Race-Day Reel: Normalize Mixed Footage for Vertical Cut

Problem Description

You're helping build a social reel from a regional triathlon. The shooter came back with a mix of footage: handheld iPhone clips at the start and finish, DJI drone footage shot in a log color profile for maximum dynamic range, and a couple of JPEG race photos from the event photographer. Everything is landscape orientation — the reel needs to end up as a 9:16 vertical short for social.

The probe step has already been run and the clip metadata is recorded in inputs/clips.json. The still photos are at inputs/raw/race_photo_finish.jpg and inputs/raw/race_photo_podium.jpg. The metadata includes color profile information, resolution, and notes about subject positioning within each frame. Use this to determine the correct normalization strategy for each clip.

Your job is to bring all this footage to a common mezzanine standard so it's ready for cut planning. Different clip types need different treatment — make sure each clip gets the right processing given its source characteristics.

Output Specification

Produce the following files:

1. work/clips.json — A copy of inputs/clips.json updated with any per-clip options needed before normalization. Add appropriate fields to each clip object based on its source type and the notes.

2. work/normalize_commands.sh — A shell script with the complete normalization commands for every clip, with all relevant flags and options. Include a brief comment above each command explaining what treatment is being applied and why.

3. work/normalization_plan.md — A short summary explaining:

   • Which clips received which color treatment and why
   • How each clip's source type influenced the normalization choices
   • Any clips left with default settings and the reason