Pre-submission checklist
What should be possible that currently is not
An avatar should be able to animate its whole body turning over — a biped folding down into a car, a character dropping onto all fours — and have it play in VR.
Right now the pose is correct at the start and correct at the end, and the turn between them happens in a single frame. On desktop it plays properly; in VR the body stays upright for the whole sequence and snaps at the finish.
To be clear about what I am NOT asking for: nothing that moves the player. Position stays entirely yours. This is about the orientation of the rendered body during an authored animation the wearer triggered themselves.
What I tried, measured on a real avatar worn in game:
Root motion. Left as root rotation, it does nothing — the character controller keeps the capsule upright and root motion is discarded. Correct behaviour as far as I can tell, and I am not asking for it to change.
Body Control, all seven masks at weight 0 (Head, Pelvis, LeftArm, RightArm, LeftLeg, RightLeg, Locomotion) for the duration of the state. The limbs were freed, but the body still stayed upright. This is the part I would most like a second opinion on — I expected Pelvis at 0 to hand the pelvis orientation to the animation, and it did not. If that is intended, then there is simply no mask for this; if it is not intended, it may be a bug worth its own report.
Baking the rotation into the bones (Unity's Bake Into Pose, so it becomes muscle data rather than root motion). This works on desktop and is overridden by the IK solver in VR.
Route 3 is the interesting one, because it means the animation data is already in the right form — it is muscle/pose data, not root data. Two things follow:
It sits on the content side of your authority boundary. Player position and facing stay authoritative and untouched.
It should need no netcode change. Remote avatars replay transmitted muscle values rather than re-solving, so whatever the wearer's local solve produces already propagates. If the local solve stopped overriding the animation's spine/pelvis orientation, remote viewers would see it for free.
The ask, smallest version first:
A Body Control mask that covers spine/pelvis orientation, so an animation can hold the body's own facing while a state is active — or confirmation that the existing Pelvis mask is meant to do this, in which case the above is a bug rather than a feature request.
Failing that, any bounded, avatar-declared way for an animation to own body orientation for the duration of a state, returning it on exit the way Body Control tasks already do.
Other remarks
I want to be upfront that I understand why this is closed off today, because it is the same reason I am asking narrowly. Player position and orientation are authoritative state, and an avatar being able to author them is exactly what "never trust the client" exists to prevent. I am not asking for the capsule, or for anything that relocates or reorients the player in the world — only for the displayed body, for the length of an animation the wearer started, restored on exit.
Risks worth naming, since I would rather you weigh them than have me pretend they are absent:
Disorientation in VR. If the rendered body turns while the headset does not, that is a comfort question. Scoping it to a state's duration with a hard restore on exit, exactly as Body Control tasks already work, is the conservative shape.
Griefing surface. It is bounded by the same limits as any Body Control task — the wearer's own avatar, their own animator, no effect on anyone else's view of the world. It should be possible to abuse this no more than a full-body emote already can.
If a spine/pelvis orientation mask is too broad, even a narrower "the animation may set body pitch/roll while this state is active" would cover the transforming-avatar case entirely.
Context, in case it is useful: this comes from converting VRChat avatars to ChilloutVR. VRChat solves the same problem by letting an avatar raise its Action playable layer's weight, which hands that sequence the whole body. That is a much broader mechanism than what I am asking for here, and I do not think you should copy it — a scoped Body Control mask fits your model far better and does not hand avatars anything they cannot already do to their own limbs.
Happy to supply the test avatar, the converted controller, or the measurements from any of the three routes above if that helps.
Pre-submission checklist
What should be possible that currently is not
An avatar should be able to animate its whole body turning over — a biped folding down into a car, a character dropping onto all fours — and have it play in VR.
Right now the pose is correct at the start and correct at the end, and the turn between them happens in a single frame. On desktop it plays properly; in VR the body stays upright for the whole sequence and snaps at the finish.
To be clear about what I am NOT asking for: nothing that moves the player. Position stays entirely yours. This is about the orientation of the rendered body during an authored animation the wearer triggered themselves.
What I tried, measured on a real avatar worn in game:
Root motion. Left as root rotation, it does nothing — the character controller keeps the capsule upright and root motion is discarded. Correct behaviour as far as I can tell, and I am not asking for it to change.
Body Control, all seven masks at weight 0 (Head, Pelvis, LeftArm, RightArm, LeftLeg, RightLeg, Locomotion) for the duration of the state. The limbs were freed, but the body still stayed upright. This is the part I would most like a second opinion on — I expected Pelvis at 0 to hand the pelvis orientation to the animation, and it did not. If that is intended, then there is simply no mask for this; if it is not intended, it may be a bug worth its own report.
Baking the rotation into the bones (Unity's Bake Into Pose, so it becomes muscle data rather than root motion). This works on desktop and is overridden by the IK solver in VR.
Route 3 is the interesting one, because it means the animation data is already in the right form — it is muscle/pose data, not root data. Two things follow:
It sits on the content side of your authority boundary. Player position and facing stay authoritative and untouched.
It should need no netcode change. Remote avatars replay transmitted muscle values rather than re-solving, so whatever the wearer's local solve produces already propagates. If the local solve stopped overriding the animation's spine/pelvis orientation, remote viewers would see it for free.
The ask, smallest version first:
A Body Control mask that covers spine/pelvis orientation, so an animation can hold the body's own facing while a state is active — or confirmation that the existing Pelvis mask is meant to do this, in which case the above is a bug rather than a feature request.
Failing that, any bounded, avatar-declared way for an animation to own body orientation for the duration of a state, returning it on exit the way Body Control tasks already do.
Other remarks
I want to be upfront that I understand why this is closed off today, because it is the same reason I am asking narrowly. Player position and orientation are authoritative state, and an avatar being able to author them is exactly what "never trust the client" exists to prevent. I am not asking for the capsule, or for anything that relocates or reorients the player in the world — only for the displayed body, for the length of an animation the wearer started, restored on exit.
Risks worth naming, since I would rather you weigh them than have me pretend they are absent:
Disorientation in VR. If the rendered body turns while the headset does not, that is a comfort question. Scoping it to a state's duration with a hard restore on exit, exactly as Body Control tasks already work, is the conservative shape.
Griefing surface. It is bounded by the same limits as any Body Control task — the wearer's own avatar, their own animator, no effect on anyone else's view of the world. It should be possible to abuse this no more than a full-body emote already can.
If a spine/pelvis orientation mask is too broad, even a narrower "the animation may set body pitch/roll while this state is active" would cover the transforming-avatar case entirely.
Context, in case it is useful: this comes from converting VRChat avatars to ChilloutVR. VRChat solves the same problem by letting an avatar raise its Action playable layer's weight, which hands that sequence the whole body. That is a much broader mechanism than what I am asking for here, and I do not think you should copy it — a scoped Body Control mask fits your model far better and does not hand avatars anything they cannot already do to their own limbs.
Happy to supply the test avatar, the converted controller, or the measurements from any of the three routes above if that helps.