Files
Papay-Shooter/tools/retarget.py
T
Nicholas ButzkeandClaude Opus 5 27c4117c25 fix(pipeline): stand the character up before scaling; reach bones by role
Fixes the two root causes behind four of the seven reported breakages, and
rejects the source that cannot be fixed.

STAND UP FIRST. flatten_and_scale() now derives the up axis from the
skeleton and rotates the model upright before measuring anything. Aria and
Momo are correct: 1.75 m tall, 1.43 x 0.41 and 1.52 x 0.39 across, verified
by render.

The up vector is measured from the FEET to the HIPS, not from the hips to
the head. The head is not a reliable landmark — the spine walk ends on the
last non-cosmetic bone in the chain, which on a rig with a facial skeleton
can sit BELOW the hips. Momo's did, so the first cut of this fix stood her
neatly on her head: right size, right proportions, upside down. Feet cannot
be mistaken.

REACH BONES BY ROLE. SkinnedPlayerModel gained _role_bone(), and set_weapon
uses it. Four characters could not hold a gun because one hardcoded lookup
knew three spellings and their hands are called "Right wrist" and
"J_Bip_R_Hand" — both resolved perfectly in the sidecar the whole time.

HIKARI IS REJECTED. She now fails the gate: her feet and spine disagree
about which way is up, so the stand-up correction cannot resolve her
either, on top of zero-length cosmetic bones and a second armature that was
smuggling its own clips into the export. That is not a tuning problem, it
is a file that has been through two toolchains. De-registered and removed
rather than shipped broken — which is what the gate is for.

Six GLB skins remain, all passing. Smoke 0 failures, 11/11 movement tests.

Still open, recorded in the skill: kiyoko faces backwards, miku's grown
hair stretches under animation, the mannequin's rifle hold does not
convince, and taila's front skirt clipping.

Co-Authored-By: Claude Opus 5 <[email protected]>
2026-07-26 16:06:51 -04:00

1441 lines
60 KiB
Python

#!/usr/bin/env python3
"""
Put the animation library onto a character WITHOUT touching how it deforms.
This replaces the old strip_rig -> autorig -> merge_animations route, which
solved a naming problem by destroying the asset. That route threw away the
character's skeleton, joined every mesh into one blob, and rebound the result
with nearest-four-bones Euclidean weights. Measured on the shipped taila.glb it
produced 2817 vertices pulled by BOTH legs (worst a dead 50/50 split) and 86%
of all vertices carrying the full four influences — while the ORIGINAL file it
was built from had zero cross-leg bleed, one mesh per material, and dedicated
bone chains for the skirt and the hair. Every runtime "repair" in
characters/skin_leg_repair.gd exists to undo damage done right here.
So: keep the character's own rig, weights, per-part meshes and cloth chains,
and move the ANIMATION onto it instead.
1. Rebuild parenting. A Rigify DEF-rig exports its chain roots parented
straight to the armature root, because Rigify drives them by constraint
rather than hierarchy. Left that way, rotating the hips would leave the
legs, skirt and hair floating in place. Orphans are re-attached by
anatomy where it is known and by rest geometry otherwise — and cloth may
only ever attach to the trunk, never to a limb, or a skirt would ride one
thigh.
2. Retarget by ROLE, not by name (see tools/rig_map.py).
3. Bake each clip as a rest-relative delta:
R_world = src_pose_rot * src_rest_rot^-1 (what the clip does)
tgt_rot = R_world * tgt_rest_rot (done to THIS rig)
Copying absolute world orientation instead — which is what the old
constraint bake did — forces the library's bone roll onto a mesh that was
bound with a different one, and twists every limb by a constant offset.
4. Drive ONLY the body. Skirt, hair, twist and face bones are left with no
keys at all, so they rest relative to their parents and are free for the
spring solver at runtime. That split — clips animate the body, physics
animates the cloth — is the whole point.
Usage:
blender --background --python tools/retarget.py -- \
<character.glb> <animations_dir> <output.glb> [--height 1.75] [--keep-root-motion]
Writes <output>.rig.json beside the GLB: resolved bone roles, cloth chains and
twist pairs, so the runtime never has to re-guess the skeleton's anatomy.
"""
import bpy
import math
import json
import os
import sys
import tempfile
from collections import defaultdict
from mathutils import Matrix, Quaternion, Vector
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
import gltf_fix
import rig_map
from rig_map import RigRoles, build_map, is_cosmetic, is_segment_of, tokens
argv = sys.argv
argv = argv[argv.index("--") + 1:] if "--" in argv else []
if len(argv) < 3:
print(__doc__)
sys.exit(1)
CHARACTER, ANIM_DIR, OUTPUT = argv[0], argv[1], argv[2]
STRIP_ROOT_MOTION = "--keep-root-motion" not in argv
TARGET_HEIGHT = 1.75
if "--height" in argv:
TARGET_HEIGHT = float(argv[argv.index("--height") + 1])
OVERRIDES = {}
if "--bone-map" in argv:
with open(argv[argv.index("--bone-map") + 1], "r", encoding="utf-8") as f:
OVERRIDES = json.load(f)
# Library clip -> the game's canonical clip name. Unmapped clips are skipped so
# character GLBs stay small.
LIBRARY_CLIP_MAP = {
"Idle_Loop": "Idle",
"Walk_Loop": "Walk",
"Jog_Fwd_Loop": "Run",
"Sprint_Loop": "Sprint",
"Jump_Start": "Jump",
"Jump_Loop": "Fall",
"Jump_Land": "Land",
"Crouch_Idle_Loop": "CrouchIdle",
"Crouch_Fwd_Loop": "CrouchWalk",
"Roll": "Dash",
"Death01": "Death",
"Hit_Chest": "Hit",
"Dance_Loop": "Dance",
"Swim_Fwd_Loop": "Grapple",
"Pistol_Idle_Loop": "PistolIdle",
"Pistol_Shoot": "PistolShoot",
"Pistol_Reload": "PistolReload",
"Sword_Attack": "Throw",
}
UP = Vector((0.0, 0.0, 1.0))
# Which cosmetic chains get secondary motion. A face-shape or eye chain is
# cosmetic but must never swing, so this is deliberately narrower than
# rig_map.COSMETIC.
SPRING_CLASSES = {"hair", "skirt", "cloth", "ribbon", "tail", "cape", "coat",
"scarf", "sleeve", "breast", "bust", "feather", "strap",
"antenna", "wing"}
def spring_class(name):
"""Which secondary-motion class this bone belongs to, or None.
Whole token first, then a SHORT positional suffix — the same rule as
rig_map.is_cosmetic, and it must be the same rule, because a bone that reads
as cosmetic there and as nothing here is left out of every chain and its
geometry never moves. DANDADAN's hair is HairFL / HairFR / HairF_Top, which
tokenise to "hairfl" and matched no class at all: she imported with six
chains, all of them her bust, and not one strand of hair.
"""
for t in tokens(name):
if t in SPRING_CLASSES:
return t
for c in SPRING_CLASSES:
if len(t) - len(c) <= 2 and t.startswith(c) and len(t) > len(c):
return c
return None
# --------------------------------------------------------------------- import
def clear_scene():
bpy.ops.object.select_all(action="SELECT")
bpy.ops.object.delete()
for blocks in (bpy.data.meshes, bpy.data.armatures, bpy.data.actions):
for b in list(blocks):
if b.users == 0:
blocks.remove(b)
def import_any(path):
ext = os.path.splitext(path)[1].lower()
if ext in (".glb", ".gltf"):
bpy.ops.import_scene.gltf(filepath=path)
elif ext == ".fbx":
bpy.ops.import_scene.fbx(filepath=path)
else:
print(f"ERROR: unsupported character format {ext}")
sys.exit(1)
def skinned_meshes(arm):
"""Meshes actually driven by this armature.
Anything else in the file is scene dressing — Sketchfab models routinely
ship a display base or a diorama, and those must not become part of the
player.
"""
out = []
for o in [o for o in bpy.data.objects if o.type == "MESH"]:
driven = any(m.type == "ARMATURE" and m.object is arm for m in o.modifiers)
if driven or (o.parent is arm and o.vertex_groups):
out.append(o)
return out
def strip_import_suffixes(arm, meshes):
"""`DEF-thigh.L_16` -> `DEF-thigh.L`.
The glTF importer appends the node index to every bone name. Those names
ship in the exported GLB and are what the game code matches on, so clean
them up here rather than teaching every consumer about the suffix.
"""
renames = {}
taken = set(b.name for b in arm.data.bones)
for bone in arm.data.bones:
clean = rig_map.strip_gltf_suffix(bone.name)
if clean != bone.name and clean not in taken:
renames[bone.name] = clean
taken.discard(bone.name)
taken.add(clean)
for old, new in renames.items():
arm.data.bones[old].name = new
# Vertex groups are matched to bones by NAME and are not renamed for us.
for m in meshes:
for vg in m.vertex_groups:
new = renames.get(vg.name)
if new and new not in m.vertex_groups:
vg.name = new
print(f"Cleaned {len(renames)} bone names")
# ----------------------------------------------------------------- hierarchy
def _seg_distance(p, a, b):
ab = b - a
d2 = ab.dot(ab)
t = 0.0 if d2 < 1e-12 else max(0.0, min(1.0, (p - a).dot(ab) / d2))
return (p - (a + ab * t)).length
def rebuild_hierarchy(arm, roles):
"""Re-attach chain roots that exported parented to the armature root.
Anatomy first (a thigh belongs to the hips, a forearm to the upper arm),
rest geometry second. Cloth and hair are only ever allowed to attach to the
TRUNK: pick anchors by raw proximity and a skirt panel hanging beside a leg
attaches to that thigh and rides it like a trouser leg.
"""
trunk = [roles.hips] + list(roles.spine)
trunk = [n for n in trunk if n]
limb = roles.limb
def anat(role, side):
return limb.get((role, side))
fixed = {}
for side in ("L", "R"):
chest = trunk[-3] if len(trunk) >= 3 else (trunk[-1] if trunk else None)
pairs = [
(("thigh", side), roles.hips),
(("shin", side), anat("thigh", side)),
(("foot", side), anat("shin", side)),
(("toe", side), anat("foot", side)),
(("shoulder", side), chest),
(("upper_arm", side), anat("shoulder", side) or chest),
(("forearm", side), anat("upper_arm", side)),
(("hand", side), anat("forearm", side)),
]
for key, parent in pairs:
name = limb.get(key)
if name and parent:
fixed[name] = parent
bpy.context.view_layer.objects.active = arm
bpy.ops.object.mode_set(mode="EDIT")
eb = arm.data.edit_bones
# Where each anchor bone actually EXTENDS TO, from authored data only.
#
# glTF stores joints as nodes with no tail, so the tails Blender reports are
# invented — every one of Taila's skirt bones comes back 0.78 m long. Using
# them, DEF-spine's phantom tail runs straight down through the whole skirt
# and every panel measures ~0.05 m from "the hips", which beat each panel's
# real 0.15 m link to its own chain root and flattened all 7 chains.
# A bone's true extent is the head of the next bone along.
span = {}
for i, name in enumerate(trunk):
nxt = trunk[i + 1] if i + 1 < len(trunk) else None
span[name] = (eb[name].head,
eb[nxt].head if nxt and nxt in eb else eb[name].tail)
for name in limb.values():
if name not in eb:
continue
kids = [c for c in eb[name].children if not is_cosmetic(c.name)]
span[name] = (eb[name].head, kids[0].head if kids else eb[name].tail)
def anchor_distance(point, name):
a, b = span.get(name, (eb[name].head, eb[name].tail))
return _seg_distance(point, a, b)
def is_root_like(bone):
return bone is None or any(
t in ("root", "master", "armature", "scene", "rootjoint")
for t in tokens(bone.name))
def descendants(bone):
out = {bone.name}
stack = list(bone.children)
while stack:
b = stack.pop()
out.add(b.name)
stack.extend(b.children)
return out
trunk_set = set(trunk)
core = trunk_set | set(limb.values())
orphans = [b for b in eb
if b.name != roles.hips and b.name not in trunk_set
and is_root_like(b.parent)]
reparented = 0
cosmetic_left = []
for bone in orphans:
if is_cosmetic(bone.name):
cosmetic_left.append(bone)
continue
target = fixed.get(bone.name)
if target is None:
banned = descendants(bone)
cands = [n for n in core if n in eb and n not in banned]
if not cands:
continue
target = min(cands, key=lambda n: anchor_distance(bone.head, n))
if target in eb and target != bone.name:
bone.parent = eb[target]
bone.use_connect = False
reparented += 1
# Cloth and hair hang from the TRUNK — never from a limb, and never from
# each other.
#
# Attaching cloth to whatever bone is nearest puts 16 of Taila's 21 skirt
# bones on a thigh, where the panel rides one leg like a trouser leg. The
# trunk restriction fixes that.
#
# Reconstructing multi-bone chains is deliberately NOT attempted. The
# temptation is obvious — Taila's skirt is really 7 panels of 3 — but the
# information is not in the file. glTF stores no bone tails, the panel
# numbering is not sequential (`skirt` -> `skirt.011` -> `skirt.002`), and
# neighbouring panel roots ring the waist 0.04 m apart, far closer than any
# of them is to the trunk. Successive attempts at distance, chain-direction
# and grow-outward rules each produced a topology that was still wrong
# somewhere — stitching panels together sideways, or hanging hair off an
# eye bone. A wrong chain is worse than no chain: the solver then swings
# bones along axes the mesh was never weighted for, and tears it.
#
# So each orphan becomes its own pendulum from the body. Chains the ARTIST
# authored survive untouched, because only chain ROOTS are orphans — which
# is why Taila's hair keeps its real 4-7 bone strands while her
# flat-exported skirt becomes per-panel pendulums. Both look like cloth;
# only the authored one gets true multi-segment drape.
for bone in cosmetic_left:
cands = [n for n in trunk_set if n in eb]
if not cands:
break
bone.parent = eb[min(cands, key=lambda n: anchor_distance(bone.head, n))]
bone.use_connect = False
reparented += 1
bpy.ops.object.mode_set(mode="OBJECT")
print(f"Re-attached {reparented} orphaned bones")
# ----------------------------------------------------------------- normalize
def flatten_and_scale(arm, meshes, target_height, roles=None):
"""Bake the import hierarchy away and set the character's real-world size.
Sketchfab wraps everything in scaled/rotated empties. Left in place they
turn up as a scale on the exported Skeleton3D, and every measurement the
game makes off bone rests reads in the wrong units.
"""
if arm.animation_data:
arm.animation_data_clear()
for pb in arm.pose.bones:
pb.matrix_basis = Matrix()
for obj in [arm] + meshes:
world = obj.matrix_world.copy()
obj.parent = None
obj.matrix_world = world
for o in [o for o in bpy.data.objects if o.type == "EMPTY"]:
bpy.data.objects.remove(o, do_unlink=True)
def apply_all():
bpy.ops.object.select_all(action="DESELECT")
for obj in [arm] + meshes:
obj.select_set(True)
bpy.context.view_layer.objects.active = arm
bpy.ops.object.transform_apply(location=True, rotation=True, scale=True)
apply_all()
# STAND THE CHARACTER UP before measuring anything.
#
# The scale below normalises the bounding box along Z because Z is up in
# Blender. For a model that arrives lying along another axis that measures
# the character's THICKNESS — about 0.25 m — so it gets scaled by ~7 and left
# on its back. One assumption, both symptoms, and invisible afterwards
# because the exporter maps Blender Z to glTF Y: "is the height 1.75" comes
# out true on a character who is 7.5 m tall lying down.
#
# Which way is up is not a convention to assume, it is a property of the
# skeleton: the head is above the hips. Snapped to the nearest axis rather
# than aligned exactly, so a character with a slight lean in their rest pose
# is stood up, not straightened.
# Measured from the FEET to the HIPS, not from the hips to the head.
#
# The head is not a reliable landmark. The spine walk ends on whatever the
# last non-cosmetic bone in the chain is, and on a rig with a facial skeleton
# that can be a bone sitting BELOW the hips — which points this vector
# downwards and stands the character neatly on her head. Momo did exactly
# that. Feet cannot be mistaken: they are the bottom of a standing character
# on every rig, and foot.L/R resolve on every source met so far.
if roles is not None and roles.hips:
hips_b = arm.data.bones.get(roles.hips)
feet = [arm.data.bones.get(roles.limb.get(("foot", s), ""))
for s in ("L", "R")]
feet = [f for f in feet if f]
ref = None
if feet:
ref = sum((f.head_local for f in feet), Vector()) / len(feet)
elif roles.head and arm.data.bones.get(roles.head):
# No feet — fall back to the head, and accept the risk above.
ref = hips_b.head_local - (
arm.data.bones[roles.head].head_local - hips_b.head_local)
if hips_b and ref is not None:
d = hips_b.head_local - ref
k = max(range(3), key=lambda i: abs(d[i]))
src = Vector((0.0, 0.0, 0.0))
src[k] = 1.0 if d[k] > 0 else -1.0
up = Vector((0.0, 0.0, 1.0))
if src.dot(up) < 0.999:
axis = src.cross(up)
if axis.length < 1e-6:
axis = Vector((1.0, 0.0, 0.0)) # upside down
R = Matrix.Rotation(src.angle(up), 4, axis.normalized())
for obj in [arm] + meshes:
obj.matrix_world = R @ obj.matrix_world
apply_all()
print(f"Stood the character up: feet->hips ran along "
f"{'XYZ'[k]}{'+' if d[k] > 0 else '-'}, rotated to Z+")
lo = Vector((1e9, 1e9, 1e9))
hi = -lo.copy()
for m in meshes:
for corner in m.bound_box:
p = m.matrix_world @ Vector(corner)
lo = Vector((min(lo.x, p.x), min(lo.y, p.y), min(lo.z, p.z)))
hi = Vector((max(hi.x, p.x), max(hi.y, p.y), max(hi.z, p.z)))
height = hi.z - lo.z
if height > 1e-4:
s = target_height / height
for obj in [arm] + meshes:
obj.scale = (s, s, s)
apply_all()
lo *= s
hi *= s
print(f"Scaled by {s:.4f} to {target_height:.2f} m")
offset = Vector((-(lo.x + hi.x) * 0.5, -(lo.y + hi.y) * 0.5, -lo.z))
for obj in [arm] + meshes:
obj.location = offset
apply_all()
# Re-parent meshes under the armature so the export writes one clean skin.
for m in meshes:
m.parent = arm
m.matrix_parent_inverse = Matrix()
if not any(mod.type == "ARMATURE" and mod.object is arm for mod in m.modifiers):
mod = m.modifiers.new("Armature", "ARMATURE")
mod.object = arm
def fix_unlit_materials(meshes):
"""Route each material's texture into Base Color.
Anime models are commonly authored UNLIT: black base colour with the albedo
wired to emission. Our toon shader reads ALBEDO, so left alone the
character renders pitch black.
"""
seen = set()
for m in meshes:
for mat in m.data.materials:
if not mat or not mat.use_nodes or mat.name in seen:
continue
seen.add(mat.name)
nt = mat.node_tree
tex = next((n for n in nt.nodes if n.type == "TEX_IMAGE" and n.image), None)
bsdf = next((n for n in nt.nodes if n.type == "BSDF_PRINCIPLED"), None)
if not bsdf or not tex:
continue
base = bsdf.inputs["Base Color"]
if not base.links:
nt.links.new(tex.outputs["Color"], base)
print(f"Material '{mat.name}': routed '{tex.image.name}' to base colour")
if "Emission Strength" in bsdf.inputs:
bsdf.inputs["Emission Strength"].default_value = 0.0
# ------------------------------------------------------------------ retarget
def world_rest(arm):
mw = arm.matrix_world
return {b.name: mw @ b.matrix_local for b in arm.data.bones}
def rig_forward(arm, roles):
"""Which way the rest pose faces, on the ground plane, from the feet."""
mw = arm.matrix_world
acc = Vector((0.0, 0.0, 0.0))
for side in ("L", "R"):
for role in ("toe", "foot"):
name = roles.limb.get((role, side))
if not name:
continue
b = arm.data.bones[name]
v = (mw @ b.tail_local) - (mw @ b.matrix_local.translation)
v.z = 0.0
if v.length > 1e-5:
acc += v.normalized()
break
return acc.normalized() if acc.length > 1e-5 else None
def facing_correction(src_arm, src_roles, tgt_arm, tgt_roles):
"""Yaw that carries the source rig's forward onto the target's.
Without it a library that rests facing -Y drives a character that rests
facing +Y and every clip plays backwards.
"""
a = rig_forward(src_arm, src_roles)
b = rig_forward(tgt_arm, tgt_roles)
if a is None or b is None:
return Quaternion()
# angle_signed is 2D-only in mathutils, which is what we want anyway: the
# correction is a yaw about world up, never a tilt.
angle = Vector((a.x, a.y)).angle_signed(Vector((b.x, b.y)), 0.0)
if abs(angle) < 1e-4:
return Quaternion()
print(f"Facing correction: {angle * 57.2958:.1f} deg")
return Quaternion(UP, angle)
def bone_order(arm):
"""Every bone, parents before children."""
out = []
def walk(b):
out.append(b.name)
for c in b.children:
walk(c)
for b in arm.data.bones:
if b.parent is None:
walk(b)
return out
def solve_pose(arm, order, rest_w, desired_rot, hips, hips_head):
"""Turn desired WORLD orientations into per-bone local basis transforms.
Done arithmetically rather than by setting `pose_bone.matrix` and letting
Blender solve, because that needs a depsgraph update per bone — 150 bones
across 18 clips is tens of thousands of scene evaluations.
Blender relates pose to rest as
pose = parent_pose * parent_rest^-1 * rest * basis
so with M standing for everything left of `basis`, a rotation-only basis of
M.rot^-1 * desired lands the bone on `desired` exactly.
"""
pose_w = {}
basis = {}
bones = arm.data.bones
for name in order:
b = bones[name]
rest = rest_w[name]
if b.parent is not None:
M = pose_w[b.parent.name] @ rest_w[b.parent.name].inverted() @ rest
else:
M = rest
q = Quaternion()
if name in desired_rot:
q = M.to_quaternion().inverted() @ desired_rot[name]
loc = Vector((0.0, 0.0, 0.0))
if name == hips and hips_head is not None:
loc = M.inverted() @ hips_head
basis[name] = (loc, q)
pose_w[name] = M @ Matrix.Translation(loc) @ q.to_matrix().to_4x4()
return basis
def retarget_clip(src_arm, src_roles, tgt_arm, tgt_roles, mapping, action,
clip_name, yaw, scale):
src_rest = world_rest(src_arm)
tgt_rest = world_rest(tgt_arm)
order = bone_order(tgt_arm)
src_rest_rot = {n: m.to_quaternion() for n, m in src_rest.items()}
tgt_rest_rot = {n: m.to_quaternion() for n, m in tgt_rest.items()}
yaw_inv = yaw.inverted()
src_hips = src_roles.hips
tgt_hips = tgt_roles.hips
src_hips_rest = src_rest[src_hips].translation.copy()
tgt_hips_rest = tgt_rest[tgt_hips].translation.copy()
assign_action(src_arm, action)
f0, f1 = (int(round(v)) for v in action.frame_range)
baked = bpy.data.actions.new(clip_name)
assign_action(tgt_arm, baked)
for pb in tgt_arm.pose.bones:
pb.rotation_mode = "QUATERNION"
scene = bpy.context.scene
for frame in range(f0, f1 + 1):
scene.frame_set(frame)
dg = bpy.context.evaluated_depsgraph_get()
src_eval = src_arm.evaluated_get(dg)
smw = src_eval.matrix_world
desired = {}
for tgt_name, src_name in mapping.items():
if src_name not in src_eval.pose.bones or tgt_name not in tgt_rest_rot:
continue
pose_rot = (smw @ src_eval.pose.bones[src_name].matrix).to_quaternion()
delta = pose_rot @ src_rest_rot[src_name].inverted()
desired[tgt_name] = (yaw @ delta @ yaw_inv) @ tgt_rest_rot[tgt_name]
hips_head = None
if src_hips in src_eval.pose.bones:
moved = (smw @ src_eval.pose.bones[src_hips].matrix).translation
d = yaw @ ((moved - src_hips_rest) * scale)
if STRIP_ROOT_MOTION:
d.x = 0.0
d.y = 0.0 # gameplay code moves the body; keep the vertical bob
hips_head = tgt_hips_rest + d
basis = solve_pose(tgt_arm, order, tgt_rest, desired, tgt_hips, hips_head)
for name in mapping:
if name not in basis:
continue
pb = tgt_arm.pose.bones[name]
pb.rotation_quaternion = basis[name][1]
pb.keyframe_insert("rotation_quaternion", frame=frame)
if hips_head is not None:
pb = tgt_arm.pose.bones[tgt_hips]
pb.location = basis[tgt_hips][0]
pb.keyframe_insert("location", frame=frame)
assign_action(tgt_arm, None)
return baked
def assign_action(obj, action):
if not obj.animation_data:
obj.animation_data_create()
obj.animation_data.action = action
if action is None:
return
try: # Blender 4.4+ slotted actions
if not obj.animation_data.action_slot and len(action.slots):
obj.animation_data.action_slot = action.slots[0]
except (AttributeError, TypeError):
pass
def add_nla_clip(arm, action, name):
action.name = name
track = arm.animation_data.nla_tracks.new()
track.name = name
strip = track.strips.new(name, 0, action)
strip.name = name
track.mute = True
action.use_fake_user = True
# -------------------------------------------------------------------- sidecar
def _dominant_vertices(meshes, arm, min_weight=0.25):
"""bone name -> world positions of the vertices it mostly owns.
"Mostly" as in holds the largest share — a vertex belongs to one bone for
the purpose of measuring what that bone covers, even though it is skinned
to several.
`min_weight` is how strongly a vertex must belong to its bone to count.
The collider pass raises it: a vertex split 0.3/0.3/0.4 across hip, thigh
and glute is a BLEND, and letting those in put the hip flare back into the
thigh's band samples — the fitted capsule came out 0.18 m at the head.
"""
out = defaultdict(list)
for m in meshes:
gname = {g.index: g.name for g in m.vertex_groups}
mw = m.matrix_world
for v in m.data.vertices:
best = None
for g in v.groups:
if best is None or g.weight > best.weight:
best = g
if best is not None and best.weight > min_weight:
out[gname.get(best.group, "")].append(mw @ v.co)
return out
def _bone_tip(arm, bone, chain, index, owned, fallback):
"""Where a bone effectively points, in its own rest space.
The next bone along when there is one. Otherwise the centroid of the
geometry this bone actually drives — which is the only real answer for
Taila's skirt, whose 21 panel bones export with no children and no usable
tail, so there is nothing in the skeleton to say which way a panel hangs.
"""
rest_world = arm.matrix_world @ arm.data.bones[bone].matrix_local
if index + 1 < len(chain):
nxt = arm.matrix_world @ arm.data.bones[chain[index + 1]].matrix_local
return rest_world.inverted() @ nxt.translation
pts = owned.get(bone, [])
if pts:
centroid = sum(pts, Vector((0.0, 0.0, 0.0))) / len(pts)
local = rest_world.inverted() @ centroid
# The centroid sits mid-panel, so the far edge is roughly twice out.
if local.length > 1e-4:
return local * 2.0
return fallback
def _hull_samples(arm, bone, owned, limit=10):
"""A few points spread across the geometry a cloth bone actually drives,
in that bone's own rest space.
The runtime collides THESE, not points along the bone. A skirt panel is a
wide sheet hanging off a single stick from the waist, so keeping the stick
out of the thigh is nearly meaningless: measured over a movement sweep, the
bones were clear by ~1 mm while the leg was 85 mm inside the skirt MESH with
190 vertices swallowed.
Chosen by farthest-point sampling so the handful of points spans the panel
(edges, hem, middle) instead of clustering wherever the mesh is dense.
"""
pts = owned.get(bone, [])
if len(pts) < 4:
return []
inv = (arm.matrix_world @ arm.data.bones[bone].matrix_local).inverted()
local = [inv @ p for p in pts]
picked = [max(local, key=lambda v: v.length)]
while len(picked) < min(limit, len(local)):
far = max(local, key=lambda v: min((v - q).length for q in picked))
if min((far - q).length for q in picked) < 1e-4:
break
picked.append(far)
return [[round(v.x, 5), round(v.y, 5), round(v.z, 5)] for v in picked]
# How far off the skin a garment sits — its own thickness, plus the fact that
# cloth drapes over a limb rather than being painted onto it.
CLOTH_CLEARANCE = 0.008
def _pct(sorted_values, p):
"""Value at percentile `p` of an already-sorted list."""
i = int(round(p * (len(sorted_values) - 1)))
return sorted_values[max(0, min(len(sorted_values) - 1, i))]
def _body_points(meshes):
"""Every skinned vertex that is NOT cloth, in world space.
The waist lid is sized from these. Including the garment measured the skirt
itself — a 0.24 m radius that would have held it out in a bell.
"""
out = []
for m in meshes:
gname = {g.index: g.name for g in m.vertex_groups}
mw = m.matrix_world
for v in m.data.vertices:
cloth_w = sum(g.weight for g in v.groups
if spring_class(gname.get(g.group, "")))
if cloth_w < 0.35:
out.append(mw @ v.co)
return out
def _leg_colliders(arm, roles, owned, body_pts=None):
"""TAPERED capsules for the legs, sized from the body geometry itself.
The skirt has to be kept off the thighs, and a guessed radius either lets it
clip through or holds it out in a bell. Three things the obvious version got
wrong, all measured on Taila:
* A limb is not a cylinder. Her thigh is ~0.10 m across at the hip and
~0.055 m just above the knee, so one radius is either too fat at the knee
or too thin at the hip. Head and tail radii are stored separately and
interpolated at runtime.
* A leg's own vertices are not the leg. Most of the thigh belongs to the
TWIST bone (`DEF-thigh.L.001`); what is left dominated by `DEF-thigh.L`
is mostly the hip flare, which measured a 0.154 m radius — a 30 cm thigh.
Twist children are folded in.
* Neither a low percentile nor a high one works on a POOLED bucket. The
70th tracked that flare; the median then left half the limb's surface
outside its own collider, so cloth pushed out to it was clear of the
capsule while the thigh was visibly through it in the render; and the
88th over-measured the shaft by 30% because the top bucket still holds
the hip. Measured per band along the bone instead and fitted as the line
it actually is, dropping the contaminated end bands.
* Garments have thickness and hang OFF a leg rather than painted onto it,
so a small clearance is added on top. Without it the cloth's rest state
is exactly tangent to the limb and every frame is a contact.
"""
names = set(b.name for b in arm.data.bones)
out = []
# A LID across the waist first.
#
# Magica Cloth 2's skirt guide is blunt about this: put "one big sphere
# collider on your waist", because it "acts as a lid that prevents particles
# in the skirt from slipping into the body". Leg capsules alone only stop
# cloth going through a thigh — nothing stops a panel being swung INWARD
# between the legs and ending up inside the pelvis, which is where several
# of the worst contacts here were sitting.
trunk = [roles.hips] + [n for n in roles.spine if n != roles.hips]
if len(trunk) >= 2 and trunk[0] in names and trunk[1] in names:
a = (arm.matrix_world @ arm.data.bones[trunk[0]].matrix_local).translation
b = (arm.matrix_world @ arm.data.bones[trunk[1]].matrix_local).translation
# Sized from the geometry that actually surrounds the pelvis, not from
# the hip bone's own vertices: on a Rigify rig the hips own almost
# nothing (2 vertices here, and spine.001 none) because the torso
# belongs to spine.002, so there is nothing there to measure.
ab = b - a
d2 = ab.dot(ab)
pts = []
if d2 > 1e-9:
for p in (body_pts or []):
t = (p - a).dot(ab) / d2
if 0.0 <= t <= 1.0:
pts.append(p)
if len(pts) >= 12:
rr = sorted(_seg_distance(p, a, b) for p in pts)
# 60th percentile, not the 90th used for limbs: the lid only has to
# stop cloth being swung INTO the body. Sized to the widest thing
# near the hips it would hold the whole skirt out in a bell.
r = _pct(rr, 0.60) + CLOTH_CLEARANCE
out.append({
"bone": trunk[0], "child": trunk[1],
"from": 0.0,
"radius_head": round(r, 4),
"radius_tail": round(r, 4),
"radius": round(r, 4),
# A lid is something to stay OUT of, not a limb to be carried by.
"lid": True,
})
for role, child_role in (("thigh", "shin"), ("shin", "foot")):
for side in ("L", "R"):
name = roles.limb.get((role, side))
child = roles.limb.get((child_role, side))
if not name or not child:
continue
a = (arm.matrix_world @ arm.data.bones[name].matrix_local).translation
b = (arm.matrix_world @ arm.data.bones[child].matrix_local).translation
ab = b - a
d2 = ab.dot(ab)
if d2 < 1e-9:
continue
# This bone plus any twist segment hanging off it — together they
# are the limb.
pts = list(owned.get(name, []))
for other in names:
if other != name and other.startswith(name + ".") \
and is_segment_of(other, names):
pts.extend(owned.get(other, []))
if len(pts) < 12:
continue
# A limb is a TAPER, so measure it as one. Ten bands along the
# bone, the 90th percentile radius in each, and a least-squares
# line through them. Two pooled buckets could not do this: the top
# bucket is contaminated by the hip flare and the bottom one by the
# knee and boot, so whatever percentile was chosen came out wrong
# at one end or the other — a median under-measured the limb by
# half its surface, and a high percentile over-measured it by 30%
# along the whole shaft. The two end bands are dropped for exactly
# that reason; the fit extrapolates back through them.
bands = [[] for _ in range(10)]
for p in pts:
t = max(0.0, min(1.0, (p - a).dot(ab) / d2))
bands[min(int(t * 10.0), 9)].append(_seg_distance(p, a, b))
samples = []
for k in range(1, 9):
if len(bands[k]) < 3:
continue
bands[k].sort()
samples.append(((k + 0.5) / 10.0, _pct(bands[k], 0.90)))
if len(samples) < 3:
continue
n = len(samples)
mt = sum(t for t, _ in samples) / n
mr = sum(r for _, r in samples) / n
den = sum((t - mt) ** 2 for t, _ in samples)
slope = sum((t - mt) * (r - mr) for t, r in samples) / den if den > 1e-9 else 0.0
head = mr + slope * (0.0 - mt) + CLOTH_CLEARANCE
tail = mr + slope * (1.0 - mt) + CLOTH_CLEARANCE
# A limb never widens toward the joint below it, and a fit through
# noisy bands occasionally says otherwise.
tail = max(0.01, min(tail, head))
head = max(head, tail)
# The capsule starts BELOW the hip joint.
#
# The top of a thigh is not a free limb, it is the hip, and it is
# buried inside the body the skirt hangs from. Colliding against it
# asks the solver for something it cannot do: those cloth points sit
# 20-30 mm from their own bone's head, and rotating a bone moves a
# point near its pivot by almost nothing — measured, 24 mm of lever
# against 86 mm of overlap, where the most any rotation can achieve
# is twice the lever. The solver spent all six passes saturated at
# its per-pass cap and still left 60-90 mm.
#
# Only the sphere cap buried in the pelvis is cut. The upper thigh
# itself stays covered, because the runtime can also SHIFT a chain
# bodily (SpringBones.PUSH_MAX) and a shift does not care how much
# lever the bone has: rotation handles the contacts with leverage,
# translation handles the ones without. Trimming 30% instead of 10%
# stopped the solver even trying across the top of the thigh, and
# that band is exactly what then showed through the skirt.
head_t = 0.10
out.append({
"bone": name, "child": child,
"from": head_t,
"radius_head": round(head + (tail - head) * head_t, 4),
"radius_tail": round(tail, 4),
# Kept so an older runtime still gets a usable single radius.
"radius": round(tail, 4),
})
return out
def _cloth_neighbours(meshes, cloth_names):
"""bone -> {neighbouring bone: how strongly they share the same mesh}.
Two cloth bones are neighbours when the SAME VERTICES are weighted to both.
That is the only definition that matters here: a vertex driven half by one
skirt panel and half by the next is the piece of mesh that has to absorb any
difference between them, and linear-blend skinning absorbs it by pulling
itself apart. Measured with debug/cloth_stretch_check.gd, adjacent panels
taking drape shares of 0.85 and 0.48 stretched the edge between them to 3.3x
its rest length — an 80 mm hole in the front of the skirt, which is the
skirt "breaking" around the thigh rather than deforming over it.
Adjacency by NAME or by rest distance would both be guesses; the artist
already answered the question in the weights.
"""
shared = defaultdict(lambda: defaultdict(float))
for m in meshes:
gname = {g.index: g.name for g in m.vertex_groups}
for v in m.data.vertices:
here = [(gname.get(g.group, ""), g.weight) for g in v.groups
if gname.get(g.group, "") in cloth_names and g.weight > 0.05]
for a_name, aw in here:
for b_name, bw in here:
if a_name != b_name:
shared[a_name][b_name] += aw * bw
return {a: dict(d) for a, d in shared.items()}
def describe_rig(arm, roles, mapping, meshes):
"""Record what we worked out, so the runtime never re-guesses anatomy."""
names = set(b.name for b in arm.data.bones)
owned = _dominant_vertices(meshes, arm)
roles_out = {"hips": roles.hips, "head": roles.head, "neck": roles.neck,
"spine": list(roles.spine)}
for (role, side), name in roles.limb.items():
roles_out[f"{role}.{side}"] = name
driven = set(mapping)
chains = []
springy = {b.name: b for b in arm.data.bones
if b.name not in driven
and spring_class(b.name)}
neighbours = _cloth_neighbours(meshes, set(springy))
for name, bone in springy.items():
if bone.parent is not None and bone.parent.name in springy:
continue # not the root of a chain
# One chain per leaf path, so each strand solves independently.
stack = [[name]]
while stack:
path = stack.pop()
kids = [c.name for c in arm.data.bones[path[-1]].children
if c.name in springy]
if not kids:
tips = []
hulls = []
fallback = Vector((0.0, 0.0, -0.06))
for i in range(len(path)):
t = _bone_tip(arm, path[i], path, i, owned, fallback)
tips.append([round(t.x, 5), round(t.y, 5), round(t.z, 5)])
hulls.append(_hull_samples(arm, path[i], owned))
fallback = t
chains.append({
"class": spring_class(path[0]) or "cloth",
"root_parent": bone.parent.name if bone.parent else None,
"bones": path,
"tips": tips,
"hulls": hulls,
# Which other cloth bones share mesh with each of these,
# so the runtime can stop neighbours drifting apart. See
# _cloth_neighbours.
"neighbours": [neighbours.get(n, {}) for n in path],
})
continue
for k in kids:
stack.append(path + [k])
# Twist bones only — a hair link is also `X.001`, but it is cloth, and
# listing it here would have the twist distributor and the spring solver
# both writing the same bone.
twist = []
for b in arm.data.bones:
if b.name in driven or is_cosmetic(b.name) or not is_segment_of(b.name, names):
continue
if b.parent is not None:
twist.append({"bone": b.name, "parent": b.parent.name,
"child": b.children[0].name if b.children else None})
return {"roles": roles_out, "chains": chains, "twist": twist,
"colliders": _leg_colliders(
arm, roles, _dominant_vertices(meshes, arm, min_weight=0.6),
_body_points(meshes)),
"weights_authored": _weights_look_authored(meshes, roles),
"driven_bones": sorted(driven)}
def _weights_look_authored(meshes, roles):
"""Were these weights painted, or solved by a nearest-bone fit?
The runtime decides from this whether to run its destructive load-time
weight repair, so it is MEASURED rather than inferred from which pipeline
branch ran — a model that arrives unrigged still goes through autorig and
out through this same tool, and must not be handed a sidecar that says its
weights are fine when they are not.
Two signatures, both taken from the shipped-vs-source comparison that
started this rework: the nearest-four-bones fit left 16% of vertices pulled
by BOTH legs and gave 86% of them the full four influences, where the
artist's own weights had 0.1% and 26%.
"""
legs = {}
for (role, side), name in roles.limb.items():
if role in ("thigh", "shin", "foot", "toe"):
legs[name] = -1 if side == "L" else 1
def side_of(group_name):
for name, s in legs.items():
if group_name == name or group_name.startswith(name + "."):
return s
return 0
total = 0
bleeding = 0
four = 0
for m in meshes:
gside = {g.index: side_of(g.name) for g in m.vertex_groups}
for v in m.data.vertices:
wl = wr = 0.0
n = 0
for g in v.groups:
if g.weight <= 0.005:
continue
n += 1
s = gside.get(g.group, 0)
if s == -1:
wl += g.weight
elif s == 1:
wr += g.weight
total += 1
if n >= 4:
four += 1
if wl > 0.005 and wr > 0.005:
bleeding += 1
if total == 0:
return False
authored = bleeding / total < 0.02 and four / total < 0.5
print(f"Weights: {bleeding} cross-leg ({bleeding / total * 100:.1f}%), "
f"{four / total * 100:.0f}% at four influences — "
f"{'authored' if authored else 'solved, runtime repair stays on'}")
return authored
# ------------------------------------------------------------------------ main
def main():
clear_scene()
print(f"Importing character {CHARACTER}")
# Normalise emissive-albedo materials FIRST. Blender honours
# KHR_materials_unlit and reads only base colour, so an unlit model with its
# texture in the emissive slot imports with no images at all and exports a
# black silhouette — there is no node graph left to repair afterwards.
import_any(gltf_fix.prepare(CHARACTER, tempfile.gettempdir()))
arms = [o for o in bpy.data.objects if o.type == "ARMATURE"]
if not arms:
print("ERROR: character has no armature. Rig it first (see docs/3D_ASSET_PIPELINE.md).")
sys.exit(1)
arm = max(arms, key=lambda a: len(a.data.bones))
meshes = skinned_meshes(arm)
if not meshes:
print("ERROR: no skinned meshes bound to the armature")
sys.exit(1)
for o in [o for o in bpy.data.objects if o.type == "MESH" and o not in meshes]:
print(f"Dropping unskinned prop mesh: {o.name}")
bpy.data.objects.remove(o, do_unlink=True)
print(f"Character: {len(arm.data.bones)} bones, {len(meshes)} meshes "
f"({sum(len(m.data.vertices) for m in meshes)} verts) — weights kept as authored")
strip_import_suffixes(arm, meshes)
for a in list(bpy.data.actions):
bpy.data.actions.remove(a) # the character's own clip is not ours
roles = RigRoles(arm)
missing = roles.missing_core()
if missing:
print(f"ERROR: could not identify these bones on the character rig: {missing}")
print("Resolved so far:\n" + roles.describe())
sys.exit(1)
rebuild_hierarchy(arm, roles)
subdivide_cloth_panels(arm, meshes, roles)
unbind_cloth_from_legs(arm, meshes, roles)
flatten_and_scale(arm, meshes, TARGET_HEIGHT, roles)
fix_unlit_materials(meshes)
roles = RigRoles(arm) # rest positions moved; re-read
library = os.path.join(ANIM_DIR, "_library.glb")
if not os.path.exists(library):
print(f"ERROR: animation library not found: {library}")
sys.exit(1)
before = set(bpy.data.objects)
before_actions = set(bpy.data.actions)
bpy.ops.import_scene.gltf(filepath=library)
new_objects = [o for o in bpy.data.objects if o not in before]
src_arm = next((o for o in new_objects if o.type == "ARMATURE"), None)
if not src_arm:
print("ERROR: no armature in the animation library")
sys.exit(1)
src_roles = RigRoles(src_arm)
mapping = build_map(src_roles, roles, OVERRIDES)
print("\nLibrary rig:\n" + src_roles.describe())
print("\nCharacter rig:\n" + roles.describe())
print(f"\nDriving {len(mapping)} of {len(arm.data.bones)} bones from the library; "
f"{len(arm.data.bones) - len(mapping)} left free for secondary motion.")
yaw = facing_correction(src_arm, src_roles, arm, roles)
src_h = (src_arm.matrix_world @ src_arm.data.bones[src_roles.hips].matrix_local).translation.z
tgt_h = (arm.matrix_world @ arm.data.bones[roles.hips].matrix_local).translation.z
scale = tgt_h / src_h if src_h > 1e-5 else 1.0
print(f"Hips height: library {src_h:.3f} m, character {tgt_h:.3f} m (scale {scale:.3f})")
if not arm.animation_data:
arm.animation_data_create()
for track in list(arm.animation_data.nla_tracks):
arm.animation_data.nla_tracks.remove(track)
merged = 0
kept_clips = set()
for action in [a for a in bpy.data.actions if a not in before_actions]:
clip = LIBRARY_CLIP_MAP.get(action.name.split(".")[0])
if not clip:
continue
print(f" {action.name} -> {clip}")
baked = retarget_clip(src_arm, src_roles, arm, roles, mapping, action,
clip, yaw, scale)
add_nla_clip(arm, baked, clip)
kept_clips.add(clip)
merged += 1
if merged == 0:
print("ERROR: no clips retargeted")
sys.exit(1)
# Drop anything animated that we did not put here.
#
# Clearing bpy.data.actions before the library import is not enough: a
# character can carry its own clip in an NLA track that survives, and it is
# then exported alongside ours. `hikari` shipped two — " Girl|Loli Anime
# GirlAction" — which are keyed for a rig that no longer exists after the
# retarget, so they export as frozen rest-pose statues and the build fails a
# check that is right to fail it.
# Across EVERY object, not just the armature we retargeted onto. The export
# runs in NLA_TRACKS mode, so anything sitting in an NLA track anywhere in
# the file ships — including a second armature the model happened to carry,
# which is where hikari's two were hiding.
def _strip_foreign(ad):
if ad is None:
return
for track in list(ad.nla_tracks):
if not any(st.action and st.action.name in kept_clips
for st in track.strips):
ad.nla_tracks.remove(track)
if ad.action is not None and ad.action.name not in kept_clips:
ad.action = None
for obj in list(bpy.data.objects):
_strip_foreign(obj.animation_data)
if obj.type == "MESH" and obj.data.shape_keys:
_strip_foreign(obj.data.shape_keys.animation_data)
for a in list(bpy.data.actions):
if a.name not in kept_clips:
a.use_fake_user = False
bpy.data.actions.remove(a)
print(f"Kept {len(kept_clips)} clips; dropped everything else")
for o in new_objects:
bpy.data.objects.remove(o, do_unlink=True)
os.makedirs(os.path.dirname(os.path.abspath(OUTPUT)), exist_ok=True)
sidecar = os.path.splitext(OUTPUT)[0] + ".rig.json"
info = describe_rig(arm, roles, mapping, meshes)
with open(sidecar, "w", encoding="utf-8") as f:
json.dump(info, f, indent=2)
print(f"Wrote {os.path.basename(sidecar)}: {len(info['chains'])} cloth chains, "
f"{len(info['twist'])} twist bones, {len(info['colliders'])} leg colliders")
print(f"Merged {merged} clips. Exporting {OUTPUT}")
bpy.ops.export_scene.gltf(
filepath=OUTPUT,
export_format="GLB",
export_yup=True,
export_apply=False,
export_skins=True,
export_animations=True,
export_animation_mode="NLA_TRACKS",
# Export ONLY the bones the clips actually key. Baking every bone
# writes rest-pose tracks for the skirt and hair too, which triples the
# clip data and — worse — has the AnimationPlayer overwrite exactly the
# bones the spring solver is meant to own.
export_bake_animation=False,
export_optimize_animation_size=True,
# "keep_anim_armature" is what forces a track onto every bone whether
# or not the clip touches it. Off, so the skirt and hair export with no
# tracks at all and belong entirely to the spring solver.
export_optimize_animation_keep_anim_armature=False,
)
print("Done.")
def subdivide_cloth_panels(arm, meshes, roles, segments=4):
"""Split single-bone cloth panels into a CHAIN so they can bend.
A skirt panel modelled as one rigid bone off the waist cannot ride up over a
thigh, because riding up is a FOLD: the top of the panel stays at the waist
while the lower part lifts. One bone can only rotate the whole sheet about
the waist, so the hem swings forward while the middle stays put and the
thigh comes straight through it. No amount of collision or drape strength
fixes that — the degree of freedom does not exist.
Taila's skirt is 21 such bones (see rebuild_hierarchy: the panels export
flat, and their real 3-segment chains are not recoverable from the file).
So the segments are BUILT here: each panel bone gains `segments - 1`
children strung along the direction it actually covers, and the vertices it
drives are handed to whichever segment spans them. The runtime spring
solver already treats chains properly, so the lower segment can lift over
the leg while the top stays anchored.
SIX segments, not three. Magica Cloth 2's skirt guide: "if the bone position
deviates from the shape of the skirt, or if the bones are spaced too far
apart, the accuracy of collision detection will be significantly reduced."
That is measurable here — a collision is resolved by ROTATING a bone, which
moves a point at most twice its distance from that bone's head, and with
three segments the front panels had contacts sitting 10-48 mm from their own
pivot against a thigh 100 mm inside them. No amount of tuning can clear that;
the pivot has to be closer to the contact.
Runs before flatten_and_scale, in the source's own units.
"""
owned = _dominant_vertices(meshes, arm)
panels = []
for bone in arm.data.bones:
if not is_cosmetic(bone.name):
continue
cls = spring_class(bone.name)
if cls is None or cls == "hair":
continue
if bone.children:
continue # already a chain — the artist's own topology wins
pts = owned.get(bone.name, [])
if len(pts) < 12:
continue
rest = arm.matrix_world @ bone.matrix_local
tip = _bone_tip(arm, bone.name, [bone.name], 0, owned,
Vector((0.0, 0.0, -0.06)))
if tip.length < 1e-4:
continue
panels.append((bone.name, rest, tip))
if not panels:
return 0
bpy.context.view_layer.objects.active = arm
bpy.ops.object.mode_set(mode="EDIT")
eb = arm.data.edit_bones
made = {}
for name, rest, tip_local in panels:
parent = eb[name]
head = parent.head.copy()
step = (rest.to_3x3() @ tip_local) / segments
chain = [name]
prev = parent
for s in range(1, segments):
seg = eb.new("%s.seg%d" % (name, s))
seg.head = head + step * s
seg.tail = head + step * (s + 1)
seg.parent = prev
seg.use_connect = False
prev = seg
chain.append(seg.name)
parent.tail = head + step
made[name] = chain
bpy.ops.object.mode_set(mode="OBJECT")
# Spread each vertex across the TWO nearest segments, so the panel bends
# where the geometry is and the mesh stays continuous while it does.
#
# Handing every vertex to exactly one segment puts a hard cut through the
# sheet: the vertices either side of a segment boundary are then driven
# 100% by different bones, and the moment those bones rotate apart the mesh
# opens along that line. Measured with debug/cloth_stretch_check.gd over a
# movement sweep, an edge inside DEF-skirt.L.003 grew 270 mm — 11.7x its own
# rest length. That is the skirt visibly BREAKING around a thigh instead of
# stretching over it, and it looks nothing like a collision bug.
#
# Weighting to the two nearest segment CENTRES instead makes the transition
# linear, which is the same trick that stops a limb tearing at an elbow.
for m in meshes:
for name, chain in made.items():
vg = m.vertex_groups.get(name)
if vg is None:
continue
groups = [vg] + [m.vertex_groups.new(name=n) for n in chain[1:]]
rest = arm.matrix_world @ arm.data.bones[name].matrix_local
origin = rest.translation
axis = (rest.to_3x3() @ _bone_tip(
arm, name, [name], 0, owned, Vector((0.0, 0.0, -0.06))))
span = axis.length
if span < 1e-5:
continue
axis = axis / span
mw = m.matrix_world
for v in m.data.vertices:
w = 0.0
for g in v.groups:
if g.group == vg.index:
w = g.weight
if w <= 0.0:
continue
t = max(0.0, min(1.0, (mw @ v.co - origin).dot(axis) / span))
# Position measured in segment-CENTRE space: the ends of the
# panel sit on one segment outright, everything between is a
# blend of the two it lies between.
u = t * segments - 0.5
lo = int(math.floor(u))
f = u - lo
if lo < 0:
lo, f = 0, 0.0
elif lo >= segments - 1:
lo, f = segments - 1, 0.0
groups[0].remove([v.index])
if w * (1.0 - f) > 1e-5:
groups[lo].add([v.index], w * (1.0 - f), "REPLACE")
if f > 1e-5:
groups[lo + 1].add([v.index], w * f, "REPLACE")
print(f"Subdivided {len(made)} cloth panels into {segments} segments each")
return len(made)
def unbind_cloth_from_legs(arm, meshes, roles):
"""Take the LIMBS out of any vertex the cloth chains own.
A source model auto-weighted in Rigify leaves a band of skirt vertices
partly weighted to the thigh, and the vertices immediately next to them not
at all. Nothing about that is smooth: when the thigh swings, one vertex
follows it and its neighbour does not, and the mesh splits along the join.
Measured with debug/cloth_stretch_check.gd during a slide, two adjacent
front-panel vertices — identical weights except that one carried
DEF-thigh.L at 0.24 — pulled 80 mm apart, 3.3x their rest length. 0.24 of
the thigh's ~0.35 m of travel is 84 mm, so that is the whole of it. On
screen it is the skirt "breaking" around the thigh instead of stretching
over it.
Smoothing the field instead of deleting it does NOT work, and the arithmetic
says why: differential motion is the weight GRADIENT times how far the limb
travels. Holding an edge to under 10 mm against a thigh that moves 350 mm
needs less than 0.03 of weight difference across it, and this skirt's edges
are ~48 mm long, so the falloff would have to span most of the character.
Tried it — a 1.7-radius falloff put 0.24 on one vertex and 0.00 on its
neighbour and tore by 95 mm. Any leg weighting on this cloth tears.
Which is the project's own rule anyway: cloth moves by spring bones, never
by weights. Keeping the leg out of the weights is also what leaves the
spring solver able to move these vertices at all — see the note below on
bind_cloth_to_legs, which was this same mistake made deliberately.
"""
limbs = set()
for (role, side), name in roles.limb.items():
if role not in ("thigh", "shin", "foot"):
continue
limbs.add(name)
# ...and the twist segments hanging off them, which are the same limb.
for b in arm.data.bones:
if b.name != name and b.name.startswith(name + "."):
limbs.add(b.name)
cloth = set(b.name for b in arm.data.bones
if any(t in SPRING_CLASSES for t in tokens(b.name)))
if not limbs or not cloth:
return 0
freed = 0
for m in meshes:
gname = {g.index: g.name for g in m.vertex_groups}
for v in m.data.vertices:
cloth_w = 0.0
limb_w = 0.0
for g in v.groups:
n = gname.get(g.group, "")
if n in cloth:
cloth_w += g.weight
elif n in limbs:
limb_w += g.weight
# The cloth has to actually own the vertex. A thigh vertex with a
# little skirt on it is a LEG and must keep following the leg.
if limb_w <= 0.0 or cloth_w < 0.5 or cloth_w <= limb_w:
continue
scale = (cloth_w + limb_w) / cloth_w
for g in list(v.groups):
n = gname.get(g.group, "")
if n in limbs:
m.vertex_groups[g.group].remove([v.index])
elif n in cloth:
m.vertex_groups[g.group].add(
[v.index], min(1.0, g.weight * scale), "REPLACE")
freed += 1
print(f"Freed {freed} cloth vertices from the legs they were skinned to")
return freed
# Cloth is NOT skinned to the legs.
#
# There was a bind_cloth_to_legs() here that gave every cloth vertex near a
# thigh a share of that thigh, so the skirt would ride the leg the way a real
# one does. It was written because the drape and collision in
# characters/spring_bones.gd appeared to do nothing however hard they were
# driven — and they did nothing because of two bugs since fixed there (the
# drape rotated the spring's target but not the bone, so it cancelled out at
# equilibrium; and the collision's answer was read back off the bone tip, which
# discards the twist that is most of what lifts a sheet off a leg).
#
# With those fixed the binding is actively harmful. A vertex weighted 0.9 to a
# thigh cannot be moved by its own cloth bone, so the solver loses the authority
# to push it out of the leg — and the leg still overtakes it, because 0.9 of a
# rotation always lags the surface doing 1.0 of it. It also poisoned the
# collider measurement below: 2258 skirt vertices counted as thigh geometry and
# fitted a 0.28 m thigh.
#
# Measured on the run cycle with debug/skirt_clip_view.gd, which paints cloth
# magenta so leg-in-front-of-cloth is unambiguous: removing it made the frames
# it was supposed to fix cleaner, not worse.
main()