Script & shots, in parallel

Why the shots are in this order

They follow Bryan's script, not Jesse's. There are two scripts in play and they are structured differently.

• Bryan's short-film script — 26 scenes, 151 shots, IDs A1-01…A1-87 then A2-01…A2-64. It dramatises the Aztec story in straight chronological order: the gate at dawn → the call → the disc → inside → the military → the recovery → the cleanup → Scully and the debunk → the trial → the coda. Every AI shot was generated from this document, so the shot order is its scene order.
• Jesse's documentary script — an essay structure: cold open → the "hoax" framing → main interview → witnesses to the crash → recovery → the debunk → MIB → and the dramatised crash scene at the end. It is driven by VO and real footage, with camera-card timecodes (D1/D3/D4/D5) for the documentary B-roll.

So these shots are inserts, held in story order, for Jesse's edit to place where the narration calls for them. The numbering is a filing system, not a running order for the finished film. The one place the two scripts meet directly is the dramatised crash scene — diner, night highway, the homestead at dawn, Hart Canyon — which is new material and is tracked separately in docs/new-script-2026-09-16/.

The five columns

• SCRIPT — Bryan's line for the shot, plus the intended motion and the original generation spec.
• FIRST FRAME — the locked still. This is the frame the video models are conditioned on.
• LAST FRAME — where the shot should end up. Only a handful exist so far; the rest are placeholders and most will need generating.
• FREE GEN — the local render, MiniMax H3 or LTX 2.5. Costs nothing but GPU time.
• PAID GEN — reserved for Kling or Seedance on the shots where the free render is not good enough. Empty until we spend.