Modelling for Efficient Rendering

Medhue Simoni has been producing a series on the use of Level of Detail (LOD) to help with generating rendering efficient models in Second Life. Part 1 is embedded below, and Part 2 can be found here. A third part will be available soon, and hopefully, he and I will be producing a companion article in this blog once that part is available on-line.

Efficient modelling for Second Life has long been a problem within the platform, and something exacerbated by the arrival of full mesh support. Given this:

• The Lab is considering revising the rendering cost calculations “at some point” to encourage people to consider more efficient models (e.g. making more use of normal maps to add detail to models where appropriate, rather than rely on always producing an extremely high poly count model).
• It has been suggested that providing some basic indicators of what might be considered “reasonable” number – poly counts, proportionate scaling of LODs, etc. – for models could be produced, together with videos (by content creators with a solid understanding of the subject and Second Life)  on efficient use of normal and specular maps

The discussion broadened to cover awareness among content creators as to what actually works and where falsehoods / misunderstanding lie. A cited example in the meeting was that of mesh clothing makers avoiding the use of normal maps because they want their clothing to look the “same” to everyone, even though doing so can severely impact the user experience for those on lower-end system, and thus discourage users from buying their goods.

.BVH Animations and Animation Playback

Animations can be uploaded to Second Life in one of two formats, .ANIM and .BVH. The latter are optimised as a part of the upload process, and this is proving to be a particular problem for Maya users when animating facial expressions. These require finer bone movements in the animation, which the optimisation process is filtering out, requiring Maya users to use exaggerated movements. Blender users can avoid the issue by using .ANIM, which does not pass through any form of optimisation / filtering.

While it is recognised that the thresholds used by the .BVH optimisation / filtering process may not be the best for very small bone movements, there are currently no plans to alter / refine the .BVH uploader. Nor is it really feasible to adjust the thresholds for hand and face bones, as this could have an adverse effect where these bones are re-purposed for other uses (as Bento is intended to allow).

There are two possible workarounds which may help with these issues for anyone using the .BVH format:

• Swap to using .ANIM files, which do not go through any similar optimisation process. Unfortunately, this isn’t an option for Maya users, as there is current no .ANIM exporter for Maya, although Aura Linden is working on one in her own time, and is hoping to get time in about three weeks to sit down and finish it
• Alter the frame rate of the animation itself – so rather than creating it at 30 fps, try 15 or 10 fps, depending on the animation.

There was also some confusion over the maximum file size for animations, as per my 2016 week #25 report, this was increased from 120 Kb to 250 Kb in June 2016. The wiki page on SL limits has now been correctly updated to reflect this. It’s also worth noting, as an aside, that animations will be moving to delivery via the CDN network in the future.

There was an extended conversation around Outfits and the Outfits folder. While much of this revolved around the Visual Outfits Browser option for Outfits, a couple of significant items were discussed.

The first was on the subject of saving gestures with outfits. As noted in my Bento update #26 and Bento update #27, gestures cannot be automatically saved with an outfit, but can be manually added as links / copies. However, Medhue Simoni has found a catch: should the outfit ever be re-saved, the gestures are removed. Expect a JIRA soon

A common request for the Outfit system to allow nested folders once more the ability was removed with viewer 2.1 (see VWR-19774), and while it was at the time noted as a possible “priority” item for consideration by the Lab, the topic has only recently again come up in internal conversations as a result of feature request BUG-41826. However, the amount of work involved to make it happen makes it unclear if the request will be accepted.

One request for Outfits which is unlikely to be acted upon, is to have links to other folders in addition to objects. This is seen as even more complex than allowing nested folders within the Outfits hierarchy.

Next Meeting

The Next Content Creation User Group meeting will be on Thursday, February 16th, 2017.