treatment record

glitch-envelope-02

Time-varying glitch with an AUTHORED cut list. Long stretches carry the CV overlay — contours, skeletons, id boxes — over an essentially clean image at a 0.04 displacement floor, and bursts rise and decay over about 1.5s (cut-half-life 26).

status: preferred
upstream: pass-02 frames 1–3,600
evidence: Video only. No frame sequence was written, so the encode is the record.

frame 1 · 0:0021,600 · 15:0043,200 · 30:00

glitch-envelope-021–3,600

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 1–3,600, before the treatment was applied.

departs from the baseline

glitch-envelope-01 detected its cuts automatically, which followed the render's OWN churn: at peak saturation the source sits at denoise 0.77 with a high frame diff, so it fired 52 cuts after frame 2,400 with a median shot of 8 frames and 76 of 97 shots under a second. The envelope never decayed and the densest passage was torn apart entirely — the one place the image most needs to be legible. Here the 14 cuts are placed: the eight chain breaks stay, because they are real discontinuities in the material, and through peak saturation the spacing is WIDER not narrower, ~220 frames (9.2s). The longest calm in the piece, 470 frames / 19.6s, sits at 1,969-2,439 as saturation arrives, so the dense frame gets an uninterrupted look before chaos punctuates it.

Upstream

pass: pass-02
frames: 1–3,600
source: regenerable: hstack of passes/pass-07-rehearsal/{frames,text} 1-3600 (image left, text right)
source sha256: be311e6eb26b

Regenerating the source

ffmpeg -framerate 24 -start_number 1 -i frames/frame_%06d.png -framerate 24 -start_number 1 -i text/frame_%06d.png -frames:v 3600 -filter_complex '[0:v][1:v]hstack=inputs=2' -c:v prores_ks -profile:v 4 -pix_fmt yuv444p10le source.mov

Render

composite: under
device: cpu
frames: 3,600
clip length: 2 m 30 s
time to render it: 21 m 15 s
chunking: 160-frame windows, 23 of them, 124-frame lead-in discarded
overlay draws: box, mask, skeleton, face_mesh
overlay hold: 1 frame
stroke / bracket / label floor: 5 / 56 / 4,800 px²

Displacement layers

grid	max offset	angle	sparsity	jitter	hold	stagger	seed	envelope
4×3	32	90°	0.5	0.8	8	0.6	5	cut
8×7	22	0°	0.8	0.8	6	0.5	9	cut
20×18	14	0°	0.93	0.8	4	0.8	13	cut

What the detector found

model: yolo11n-seg.pt + yolo11n-pose.pt
confidence floor: 0.15
detections: 5,361
per frame: 1.489

label	detections
person	3,524
train	485
refrigerator	367
car	168
bench	149
suitcase	90
clock	65
potted plant	65
oven	57
tv	48

Outputs

file	frames	geometry	runtime	bytes	sha256
`previews/glitch-envelope-02.envelope-authored.f000001-003600.mp4`	1–3,600	1536x1344	2 m 30 s	740.7 MB	`9646f9df2bd6`

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

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

Rendered locally, not through glitch_treatment: the render node was mid pass-08 and the capability exposes no envelope flags, so adding them would have needed a node restart that killed a 53-hour render. No task id or lease. The cut list was written by hand into the cache path ensure_cuts() reads, which is why it survived instead of being re-detected.

← glitch-envelope-01 all treatments