Tag: Project Bento

The following notes and audio were taken from the weekly Bento User Group meeting, held on Thursday, August 25th at 13:00 SLT at the the Hippotropolis Campfire Circle. For details on the meeting agenda, please refer to the Bento User Group wiki page.

Note that this update is not intended to offer a full transcript of the meeting, nor does it present the discussion points in chronological order. Rather, it represents the core points of discussion, grouped together by subject matter were relevant / possible, whilst maintaining the overall context of the meeting. My apologies for the background birdsong. I usually turn off local sounds for these recordings – but in this instance I forgot!

Bento Viewer

The Bento project viewer updated on Thursday, August 25th to version 5.0.0.318969. This release includes the following changes / improvements, but please refer to the release notes for the full set of updates.

SLM Files

Following previous discussions (see my Bento week 23 update), MeshImportUseSLM has been set to FALSE in the viewer, so that SLM files are by default not created and not used. Switching MeshImportUseSLM to TRUE via debug settings will re-enable them once more.

Avatar Vertical Position (height above ground) Calculation

See here for further background noted. Essentially, any position or scale changes for joints that are included in the body size calculation (those up through the left leg from the foot, the pelvis, torso, chest, neck, head and skull. can trigger issues with an avatar bobbing up-and-down unexpectedly as a result of the frequency of the calculation being increased to once per frame. To help compensate, this new version of the viewer only updates the body size calculation when starting / stopping an animation, so hopefully the changes will be less visually jarring.

Overall, however, the best fix is still to avoid using animations that change the positions of the joints used to calculate the vertical position of the avatar.

Two other updates worthy of individual note are:

• The show bones display has been modified to use colours differently, distinguishing between joints that are skinned to, joints that have position overrides defined, and all other joints.
• Animation of collision volumes has been fixed.

Move To Release Candidate Status

Currently, the Lab believes there are no known active bugs which are sufficient enough to prevent the Bento viewer moving to Release Candidate status, potentially in the next few weeks. With the core AvaStar issues also seemingly resolved (see below),  creators are being encouraged to  complete any additional testing they wish to make so that any remaining potential blockers are found before the viewer goes to a wider audience. Move the viewer to RC status will expose it to a wider cross-section of users, allowing the Lab to get a better feel for the viewer’s overall stability or whether any non-Bento regressions have slipped into the code, etc.

Mesh Distortions with Altitude

Vir has looked further into the issue of facial distortions with altitude when software skinning is used (some meshes can also be distorted when right-clicking on an avatar). The problem seems to be in the base avatar mesh itself, although one idea he had for the cause of the problem didn’t pan out.

As this particular problem can be overcome by using hardware skinning, and seems to be a subset of a broader issue affecting worn meshes (floating point calculation error), it may not prevent the Bento viewer going to RC, but Vir is going to continue to look at it.

AvaStar Issues

Matrice Lavalle of the AvaStar team believes he’s resolved most of the issues affecting their software (see here for some background), and his testing indicates that bones should now be correctly positioned and adjusted when using sliders, although there are some minor nips and tucks he still wants to take care of. He also noted an issue affecting the torso bone, which came up shortly before the meeting and which may be limited to a particular way in which the torso is modified, and he is looking into that.

Maya / MayaStar Notes

Cathy Foil provided background information on some of the lessons she’s most recently learned in using Maya for avatar models, relating to joint orientation requirements, performing freeze transformations, etc.

Cathy also provided some advice to Maya users on ensuring that facial bones which have a custom position but which are not rigged to, are included in the skin cluster, otherwise the bone will be excluded from the .DAE file, leading to issues with the model in Second Life.

Matrice noted that as a result of Cathy’s discovery, AvaStar has been updated to ensure custom bones without weights are automatically exported.

When she has time, Cathy plans to produce some video tutorials on the lessons she’s learned through the Bento process  to help Maya users avoid issues like these. Currently, she is still working on updating MayaStar to work with all of the new bones, which is currently a manually intensive task.

Bone Position Information and Slider Values

Cathy points to an issue with the neck bone slider, which, rather than having a neutral position of 50 (as with most bones), has a neutral position value of 66.6667, which can lead to neck alignment problems when adjusting the neck bone position. She asked if in the future, the Lab could update the default avatar shape to ensure all slider positions as close to neutral (50) as possible. This initiated a discussion on the feasibility of doing so, and some of the broader issues slider position value calculations.

Matrice also provided an overview of how AvaStar now handles neutral positions and the default avatar, before he and Vir discuss using other tools and the avatar BLEND file.

Following the meeting proper, Cathy and Matrice further discussed the differences between Blender and Maya in handing bone positions / rigging on export through the use of weight maps (Blender) and skin clusters (Maya).

Other Items

Bento Collaboration and Future Projects

Troy reported that overall, the Lab like the outcome of the Bento project, with the level of collaboration and cooperation between content creators and the Lab, particularly vis the user group meetings etc. While not the first time the Lab has worked with SL users on a project, Troy indicated that the approach taken with Bento is something the Lab would like to continue, where appropriate. Vir seconded this view, noting how the collaboration has made Bento a far more usable product.

Bento Content Documentation

An interesting point of discussion with Bento is the need for content creators to consider documentation carefully given the way the bones can be re-purposed – or at least define a common means of annotating how they have used various bones  – in order to help avoid potential conflicts which might arise on those occasions where a user might attempt to use two mesh elements, each of which calls upon the same bone, but require it to be positioned differently to one another.

Updated: animation of collision volumes is being added – section updated section below.

The following notes and audio were taken from the weekly Bento User Group meeting, held on Thursday, August 11th at 13:00 SLT at the the Hippotropolis Campfire Circle. For details on the meeting agenda, please refer to the Bento User Group wiki page.

Note that this update is not intended to offer a full transcript of the meeting, nor does it present the discussion points in chronological order. Rather, it represents the core points of discussion, grouped together by subject matter were relevant / possible, whilst maintaining the overall context of the meeting.

Time Frame for Promotion to Release Candidate Status

Some at the Lab are keen to see Bento move forward and reach a larger audience of content creators (although there is currently no reason why creators cannot get involved now and attend meetings, help with feedback, etc). Inevitably, this means promoting the Bento viewer to release candidate status, and doing so some time in the next two weeks has been mooted.

Given the number of issues currently being investigated, not all of which are viewer-specific (see below), but for which fixes may yet have an impact on the viewer, this would seem to be a somewhat ambitious time frame.

Viewer Issues

Work at the Lab is focusing on a batch of issues for which bug reports have been raised, including:

• Several relating to joint positions:
  • BUG-37546 [BENTO] facial bone joint offset issues – which has been the subject of recent Bento updates in this blog
  • BUG-37591 [BENTO] New import of a non-human rigged mesh is deformed
  • BUG-37633 Bento – Custom joint positions and Reset Skeleton not matching
• Avatar height calculation issues:
  • BUG-20013 [Bento] When wearing a Bento avatar, observers see your hover height lower than you see it yourself
  • BUG-20169 [ BENTO ] Z-Offset not applying correctly & camera tied to Z-offset
• Eye scaling problems (referenced via SL-433, an LL internal bug report)

The eye scaling issue, the result of the scaling within the viewer for mesh eyes being different to the expected system eye size, has been corrected, and will be one of the fixes released with the next viewer update, although existing models may need tweaking as a result a test carried out during the meeting using an early build of the next project viewer iteration suggested that there are still issues, and Matrice believes part of the problem might be within AvaStar.

Facial Bones Issues / Distortions

At least some of the facial bones issues appear to lie with AvaStar exports, which Matrice and Gaia are trying to rectify. Gaia offers an explanation for the problem in the Bento forum thread. Matrice further notes that there is a discrepancy between the restpose and the SL default pose; in Blender there is no easy way to define the absolute SL restpose in order to add bind pose information on top of it, leading to the wrong information (or even a lack of information) being exported to the Collada files.

Matrice is working on improving things in AvaStar, but the work is taking time and effort. Gaia has offered a detailed explanation of joint preparation on the Machinimatrix website as a workaround as they continue to redefine and improve their workflow, and offers a brief summary of the approach through the forum thread. The workaround isn’t seen as optimal at the moment, but the hope is that once the work on AvatStar has been completed, the workflow should be more straightforward and transparent.

Some questions were asked on whether it was an AvaStar issue rather than something inherent in the viewer, prompting Vir to comment that given the viewer handles joint positions in a model and joint positions in an animation pretty similarly, it’s hard to see how the viewer could be at fault, but acknowledged the only way to be sure of this is test once AvaStar have reasonable confidence they have corrected the issues they have identified.

Later in the meeting, Matrice offered a further explanation of the Belender / AvaStar issue,  also pointing to the explanatory document on joint preparation, again noting that things are very much a work-in-progress on trying to fix things.

During the meeting, Medhue Simoni tried the recommended approach given in the documentation and found it did resolve all of his facial deformation issues, other than with his eyes, which as noted above, may require further viewer-side tweaks and additional work within AvaStar, as Matrice commented.

Avatar Height Position Issues

I’ve covered this at length in past updates (see update #21 and update #19 for more). The Lab hasn’t progressed any further with a fix, but may look to alter the frequency of updates to see if this helps.

The problem is viewed as being rooted in animating one of the bones used in the calculations for determining the avatar’s height relative to the ground (left leg, up through the pelvis and head), so the solution may ultimately be down to alternative means of animating movement without touching something like the pelvis bone – something which may not be possible when trying to replica certain types of natural movement.

Rigs Modifying Collision Volumes

There have been recent reports of people encountering problems with Bento rigs which modify collision volumes. As Vir explains in the meeting, it’s not something that the Lab planned on supporting, nor did it come up during the original Bento development meetings when rigging to attachment points was discussed, so no effort has been put into making sure it works at all. Vir is now looking at the issue, but whether the Lab will support it is questionable at this point in time; however, the focus at the moment is to more fully understand the impact.

Update, August 12th: Whirly Fizzle pointed me to a viewer repository containing an update to support the animation of collision volumes, to avoid breaking existing content.

Other Items

Slider / Bone Spreadsheet

Aki Shichiroji has been putting together a spreadsheet of all the slider settings and how they affect the Bento avatar bones and geometry. It is very much a work in progress, given things are still in a state of flux, and Aki notes she is aware that there is missing information (she has yet to run through the avatar.LAD file and the avatar skeleton file). However. she has made the spreadsheet available through Google Docs, and welcomes constructive feedback from content creators (which should be addressed to her, I’m not sure if Aki reads this blog!).

Maya / MyaStar And Custom Joint Positions

Following the discussion of Blender / AvaStar and custom joints, Cathy Foil provided some additional information to help Maya users with MayaStar.

The following notes and audio were taken from the weekly Bento User Group meeting, held on Thursday, August 4th at 13:30 SLT at the the Hippotropolis Campfire Circle. For details on the meeting agenda, please refer to the Bento User Group wiki page.

Note that this update is not intended to offer a full transcript of the meeting, nor does it present the discussion points in chronological order. Rather, it represents the core points of discussion, grouped together by subject matter were relevant / possible, whilst maintaining the overall context of the meeting.

Lab Summit Feedback

Troy and Vir attended an internal Second Life summit at the Lab, where Troy gave a further presentation on Bento to the rest of the Lab’s SL staff and senior management, all of whom where very excited about the project. During the Bento meeting, Troy again offered thanks to all who have contributed to the project and who are endeavouring to make Bento a success.

Updates and Bugs

As Vir has been on vacation, there have been no further updates to the viewer since the release over version 5.0.0.317597, which the Lab hopes marks the end of updates to the enhanced skeleton and the revised appearance sliders (although obviously, the viewer will continue through project/RC releases to a final release down the road). However, a number of bugs have been reported against this update, with discussion on the forum thread on whether the problems lay within the Bento updates or elsewhere.

One of these, BUG-37546, relates to facial bone joint offset issues, which I’ve reported upon on previous Bento updates, and which can cause the collapse /deformation of any face using custom offsets as soon as it is worn, or in the view of anyone else seeing it for the first time, with both issues requiring the wearer to play an animation against the offsets in the facial bones will witness an immediate collapse / deformation of the face until such time as the wearer runs the animation against their face.

Teager has been experimenting with the issue at length in an attempt to try to isolate possible causes. In doing so she has identified that this deformation can occur whether the mesh is weighted or not, and it doesn’t seem to be related to the facial appearance sliders or any conflicts caused by adjusting them.

Gaia Clary has also been looking into the problem, and may have come across a partial cause of why things like the lower jaw are getting “pulled back” into the face when seen in the viewer. As she explains in the Bento forum thread subsequent to the meeting:

The exported Collada files do not contain information about where the bone ends are.
In the SL Viewer the joints and the corresponding bone ends are just shifted
to the offset location. So for bones which have no connected children the “bone” rotation
never changes in the SL viewer.

But the most important news is: it does not matter how the end point bones are “rotated”.
And this is because apparently all scaling happens relative to the Avatar’s coordinate system.

There will doubtless be further follow-up on this at the next meeting.

Height Offset Issues

There was further discussion on the height offset issues which are being encountered  – see BUG-20013 and more recently BUG-20169 for details). This again came up for discussion, particularly the issue Aki has encountered with animations no longer functioning as expressed, as in the case of her wyvern’s hind legs (rigged using the new hind leg bones) no longer flexing correctly.

Vir notes that there are a number of potential causes, as previously discussed in these updates. One suggestion to try to improve things is to dial down the frequency with which the height calculation is performed (it is now being carried out more-or-less continuously), but there’s no absolute guarantee this would eliminate issues.

Part of the problem could be in the manner in which an avatar’s position relative to the ground is calculated (using a chain of bones up the left leg and then through the pelvis to the head). This has seen issues even before Bento, which has perhaps exacerbated the problem as a result of it more directly supporting quadruped / non-bipedal avatars. The problem here is that any changes to how an avatar’s height position is calculated could do more harm than good for existing avatars.

A further exacerbation with the issue is that “old” methods of fixing issues (removing / wearing a shape / avatar, etc.), no longer appear to be consistent or persistent, and a reliance on the Avatar Hover Height also isn’t always consistent (e.g. wear a shape / shoe base, etc, set hover height, then replace items and re-wear without altering Hover Height, and avatar’s height appears odd again).

Vir, Cathy Foil, Teager, and Medhue Simoni further discuss the avatar height position issue

Other Items

Updating Skeletons in AvaStar

There was some discussion / clarification on updating the skeleton in AvaStar between version (e.g. AvaStar 2 Alpha 4 to AvaStar 2 Alpha 5), and some of the problems noted with it.

Medhue Simoni, Matrice Laville and Aki Shichiroji discuss updating the skeleton in AvaStar.

Shape Update Issues

Some people have seen problems with the current (5.0.0.317597) version of the Bento viewer whereby when updating their shape others on the Bento viewer are not seeing their shape update correctly until they relog. This appears to be a baking issue, possibly the result of the Bento viewer currently not having the most recent baking service updates.

Bento Test Models

It appears the models on the Bento test page and referenced in the forum thread are not complete; then will hopefully be updated in the near future (check the dates against the models on the test page). The new models, once available should include support for all the Bento bones and (hopefully) fitted mesh support as well.

Related Links

• Bento forum thread
• SL wiki Bento test page