2022 Puppetry Project weeks #36 and #38 summary

Puppetry demonstration via Linden Lab – see below.  Demos video with the LL comment “We have some basic things working with a webcam and Second Life but there’s more to do before it’s as animated as we want.”

The following notes have been taken from chat logs and audio recordings of the September 8th and September 22nd Puppetry Project meetings held at the Castelet Puppetry Theatre on Aditi. These meetings are:

  • Generally held on alternate weeks to the Content Creation User Group (CCUG), on same day / time (Thursdays at 13:00 SLT).
  • A mixed Voice / text chat format – attendees are not obligated to use voice when asking questions, but will need to listen to voice to hear the entire meeting.

Notes in these summaries are not intended to be a full transcript of every meeting.

Project Summary

  • Previously referred to as “avatar expressiveness”,
  • Puppetry is intended to provide a means by which avatars can mimic physical world actions by their owners (e.g. head, hand, arm movements) through tools such as a webcam and using technologies like inverse kinematics (IK) and the  LLSD Event API Plug-in (LEAP) system.
    • Note that facial expressions and finger movements are not currently enabled.
    • Most movement is in the 2D plain (e.g., hand movements from side-to-side but not forward / back), due to limitations with things like depth of field tracking through a webcam, which has yet to be addressed.
  • The back-end support for the capability is only available on Aditi (the Beta grid) and within the following regions: Bunraku, Marionette, and Castelet.
  • Puppetry requires the use of a dedicated viewer, the Project Puppetry viewer, available through the official Second Life Alternate Viewers page.
  • No other special needs beyond the project viewer are required to “see” Puppetry animations. However, to use the capability to animate your own avatar and broadcast the results, requires additional work – refer to the links below.
  • There is now a Puppetry Discord channel – those wishing to join it should contact members of LL’s puppetry team, e.g. Aura Linden, Simon Linden, Rider Linden, Leviathan Linden (not a full list of names at this time – my apologies to those involved whom I have missed).

Further Information

Bugs, Feature Requests and Code Submissions

  • For those experimenting with Puppetry, Jiras (bug reports / fixes or feature requests) should be filed with “[Puppetry]” at the start of the Jira title.
  • Those wishing to submit code (plug-ins or other) or who wish to offer a specific feature that might be used with Puppetry should:

Summary of September 8th Meeting

Note: timing issues on my part meant I was unable to attend the first third of this meeting.

  • It is acknowledged that the current Puppetry viewer (viewer branch DRTVWR-558) is somewhat crashy and subject to some looping issues.
  • One aspect of Puppetry that should be highlighted is the ability for it to work alongside / in concert with existing SL animations – so you can be running a dance animation and still wave to a friend using puppeteering without the two animations clashing.
  • It is acknowledged that to ensure some reasonable smoothness of movement and to prevent things like movement conflicts between joints, there will need to be a more formalised animation constraints system. The current plan is to make this configurable via XML.
  • It is also acknowledged that tracking in general needs to be tightened within the plug-in code.
  • Puppetry does not currently interact with the Havok physics system (puppetry is largely viewer-side; physics – with the exception of some special use sub-libraries – is largely simulator-side).
  • The protocols which are used server-side to support Puppetry are not set in stone at this point; cases which require additional messaging, etc. can be discussed with the Puppetry team members from the simulator / server side of LL (e.g. Rider and Simon Linden).
  • Direct avatar interactions (e.g. shaking / holding hands, swinging a tennis racket to strike a ball, etc.): the IK system could help enable this, but it would also require a lot more work on the avatar / world mapping system to be fully possible, and this work has yet to be tackled (if it is to be tackled as a part of this initial Puppetry work).
  • The project is, at this point, fairly open as to where it might go: these initial project meetings are geared towards developers who may be interested in contributing and pushing elements of the project forward (e.g. support for full body tracking, etc.). Obviously, at some point, constraints will be placed on what is to be initially delivered.

Plugins (Pros and Cons)

  • Requests were made for the Puppetry system to support OpenXR (as well as LEAP). It was indicated that OpenXR would be considered as a default if a suitable plug-in were to be developed and contributed to Linden Lab for proper vetting and formal inclusion in the viewer.
  • The fact that the Puppetry project is using plug-ins raised concerns over system security. Plug-ins are executable, and so if accepted to run, a malicious plug-in could do considerable harm to a person’s system.
    • LL is aware of this, and is actively trying to minimise risk as far as possible.
    • However, safety also lay with users – do not download viewers from unofficial sites / sites that cannot be trusted; do not accept and run plug-ins that are passed around through forums, etc.
  • The benefits of using plug-ins was summarised as:
    • Speed of internal development / testing: there is no need to run a complete viewer build process simply because a couple on lines of code have been changed in testing; only the plug-in needs to be updated.
    • Extensibility: plugs-ins allow for more flexible support of additional creation tools or to add support for additional data formats (e.g. as with OpenXR) / hardware / programming languages (e.g. Python, C++, etc.).
    • Performance: using plug-ins allows the required additional processing such as webcam capture, processing and translation to be handed-off the separate processing threads within a computer from the viewer, thus preventing the latter losing performance by having to do the processing itself.
    • User assurance: removing things like the webcam controls to a plug-in that is not run by default as a de facto part of the viewer’s processing will (hopefully) remove fears about webcams somehow being used to “spy” on users.

Summary of September 22nd Meeting

  • It is hoped an updated version of the Puppetry Project Viewer will be available via the Alternate Viewers page in week #39 (commencing Monday, September 26th). This includes fixes and updates to the motion logic that should make avatar motion more predictable.
  • In terms of device support for puppeteering, any device that can be recognised as a joystick should be supportable within the Puppetry viewer (utilising the existing Joystick support options through Preferences) – although some refinement to the controls may be required via LL.
  • LSL support for puppeteering: nothing has been defined at present, but there are some ideas as to what might be needed / nice to have. It has been suggested LSL support is a subject for discussion at the next meeting.
  • Simon Linden has pushed a couple of capabilities:
    • A simple poser contained in a side branch of the LEAP repository. This reads a basic JSON file with bone positions (rotations)  for all 133 bones in the avatar skeleton and sends it as LEAP data to the viewer for animating the avatar. Thisfile can be live-edited, and is desgined to help those working with puppeteering  to experiment with it in an easy format – it will not be an end feature for the project.
    • Added a further branch to the Puppetry viewer repository called DRTVWR-558 Data Packing. This converts the data going from the viewer to the server onwards to a more efficient format, allowing the full animation data set to be contained in  a single packet for transmission.
      • However, this format is incompatible with the existing data format used within viewers built via DRTVWR-558; so as viewers are built using the newer code, this will not be able to show puppeteering using the older format, and vice-versa.
      • Those involved in experimenting with Puppetry should therefore switch to the viewer using the updated data format, once this is made available through the Alternate Viewer page, as it will be replacing the current data format going forward.
  • Leviathan Linden has suggested that if LL can transmit all bone data in compressed format, then they may not need to send IK targets and have the viewer manage the IK for all avatars in a scene, but rather have the viewer run the IK for a user’s avatar and then stream the avatar’s entire state, reducing the load on the viewer.

Pelvis Movement / Full Body Tracking / OpenXR Support

  • There was initial discussion about supporting local joint offsets and particularly off-setting the avatar pelvis to allow for subtle movements without actually moving the avatar.
    • This is somewhat similar to scripted animations, such as stands with an AO system – the avatar appears to step forward / back / walk in circle, but it is not physically moving as far as the simulator is concerned – the motions are the result of the avatar pelvis being offset from it’s actual position as seen by the simulator, and the animations running based on that offset.
    • There was some initial confusion over this and physically moving the avatar, as such, it was suggested this be referred to as “pelvis movement, rather than “offsetting joints / bones”.
  • Part of the reason for this discussion is because several non-Linden developers have been experimenting with partial and full-body tracking via OpenXR, and have found that not being able to move the pelvis within Puppetry can lead to issues of floating, etc., when an avatar kneels or crouches (as seen within existing SL animations) – the result of the legs being pulled up towards the pelvis, rather than the pelvis being moved towards the ground.
  • In addition this work has noted:
    • If Second Life were to return the “full” appearance data for an avatar (i.e. after allmesh transforms, slider data,, baked appearance information, etc.) has been applied, rather than the “raw” skeletal appearance, better calculations could be made around the pelvis height from the floor.
    • The approach works equally well with partially body tracking via a Rift S headset, and fully body tracking using alve headsets and Kinect devices.
    • However, it currently uses Blender as a conduit for translating movement within an OpenXR rig to the Second Life puppeteering rig, and would benefit enormously from a dedicated OpenXR plug-in, and the developers are willing to provide data data gathered from the work they’ve thus far completed to help facilitate this.
    • Separately to this, OPEN-363 “[Puppetry] [LEAP]: Add native OpenXR plugin” has been raised, but is (at the time of writing) awaiting review.
  • The above formed a nucleus of the discussion for much of the meeting with the ability to move the avatar pelvis now being seen as more of a priority requirement, with Leviathan Linden indicating they will try to look specifically at this between now and the next meeting.

Date of Next Meeting