nothing but datacenters the record

treatment record

glitch-03

The same CV layer and glitch matrix as glitch-02, composited the other way round: order 'over' draws the HUD onto the already-displaced plate, so contours and id boxes sit cleanly on top instead of tearing with the image.

status
test pass
upstream
pass-02 frames 1–1,440
evidence
A lossless frame sequence was written and manifested; the video is derived from it.
frame 1 · 0:0021,600 · 15:0043,200 · 30:00
glitch-031–1,440

Its window on pass 2's full 43,200-frame axis.

what it looks like · 12 frames

This treatment kept no sampled frames of its own, so these are the source frames it consumed — pass-02 1–1,440, before the treatment was applied.

departs from the baseline

Differs from glitch-02 by exactly one parameter - composite 'over' rather than 'under'. Source hash, detections, displacement layers and overlay sizes are identical, which is what makes the pair a controlled comparison rather than two different renders.

Upstream

pass
pass-02
frames
1–1,440
source
passes/pass-02/previews/pass-02.sbs.f000001-001440.mov
source sha256
f11ed0d68751

Render

composite
over
device
cpu
frames
1,440
clip length
1 m 00 s
time to render it
17 m 30 s
chunking
160-frame windows, 9 of them, 124-frame lead-in discarded
overlay draws
box, mask
overlay hold
1 frame
stroke / bracket / label floor
5 / 56 / 4,800 px²

Displacement layers

gridmax offsetanglesparsityjitterholdstaggerseedenvelope
4×33290°0.50.880.65none
8×7220.80.860.59none
20×18140.930.840.813none

What the detector found

model
yolo11n-seg.pt
confidence floor
0.15
detections
675
per frame
0.469
labeldetections
bench150
train62
person62
refrigerator37
umbrella32
airplane32
car31
suitcase25
boat24
kite23

Outputs

fileframesgeometryruntimebytessha256
previews/glitch-03.hud-over.f000001-001440.mp41–1,4401536x13441 m 00 s294.4 MB4c6b85516f51

Named and hashed, not hosted. Nothing in this repository is a master.

Frame sequence

directory
frames/
files
1,440
bytes
2.0 GB
geometry
1536x1344
lossless
yes
manifest
frames-manifest.json

The treatment's record, the same way a pass's frames are its record. preview.mp4 is derived from these, not the other way round — extracting frames back out of it returns the codec's opinion of the pixels.

Provenance

no task id

This treatment was run as a local process rather than dispatched through the mesh, so it has no task id, no lease and no event trail. That is a gap in the record, and it is recorded as one.

Note

Judged against glitch-02 and NOT preferred: the flat overlay reads as an annotation laid over the piece rather than something inside it. Kept because the comparison is the evidence for that, and a rejected variant with no artifact is an opinion.