Automated Vertical and Horizontal Reframing System

You are a delivery automation engineer. Build an auto-reframing system that converts productions between 16:9, 9:16, and 1:1 while protecting the story, speakers, and action. Inputs: - Source masters: [INSERT - episodic, feature, ads, live capture] - Target formats and platforms: [INSERT - 9:16, 4:5, 1:1, platform list] - Content challenges: [INSERT - two-shots, fast action, on-screen graphics, subtitles] - Style constraints: [INSERT - headroom, safe areas, brand kit] - Volume and turnaround: [INSERT - episodes/week] Deliver in modules: 1. Objective: broadcast-quality reframed versions with minimal human intervention. 2. Process: saliency analysis (faces, active speaker, action, text/graphics) -> shot-level frame paths with smooth virtual camera moves -> subtitle/graphics re-layout -> multi-format render -> exception review. 3. Toolchain: [INSERT - reframe AI, speaker diarization, graphics engine, NLE/render farm], QC player. 4. QA: subject framing, cut-off graphics, motion-speed limits to prevent sickness, safe-area and subtitle checks, human review of flagged shots. 5. Delivery: per-platform masters, sidecar graphics projects, QC reports, and batch-status dashboards. Output: shot decision rules, exception categories, render-farm design, and cost per reformatted minute. ================