Lead: the exe’s character
shaders upload cS0_VEC = (axis,
0.5), i.e. k = n·a + 0.5; the stage shaders’
upload was not found in their draw path (FORMATS
“Constants from the executable”).
(sbtsspec) The SBTS family’s weight is
decoded (pixel cS1_COLp.w = 0.5,
0x1402a78df) and used as such; only SD/SBPX
remain inferred
Now:cEYEVEC.w
decoded (1 + 0.01 × the material’s s16 at 0x40,
pdconv pd_eye_w); REFP
cTMAT0 decoded
(floorstain) (s16s at 0x44/0x46/0x48,
pdconv pd_tmat; upload
0x1402a5a30). Still ours: c0.z = 2
and the view-space sphere map of REFS, and the
vertex c0.w = 0.5 of the REFP
projection (the same pack constant as the SBTS
row below). REFP’s model space = the buffer’s
joint space (pdconv pd_refp_xf)
(refpflicker): INFERRED from the stage
vertex buffers being joint-local (the upload’s
model matrix, the first pointer of its draw
argument, not traced to the joint)
Where:pd_data.gd
Lead: the TMAT uploads of
the REFS classes; the vertex c0 upload of the
REF classes; the model matrix passed to the REFP
upload (0x1402a5a30)
SBTSREF* bump
offset c21.y/.w
Status: INFERRED
(floorstain)
Now: No offset: the bind
sets c21.x = c21.z =
the material’s s16 at 0x42 / 255 (pdconv
pd_bumpenv; 0 on every SBTSREF*
material of all 19 stages), so the offset is 0
if .y/.w are 0. They
are render states 0x17/0x19 of texture stage 2,
reset to 0 at init and otherwise set only by
other draw paths (left over from the previous
draw)
Where:pd_data.gd PRELIT_CODE (sbts
2)
Lead: which draws set
stage-2 states 0x17/0x19 before a SBTSREF* draw
(0x1402365c4 and similar)
SBTS normal
map signed flag
Status: INFERRED
(sbtsspec)
Now:rgbsigned.y (c18, REF pixel shaders
only) taken as 0: normal = (rgb − 0.5)·2 once,
as plain SBTS always does
Where:pd_data.gd PRELIT_CODE
Lead: the c18 upload of the
SBTSREF* classes (not found in their draw
0x1402a0b20)
SBTS
tangent-space pack scale c0.w
(vertex)
Status: INFERRED
(sbtsspec)
Now: 0.5: the vertex shader
packs the hemisphere and point vectors as x·c0.w
+ c0.w, the pixel shader unpacks (x −
0.5)·2
Where:pd_data.gd PRELIT_CODE
vertex()
Lead: the SBTS classes’
vertex c0 upload
SBTS output
alpha scale
Status: INFERRED
(sbtsspec)
Now: The cut-out keeps a
> 0.75 like the other stage shaders; the
game’s output alpha (4a − 3)·(sat(ramp.b)·c0.z +
c0.w), which weights the shadow pass, is not
reproduced (pixel c0 = (0.8, 0.8, 0.8, 1/124),
so its scale is decoded but unused)
Where:pd_data.gd PRELIT_CODE, PdShadow
receiver
Lead: wire the ramp’s blue
channel into the receiver’s alpha
SBTS
binormal = cross(N, T)·w instead of the stored
binormal
Status: INFERRED
(sbtsspec)
Now: The vertex frame uses
B = cross(N, T) × the stored side (glTF
TANGENT.w, Godot’s BINORMAL); the game’s vertex
shader reads the stored binormal (format 0x03,
bytes 28..31). They agree except on the
near-zero frames (T = B = 0 either way) and
about 20 unit frames per stage on st04, st14 and
st24 that are skewed off cross(N, T) (pdconv
--stats: 21 each)
Lead: pass the stored B
through as an extra vertex channel (e.g.
TEXCOORD_2/3 → Godot CUSTOM0) and use it in
vertex()
SBTS zero
stored normal
Status: INFERRED
(sbtsspec)
Now: 1–7 % of SBTS vertices
store N = T = B = 0 (st03: 1526, 673 triangles
all-zero); glTF gets zeros, Godot’s import makes
N a unit vector, so those corners are lit with
that N where the game’s tangent-space dots give
0
Lead: write a flag (e.g.
COLOR alpha or CUSTOM) for a zero normal and
zero the light vectors in vertex(); check the
game’s look on those triangles
SDLYFOG glow
term (st01)
Status: MISSING
(lightextras)
Now: Not drawn
Lead: the TMAT1 upload of
the SDLYFOG class
Cloud scroll
rate (skyfix5)
Status: DECODED
(skyglow 11-13 built)
Now: Built: UV offset =
velocity × the launch-time 30 Hz tick counter
(PdClock, whole steps) × 1e-5, wrapped to [0,
1); the counter update is inside the 30 Hz game
tick 0x1401c7630 (one step per tick, from boot,
menus included). PdSky updates the
offsets every frame
Lead: PC draws none (FOG0
unread in all 69 PS); removed. Xbox-look fog is
a separate, later question
Glow (bloom) pass
(bloom)
Status: INFERRED
Now: Decoded and drawn:
bright pass s·dot(c, (0.3, 0.6, 0.1)) > 0.5
into 256², 4-tap reduce to 64², 2×4+2×4-tap
blur, additive composite ×1.0 over the screen; s
= the stage script’s 0x2003 f eased from 0.5
(FORMATS “Glow pass”). INFERRED: the frame the
glow reads is a second camera’s render of the
same world (the game copies the back buffer;
same pixels, not the same mechanism); taps past
a tile’s edge clamp to the edge (the game reads
the neighbouring texels of its shared 512²
targets: never-written texels of slot 0, the
other tile of slot 1)
Where:pd_glow.gd
Lead: what slot 0/1 hold
outside the tiles (clear at creation?
0x14028de90). (skyglow 11-13) st02’s
white sky is not the glow: the chain matches a
CPU model of the decoded passes (0.014 mean);
the input is wrong (rows “Sky dome vertex
colour”, “Sky light set”)
Glow scale
easing rate (bloom)
Status: INFERRED
Now: One 10 % step per 1/30
s game tick
Where:pd_glow.gd TICK
Lead: who calls the update
0x14019a240 (vtable method, no direct caller)
and how often. (skyglow 11-13)
0x14019a240 is slot 3 of the battle light
class’s vtable (0x14047b530); the tick’s
256-object loop calls slot 5, so the caller is
still open
Glow
paths not modelled (bloom)
Status: MISSING
Now: Not drawn: the
composite at 0.125 in game state 0xe
(0x140530264), the view variant 0x140288b00 (bit
3 of 0x140530034), the 0x2003 colour (stored at
0x14054c2e8, no reader found), the Xbox SelColor
vertex program’s tap placement (Xbox object
0x1403fc570; the PC shader samples both taps at
one point)
Where:pd_glow.gd
Lead: the addresses named.
(skyglow 11-13) leads only: bit 3 of
0x140530034 is cleared at 0x14012bf97 (battle
flow), its setter writes the whole word; game
state 0x140530264 is written by 0x1401d8430
only
Sky
dome vertex colour (SC, format 0x09)
(skyglow 11-13)
Now: Built: pdconv writes
the u8 × 4 colour at +16 of the 20-byte format
0x09 vertex as COLOR_0 (all 19 stages
reconverted); sc and vkraylight
draw texture × vertex colour; (skycol2
r2)scal too, now decoded and
drawn unlit: texture × the vertex colour clamped
to [−1, 1], alpha = texture alpha × vertex
alpha, one / one (FORMATS “Stage shaders”,
scal); until r2 it went through the
lit path with set 2’s point light. INFERRED:
vkblink (decoded as texture ×
constant) and vksand (undecoded,
see vkSAND below) ignore the colour
(NO_VCOL_SHADERS)
Lead: vkSAND_ps.cso:
whether it reads the vertex colour (its alpha
runs 0..1 on st02)
Sky light
set (skyglow 11-13)
Status: DECODED (H PC
code 0x1401c87f0; skyglow 11-13 built)
Now: Built: stage model 2
(SC, SCAL, SVC, SVCAL, SBTSCLOUD3) is drawn
under light set 2, model 1 under set 3
(PdData.model_light_set). pdconv
writes each set’s point lights 0 and 1 (0x2005)
as points in light.json; for sets
1..3 light slot 2 is always the set’s point
light 0 (0x140199970), so apply_prelit gives it
to every set 1..3 material; the rows below say
which of them read it
Set
2 point light: the cloud layer
(sbtscloud3) (skycol
11-15)
Status: DECODED /
BUILT
Now: The cloud shader’s
point term (FORMATS “Sky colour: set 2’s point
light”) saturates over most of the dome (st03 no
longer purple). The Xbox look clamps the packed
point vector to [0, 1] (H since skybright 11-15:
the Xbox vertex program writes it through a
clamped colour output; since skycol2 the lower
bound is the material’s
cloud_p_low,
CLOUD_P_LOW_PC /
CLOUD_P_LOW_XBOX)
Where:pd_data.gd cloud shader
Set
2 point light: the unlit sky shaders
(sc, scal,
svc, svcal)
(skycol2 r2 11-15)
Status: DECODED /
BUILT
Now: None of their shaders
has a light constant (SC / SCAL: texture ×
vertex colour; SVC: vertex colour), so the point
light is uploaded but never read; all four are
drawn unlit (UNLIT_SHADERS,
vcol_sky_material). These five
shaders are all of model 2 on the 19 versus
stages (test_skycol check 6)
Where:pd_data.gd UNLIT_SHADERS
Sets
1 / 3 point light: the far scenery
(sd, sbts,
sbtsrefs on model 1) (skycol2
r2 11-15)
Status: DECODED
Now: They read slot 2
through the SD / SBTS point term, but set 3’s
point light 0 (and set 1’s colour) is black on
all 19 versus stages, so it adds nothing
Where:pd_data.gd apply_prelit
Set
2 point light on the Xbox (skycol
11-15)
Status: INFERRED (M)
Now: The Xbox is taken to
select the same point light for the sky draw
(identical stage scripts, footage); its light
slot code is not traced
Lead: default.xbe’s light
slot set-up
St03
sky brightness against the Xbox (skycol
11-15)
Now: The missing half was
the sky model’s opaque svc backdrop
dome, (76, 80, 102) over the upper sky, which
pdconv dropped (vertex format 0x0B). Built with
the next row; with it and the point light the
st03 sky box is compared against the xemu pan in
the skycol2 REPORT (a check, no tuning). Cloud
pattern phase of the footage not matched
(statistics only)
Now: pdconv converts the
0x0B layers of stage model 2 (position +
COLOR_0, pd_shader,
pd_flags, pd_blend,
index buffer order; 16 of 19 versus stages, st02
/ st13 / st22 unchanged byte for byte). Godot
(PdData.vcol_sky_material) draws
them as saturate(vertex colour), no texture /
light / fog, back faces culled (st03’s inner
dome hidden), no depth write, opaque (flag bit 4
clear) or one / one (blend word 0x11). Godot
detail, not a game value: every layer is drawn
in the transparent pass (ALPHA = 1, the opaque
ones blend_mix at the lowest render priority, so
they come before the clouds) because the
compatibility renderer writes depth for
opaque-pass materials even with
depth_draw_never. Blend words other than one /
one are not in the data and are not drawn
(error)
Where:pdconv.cxx,
pd_data.gd
Sky
backdrop fog on the Xbox (SVC fog
coordinate) (skybright 11-15)
Status: INFERRED (M)
Now: Xbox SVC
writes dot(position, c16) as the fog coordinate,
not the depth; the c16 upload is not found.
xemu’s cloud gaps read the dome’s own colour, so
the backdrop is drawn unfogged in both looks
(skycol2: the sky layers’ material never takes
the Xbox fog; XboxLook.FOGGED still
lists svc / svcal for any other use)
Where:pd_data.gd vcol_sky_material
Lead: the SVC
class’s constant upload in default.xbe /
PDUWP.exe
Far
scenery light set and fog (stage model 1)
(skyglow 11-13)
Status: DECODED (H PC
code; skyglow 11-13 built)
Now: Built: model 1 is lit
with set 3 and, in the Xbox look, fogged with
set 3’s fog
Where:pd_data.gd,
xbox_look.gd set_fog
Lead: which other stages
have a model 1 is not listed
Sky
model placement (skyglow 11-13)
Status: DECODED (H;
skyglow 11-13 built)
Now: Built:
PdSky moves the model-2 nodes to
the drawn camera position every frame (all three
axes, rotation untouched), after the camera
moved
Where:pd_sky.gd,
main.gd
SC
dome blend state (skyglow 11-13)
Status: INFERRED
Now: Every sc
surface is drawn additive; the material decides
(flag bit 4): st02’s outer dome and st22’s dome
are opaque, st02’s inner dome and st07’s are one
/ one. Same picture over the black clear colour
as long as the opaque dome draws first
Where:pd_data.gd ADDITIVE_SHADERS
Lead: decoded; the render
states 0x40 / 0x3C / 0x3B around the sky draw
are not
Now: GFilter2 and AuraBox
are rebuilt for the character aura glow only
(aura_glow_look.gd, FORMATS
“Character aura glow” 4a: tap tables, blend
words, composite). FFilter (no user) and
PreRender are not drawn
Now: Only used when a stage
has no light.json (none now)
Where:pd_data.gd STAGE_LIGHT
Stage cull mode
Status: INFERRED
(shadows)
Now: Stage surfaces drawn
double-sided, lit with the stored normal like
the game’s vertex shaders. Cull state decoded (0
none, 0x900 front, 0x901 back; back is the
default), and the scene pass that binds the SBTS
sampler slots sets none; the stage draw itself
is not traced. The st02 “growing” patch was our
camera-facing normal flip (now only with
--flip-normals)
Where:pd_data.gd PRELIT_CODE,
flip_normals
Lead: trace the stage model
draw’s cull state (FORMATS.md “Stage cull
mode”)
Character
shadow volumes
(sshadowch/sshadow on
characters)
Status: INFERRED
(shadows, narrowed by palacefix)
Now: Decoded volumes and
extrusion drawn; the stencil count is
approximated: per part, back faces mark and
front faces clear on depth fail (Godot material
stencil can only REPLACE one value), so a later
part can clear an earlier one’s mark, and one
part made of several closed pieces loses a
piece’s mark where the view ray leaves the
surface through another piece (camera
dependent). Static props no longer use it: their
volumes are extruded once on the CPU like the
vertex shader does and counted per pixel with
the stage volumes (exact z-fail count, camera
independent; --stencil-props draws
them the old way)
Lead: exact needs incr/decr
stencil (RenderingDevice, not in
Compatibility)
Stage
shadow volumes (sshadowc)
Status: INFERRED
(shadows)
Now: Drawn: the stage
shader counts the volumes per pixel along the
light (equal to the game’s z-fail count for
closed meshes, FORMATS.md “Stage volumes”), with
a fill rule on shared edges and the orientation
checked per connected part. The same count
darkens characters (their shader) and sdal
blended layers. Inferred: a volume face level
with the surface doesn’t count (point moved
STAGE_EPS = 2 cm along the light,
away from it: as with a LESSEQUAL depth test
wherever the surface faces the light, and
camera-independent; on surfaces facing away from
the light a level face may resolve the other way
(darkening weight 1/124 there); LESS would move
it toward the camera: 0.2–0.4 % of the pixels
differ); an sdal layer leaves the destination
alpha of the surface below (its blend state
isn’t traced), so the layer takes that surface’s
weight; additive layers aren’t darkened
Lead: the exe’s depth
function during the volume pass (render state
0x8F/depth func); the blend state and colour
write mask (0x43) of the sdal and additive
draws
Back-face
threshold of the volume vertex shaders
(c0.x)
Status: INFERRED
(shadows)
Now: 0
Where:pd_shadow.gd BACK_THRESHOLD
Lead: where the exe fills
the volume shaders’ c0
Length
of the light push in sshadowch
(cDL_VEC.xyz)
Status: INFERRED
(shadows)
Now: 1 unit (0.1 m)
Where:pd_shadow.gd CH_BIAS
Lead: source vector
transformed in 0x1402a3bd0
Light
direction of the volumes
Status: INFERRED
(shadows)
Now: The stage light set’s
light_dir, in the game’s order
(capsulelook-fix: x and z were swapped
before)
Now: Stage volumes darken
characters by SWBOS’s decoded output alpha
(4a − 3) · (c0.z · sat(ramp.b) + c0.w)
(FORMATS.md “Shadow volumes”), with
max(N·L, 0) standing in for the
toon ramp’s blue channel and the stage shaders’
c0 (0.8, 1/124) assumed for SWBOS
too. Character and prop volumes
(sshadowch/sshadow)
still only darken stage materials
Where:pd_data.gd
apply_character_light
Lead: the ramp texture
(slot 3) and SWBOS’s c0 upload; a
receiver pass on characters for the stencil
volumes
Shadow
colour also used as the background
Status: DECODED / BUILT
(skycol2 11-15); narrowed
(r2)
Now: Light script 0x2001 is
the shadow colour (decoded). Every clear call in
the exe (14 calls of 0x140021620) passes a
constant colour, none the stage’s: the PC battle
frame clears to black (colour 0 at 0x1401c81d3,
H); main.gd clears with
PdData.frame_clear_color() = black
whenever the stage has game light (the fitted
fallback keeps its default background). Open:
that call clears the colour target only while
bit 0x800 of the flag word 0x140530034 is set
(row below)
Where:main.gd,
pd_data.gd
St13
sky against footage (black clear behind the
additive dome) (skycol2 r2 11-15)
Status: INFERRED (open
check)
Now: St13 has no backdrop
layer and an additive sc dome, so
the clear colour shows through its whole sky.
The black clear (was the 0x2001 colour (0.071,
0.102, 0.094)) moved its Xbox-look sky box from
(0.68, 0.67, 0.41) to (0.55, 0.51, 0.26),
away from the only footage, an
xemu frame of the st13 intro (bright yellow,
another camera). Not unchanged. Decoded: no
clear in the exe uses a stage colour. Not
decoded: whether the battle frame clears the
colour target at all (the clear at 0x1401c81d3
needs flag bit 0x800, set at 0x1401aec4b /
0x1401d19f7 / 0x1401d1d33 / 0x14021846d; if not,
what the target holds before the sky is
drawn)
Where:main.gd, pd_data.gd
frame_clear_color
Lead: the states behind
flag bit 0x800; footage: PC
build frames of st13’s intro pan (and
an xemu frame at the same camera)
Now: With the backdrop the
sky box reads 72 % (st14) / 50 % (st04) / 59 %
(st24) clipped, clouds barely visible (PC look).
Split by renders (skycol2 r2 REPORT): backdrop
alone 0.51 / 0.38 / 0.41 mean (decoded data
colours, opaque on st14), the clouds add about
0.1–0.2, the glow pass (stage glow scale 1.0,
decoded) takes the clipped share from 9–13 % to
50–72 %. Not the point term: st14 has no set-2
point light, and without it st04 / st24 drop by
only 0.03. No decode error found; no footage of
these stages
Where:pd_data.gd vcol_sky_material,
pd_glow.gd
Lead: footage: PC
build frames of the st14 / st04 / st24
intro pans; then the glow chain’s input on
bright skies
Violet
specks at the st03 cloud edges (skycol2
11-15)
Status: INFERRED (open
item)
Now: Still there in the PC
look (6.9 % → 3.2 % of the frame toward the
point light, pale lavender (187, 184, 236) over
the backdrop); Xbox look 0.19 %. Source not
proven: lead = cloud texels whose perturbed
normal faces away from the point light keep the
set-2 light colour (0.68, 0.59, 1.0). The xemu
frames show faint violet fringes too
Where:pd_data.gd cloud shader
Lead: the clouds’ point
term per texel against a PC build frame toward
set 2’s point light
Character
aura glow (was “aura shells
(saura)”)
Status: decoded, draw
INFERRED in parts (auraglow build
11-14)
Now: Per fighter a glow
record (AuraGlow, rules
PdAuraRules = the aura script)
stepped every tick: colour by school, strength /
size by bucket min((level + 3) / 4, 5), the cast
flare (event 10, 140 → 53, 7.9 → 7.03) and the
aura card flare (event 8, 210 → 65, 6.5 → 5.05)
over 30 ticks, then the level values in one
step. Drawn (AuraGlowLook): the
character’s own skinned meshes as a silhouette
in the glow colour into a 256 view (the decoded
tile projection), 2× reduce to 128, the decoded
8-tap Gaussian blur in the decoded pass order
with strength / 255 of the previous frame’s tile
kept on the third pass (frame feedback, r2),
added to the frame squared, not over the body
(full-resolution body mask, r2). The client
derives it from the snapshot level and the cast
/ pickup events (schools sent with START,
NetSnapshot v22)
Lead: the INFERRED draw
parts are in rows “Aura glow draw: mask,
composite depth, frame rate”, “Aura glow
school”; Xbox build not checked
Aura
glow hidden by status slot 8 / alpha
(auraglow 11-14)
Now: The level is fed as 0
while status slot 8 is set, the fighter is dead
or out (Combatant.glow_level()), so
only a flare draws then. The record alpha is
fixed at 255: the character fade code is not
modelled
Where:aura_glow.gd alpha = 255
Lead: record alpha +0x110
from the character fade code
0x14024f3cc–0x14024f499 (L: which fades)
Aura
glow school (auraglow build 11-14)
Status: INFERRED
Now: The glow colour’s
school = the most common school among the
fighter’s data skills (0 = grey when none)
Where:arsenal.gd glow_school()
Lead: the school setter
0x140250480 is called from 0x14012d780 with a
value read through a pointer, not followed
(FORMATS “Character aura glow” 4a)
Now: Decoded in r2 and no
longer inferred: the 256 → 128 reduce, and T =
blur + T(previous frame) × strength / 255 (frame
feedback, slot 1 cleared only at the match
reset), pass 1 replacing. Still inferred: (1)
the per-pixel per-player stencil (glow not over
the own body) is replaced by a
full-screen-resolution mask: the character’s own
meshes in the scene (an extra render layer bit
per player slot) seen alone by a copy of the
frame’s camera at the frame’s interpolated
transforms (one target at the window size,
allocated once; auraglow3; drawn from the
originals, so with their alpha cut and exactly
where the frame draws them, auraglow4), so the
mask also covers body parts hidden by scenery;
(2) the composite has no depth test and is a
screen rectangle at the character’s depth; (3)
the GFilter2 luminance threshold and the add
passes’ alpha-test reference are not modelled
(taken as 0); (4) the glow is not in the bloom’s
input; (5) the chain runs once per drawn frame
(the PC build’s rate, per frame or per 30 Hz
tick, not checked), so the feedback gain depends
on the frame rate; (6) T is cleared at a
character reset (event 0) in place of the
per-match reset; (7) the silhouette shader’s
colour scale (M: the glow colour as is;
auraglow3: at the PC camera distance ours is
about half the PC footage’s glow area and peak,
and the composite has no extra gain, so this
scale is the lead; auraglow4: the colour upload
is decoded as byte / 255 into the character-draw
interface’s pixel constant 0, no doubling there,
but which shader output fills the silhouette
tile is not followed, FORMATS 4a “Silhouette
colour upload”); (8) the game’s player slots
0..3, which decide whose hidden tile keeps
decaying and whose stays frozen (decoded: the
chain runs for the players up to the highest
drawing one), are taken as the order the bodies’
glow draws were created; (9) test only, not the
draw: the GPU test’s “rim 1-2 px outside the
drawn body lit ≥ 90 %” is judged at the play
camera only; at the PC distance (5.84 m, body
~263 px) the same alpha-cut fringes of (1) leave
~12 % of the ring unlit (87.7-87.8 %, auraglow3
and r2), no decoded threshold exists for that,
so the test prints it as a NOTE there
(gpu_auraglow.gd with
--eye). Decoded in auraglow3 and
built: T is never cleared when a glow comes back
(a hidden glow’s T decays or stays frozen by
that rule; which body decays and which stays
frozen depends on the slot order of (8))
Lead: 0x14019ac40 (match
reset clear), the AuraBox box geometry
0x140289140, the filter shader’s threshold
constant, the stencil writes near 0x1402a6091;
the silhouette colour: 0x14029df00 (upload,
decoded), the material flag 0x800 loop
0x14019bb20 and the shader output that fills the
tile (not followed); the player slot of each
fighter
Capsule
pickup cue gate (auraglow build
11-14)
Status: INFERRED
Now: Cue 0x56 is played
positional at every fighter that takes a capsule
(the host plays it, the client plays it from
ORB_TAKEN)
Where:orb.gd
took()
Lead: slot-fill 0x140261bb4
plays it only when the character’s +0x10 word is
0 and the caller asks; which callers ask
(M)
Levitate
float animation (auraglow 11-14)
Status: MISSING
Now: Levitate is refused
(row “Special skills”)
Lead: decoded (M): while
status slot 11 is set, slot 44 with the stick
held, 43 without, 5-tick blend
(0x14025b170)