◐
9 of 10 checks passed — awaiting your spot-check
Automated layers are green. Nothing ships until you watch the first 30 seconds.
9/10checks passed
run #1 · 1.4s
meridian-2026-10-13-feedback-loop-part-1.mp4
3:58 · 1920×1080 · h264 + aac
generated 2026-10-13 14:05 UTC
Layer 1 — deterministic checks 5/5 ✓
✓
Stream layout uniformity
ffprobe every clip in the compose chain — identical codec, sample_rate, channels
Catches: players silently dropping audio after a scene boundary.
✓
Audio presence at 10 evenly spaced timestamps
volumedetect at t = duration × k/10 — assert mean_volume > −40 dB everywhere
Catches: dead segments, missing music bed, silent scenes.
✓
Caption single-line integrity
PIL ink-bbox height on every caption PNG — assert single line
Catches: subtitle overflow — long coaching sentences are the usual trigger.
✓
Chunk word-boundary audit
every chunk in scenes.json — no start/end mid-word fragment
Catches: TTS speaking truncated words mid-sentence.
✓
Deliverable inventory
mp4, srt, thumbnail, description.txt, qa-report.md all exist and non-empty
Catches: shipping a package with a missing piece.
Layer 2 — vision checks on the final file 4/4 ✓
✓
First and last frames
frame at t=0.5 and t=duration−0.5 — “describe what you see; anything unexpected?”
Catches: a stray default frame shipping as the video intro.
✓
Scene boundary frames
each scene start +0.5s and end −0.5s — vision-checked against the storyboard
Catches: wrong text frame, stale card from a previous render.
✓
Brand presence per scene
one frame per scene — “is the Meridian monogram present, copper rule correct?”
Catches: brand mark missing or wrong color after a kit change.
✓
Caption text readback + small-mark crispness
read caption text at 5 timestamps vs the SRT; inspect logo strokes at native scale
Catches: truncation, timing drift, pixelated corner icon.
Layer 3 — your spot-check (4 quick looks only a human can do; automated checks cannot judge pace, tone, or “something feels off”) 0/4
✓
Watch the first 30 seconds at playback speed
✓
Listen through the scene 3 → 4 transition
✓
Confirm the framework list reads Anchor · Ask · Align · Act on screen
✓
Anything that makes you pause twice becomes a scripted check next time
Publishing package — what you receive R-DELIV
Feedback Fails Late — publishing packageready after spot-check
Title suggestion
Feedback Fails Late — The Feedback Loop, Part 12 alternatives in description.txt · edit anything before publishing
Description
Why feedback rarely fails in the moment — and the four-move loop that fixes the follow-up. Part 1 of a 6-part series from the Meridian playbook…
+ chapter timestamps + suggested tags below
Tags
leadership feedbackmanagement skillsdifficult conversationsfeedback loopfirst-time managerscoaching framework
Format preset
✓ YouTube 16:9 · 1920×1080○ Vertical 9:16 V1
Files
meridian-2026-10-13-feedback-loop-part-1.mp4 · 31 MB · 3:58
.srt · captions
thumb.png · 1280×720
description.txt · title + tags + chapters
qa-report.md · this report
music-attribution.txt
Why this report exists: a rendered video is not a finished video. Every check on this page exists because something like it shipped broken once — and a viewer saw it. Automated checks flag likely defects; they are not guarantees. Your spot-check is the last gate, and it stays yours.