Video Intake Addendum - Build the Conveyor Belt 2026-06-30

Added from AI Builder Handoff - St. Pierre Video Content Engine.

Objective

Make Collins prove the raw-footage intake path, not just one finished edit.

Why it matters

The real engine is raw drop -> transcript -> rough cut -> Reel -> checks -> fanout.

You need to

Use this to score the week-1 ticket. ~5 min.

Status

Context-only. No live access or automation created.

The correction needs you
★ Default: first demo = one raw clip in, one cut/captioned vertical Reel out, with transcript, manifest, and contact sheet.
What already exists no action
AssetWhat it means for Collins
Local Pack MachineAlready turns matched raw clips into pack folders with copy, statics, carousel, transcript, QA sheet, contact sheet, and manifest. This is a major first-week source.
Remotion engineExisting renderer for vertical Reels, YouTube/VSL cuts, and show formats.
Remotion pipeline noteDefines the Drive -> cut -> render -> publish wrapper. Publishing remains TODO/held.
IG visual standardHouse look: close talking head, black top hook chip, white captions, one proof overlay when needed, no AI avatars.
OpenMontage silence cutterDead-air cutting exists, but it is not wired into the St. Pierre intake path yet.
Canonical VSL packBest demo pack and fallback raw clip.
MVP acceptance criteria no action
Build order for the intake path reference only
  1. Folder contract: raw drops, processing, rough cuts, exports, checks, manifests.
  2. Clip manifest schema: topic, show type, intended outputs, transcript path, cut status, export paths.
  3. Recording packet template: hooks, angles, bullets, CTA, proof, must-say, avoid-saying notes.
  4. Transcript generation path.
  5. Optional OpenMontage silence cutting pass.
  6. Route cut clip into the existing Remotion MVP Reel lane.
  7. Use first-party captions as the default until VEED/Descript clearly beat it.
  8. Generate contact sheet and frame checks for every export.
  9. Add VSL outline, statics, and carousel fanout after MVP Reel is stable.
  10. Only then add transcript-driven bad-take detection.
Still held guardrails
The intake addendum is a build direction, not permission to create live automations or external writes.
Path: Projects/St_Pierre_Digital/Recruiting/Collins-Onboarding/video-intake-addendum.html