fix(characters): rig-independent weapon mount, measured facing, revert Miku

WEAPON MOUNT, all models. set_weapon() seated the gun with a fixed
rotation_degrees = (0, 90, -90). A bone attachment is expressed in the
BONE's axes and no two rigs agree on those, so one constant mounted the
weapon differently on every character. It never needed to be right: the
pose layer aims the gun by rotating the wrist until the weapon's forward
lies on the aim line, so the identity means "forward is the hand bone's
-Z", which is true on any rig, and the wrist absorbs the roll. The grip
now sits at the bone origin, so the gun is in the hand rather than at an
offset from a differently-oriented bone. Verified by render on kiyoko:
rifle shouldered, both hands on it.

FACING. flatten_and_scale() now measures toes-versus-ankles and snaps the
character to face Blender -Y, the convention the runtime's blanket flip is
built around. Kiyoko was 180 degrees off. Snapped to the nearest quarter
turn so splayed feet in a rest pose are not read as a turned character.

MIKU. Re-imported without --grow-cloth. Her grown hair chains were the
cause of the stretching: cloth_bones.py clears a vertex's body weights and
re-assigns it to the fitted polyline, so a poor fit does not degrade to
"stiff", it degrades to "torn". She now has no hair simulation — stiff but
correct — until that tool blends against the weights it replaces instead
of destroying them.

ARIA'S SKIRT IS NOT SIMULATED. Worth recording plainly, because it looks
better than Taila's and the obvious conclusion is the wrong one: aria has
ZERO cloth chains. Her skirt never clips because it is rigidly skinned and
follows the legs it is weighted to. There is nothing to port to Taila
except switching her simulation off.

Co-Authored-By: Claude Opus 5 <[email protected]>
This commit is contained in:
Nicholas Butzke
2026-07-26 16:29:04 -04:00
co-authored by Claude Opus 5
parent 27c4117c25
commit c8b0337aa7
12 changed files with 609 additions and 2310 deletions
@@ -134,9 +134,13 @@ faces the other way, the two compose to a character running backwards — and
nothing anywhere measures the finished result.
**Rule:** facing is measurable from the skeleton — the toes are forward of the
ankles, and the nose/head is forward of the spine. Assert it on the OUTPUT, after
every transform has been applied. A blanket constant like `facing_flip` is a
guess about the source that must be replaced by a measurement.
ankles. `flatten_and_scale()` now snaps that to Blender -Y, the convention the
runtime flip is built around, so every character leaves the pipeline pointing the
same way whatever the source did. Kiyoko was 180 degrees off and is now correct.
Snap to the nearest QUARTER TURN, not to the measured angle: a rest pose with the
feet slightly splayed is not a character who is 7 degrees turned, and correcting
it as one puts a permanent yaw on the whole skeleton.
---
@@ -174,9 +178,21 @@ against proportions like Taila's. A rig with different arm lengths, a different
rest pose (A-pose vs T-pose) or a different bone roll will put the hands
somewhere plausible for the maths and wrong for the eye.
**Rule:** the hold has measurable success criteria — the distance from each hand
bone to the grip point it was solved onto. Assert those, per character, rather
than judging by eye.
**The real cause turned out to be #2's sibling, and it affected every model.**
`set_weapon` mounted the gun with a fixed `rotation_degrees = (0, 90, -90)`. A
bone attachment is expressed in the BONE's axes, and no two rigs agree on those,
so a constant there mounts the weapon differently on every character.
It does not need to be right at all. The pose layer aims the gun by rotating the
WRIST until the weapon's forward axis lies on the aim line, so handing it the
IDENTITY means "the gun's forward is the hand bone's -Z" — true by construction
on any rig — and the wrist absorbs whatever roll that bone happens to have. The
grip is placed at the bone's origin, so the gun sits *in* the hand instead of at
a fixed offset from a differently-oriented bone.
**Rule:** never express a mount as a constant in a bone's local frame. Either
derive the frame from the skeleton, or hand the downstream solver an identity and
let it do the work — it already knows where the gun must point.
---
+6 -6
View File
@@ -9,20 +9,20 @@
"Neck",
"Head"
],
"shoulder.L": "Left shoulder",
"shoulder.R": "Right shoulder",
"foot.R": "Right ankle",
"shoulder.L": "Left shoulder",
"foot.L": "Left ankle",
"hand.L": "Left wrist",
"foot.R": "Right ankle",
"hand.R": "Right wrist",
"forearm.L": "Left elbow",
"hand.L": "Left wrist",
"forearm.R": "Right elbow",
"shin.R": "Right knee",
"forearm.L": "Left elbow",
"shin.L": "Left knee",
"shin.R": "Right knee",
"toe.R": "Right toe",
"toe.L": "Left toe",
"upper_arm.L": "Left arm",
"upper_arm.R": "Right arm",
"upper_arm.L": "Left arm",
"thigh.L": "Left leg",
"thigh.R": "Right leg"
},
Binary file not shown.
+5 -5
View File
@@ -15,16 +15,16 @@
"shin.L": "J_Bip_L_LowerLeg",
"thigh.R": "J_Bip_R_UpperLeg",
"thigh.L": "J_Bip_L_UpperLeg",
"shoulder.R": "J_Bip_R_Shoulder",
"shoulder.L": "J_Bip_L_Shoulder",
"shoulder.R": "J_Bip_R_Shoulder",
"forearm.R": "J_Bip_R_LowerArm",
"forearm.L": "J_Bip_L_LowerArm",
"upper_arm.R": "J_Bip_R_UpperArm",
"upper_arm.L": "J_Bip_L_UpperArm",
"upper_arm.R": "J_Bip_R_UpperArm",
"toe.L": "J_Bip_L_ToeBase",
"toe.R": "J_Bip_R_ToeBase",
"foot.R": "J_Bip_R_Foot",
"foot.L": "J_Bip_L_Foot",
"foot.R": "J_Bip_R_Foot",
"hand.L": "J_Bip_L_Hand",
"hand.R": "J_Bip_R_Hand"
},
@@ -550,7 +550,7 @@
],
[
-0.02312,
0.02503,
0.02504,
0.00687
]
],
@@ -1899,7 +1899,7 @@
0.0013
],
[
-0.01334,
-0.01333,
0.01982,
-0.00015
],
Binary file not shown.
File diff suppressed because it is too large Load Diff
Binary file not shown.
+7 -7
View File
@@ -11,18 +11,18 @@
"Unused_Noname_004",
"Unused_Noname_010"
],
"shoulder.L": "Left shoulder",
"shoulder.R": "Right shoulder",
"foot.R": "Right ankle",
"shoulder.L": "Left shoulder",
"foot.L": "Left ankle",
"foot.R": "Right ankle",
"hand.L": "Left wrist",
"hand.R": "Right wrist",
"forearm.L": "Left elbow",
"forearm.R": "Right elbow",
"shin.R": "Right knee",
"shin.L": "Left knee",
"toe.R": "Toes2_R",
"toe.L": "Toes3_L",
"shin.R": "Right knee",
"toe.R": "Toes0_R",
"toe.L": "Toes4_L",
"upper_arm.L": "Left arm",
"upper_arm.R": "Right arm",
"thigh.L": "Left leg",
@@ -1804,8 +1804,8 @@
"Root_006",
"Root_007",
"Spine",
"Toes2_R",
"Toes3_L",
"Toes0_R",
"Toes4_L",
"Unused_Noname_004",
"Unused_Noname_010"
]
Binary file not shown.
File diff suppressed because it is too large Load Diff
+20 -10
View File
@@ -908,17 +908,27 @@ func set_weapon(script_path: String) -> void:
# (weapon swap while toggled), else the new weapon comes up invisible.
if (shadows_only or first_person_mode) and not _owner_visible:
_set_shadows_recursive(w)
# Undo the first-person viewmodel placement from the weapon's _ready:
# lie along the hand's grip, scaled down to character proportions.
w.position = Vector3(-0.02, 0.07, 0.0)
w.rotation_degrees = Vector3(0, 90, -90)
w.scale = Vector3(1.0, 1.0, 1.0)
# Tell the pose layer the gun's axes in hand-bone space so it can
# aim the wrist to point the muzzle exactly where the player looks.
# Seat the weapon in the hand with NO hand-relative rotation.
#
# There used to be a fixed `rotation_degrees = (0, 90, -90)` here, which
# is the offset that happens to be right for the library's own DEF-hand.R
# and is wrong for every rig whose hand bone has a different roll. A bone
# attachment is expressed in the BONE's axes, and no two rigs agree on
# those, so a constant here mounts the gun differently on every character
# — which is what "the hand mount points are totally wrong on all models"
# was.
#
# It does not need to be right, because the pose layer aims the gun by
# rotating the WRIST until the weapon's forward axis lies on the aim line
# (see _apply_rifle_hold). Handing it the identity means "the gun's
# forward is the hand bone's -Z", which is true by construction on any
# rig, and the wrist then absorbs whatever that bone's roll happens to be.
# The grip is placed at the bone's origin below, so the gun sits IN the
# hand rather than at a fixed offset from a differently-oriented bone.
w.transform = Transform3D.IDENTITY
if _pose_mod:
var b: Basis = w.transform.basis.orthonormalized()
_pose_mod.gun_fwd_hand = b * Vector3(0, 0, -1)
_pose_mod.gun_up_hand = b * Vector3(0, 1, 0)
_pose_mod.gun_fwd_hand = Vector3(0, 0, -1)
_pose_mod.gun_up_hand = Vector3(0, 1, 0)
_measure_weapon(w)
)
+39
View File
@@ -413,6 +413,45 @@ def flatten_and_scale(arm, meshes, target_height, roles=None):
print(f"Stood the character up: feet->hips ran along "
f"{'XYZ'[k]}{'+' if d[k] > 0 else '-'}, rotated to Z+")
# TURN THE CHARACTER TO FACE THE SAME WAY AS EVERY OTHER CHARACTER.
#
# Which way a model faces is a property of the file, not a constant, and two
# separate things were guessing at it: facing_correction() aligns the rest
# pose to the library's, and the runtime then applies a blanket 180 degrees
# because "glTF forward is +Z". When a source disagrees with either, the two
# compose into a character who runs backwards, and nothing measured the
# result. Kiyoko shipped like that.
#
# The skeleton knows: toes are in front of ankles. Snap that to Blender -Y,
# which is the convention the rest of the pipeline and the runtime flip are
# built around, so every character leaves here pointing the same way whatever
# the source did.
if roles is not None:
fwd = Vector((0.0, 0.0, 0.0))
n_f = 0
for side in ("L", "R"):
a = arm.data.bones.get(roles.limb.get(("foot", side), ""))
t = arm.data.bones.get(roles.limb.get(("toe", side), ""))
if a and t:
fwd += (t.head_local - a.head_local)
n_f += 1
if n_f:
fwd.z = 0.0
if fwd.length > 1e-4:
fwd.normalize()
want = Vector((0.0, -1.0, 0.0))
ang = math.atan2(fwd.x, -fwd.y) # signed yaw from -Y to fwd
# Snap to the nearest quarter turn: a rest pose with the feet
# slightly splayed must not be counted as a turn.
q = round(ang / (math.pi / 2)) * (math.pi / 2)
if abs(q) > 1e-6:
R = Matrix.Rotation(-q, 4, Vector((0.0, 0.0, 1.0)))
for obj in [arm] + meshes:
obj.matrix_world = R @ obj.matrix_world
apply_all()
print("Turned the character to face -Y: was %.0f degrees off"
% math.degrees(q))
lo = Vector((1e9, 1e9, 1e9))
hi = -lo.copy()
for m in meshes: