The production pipeline — roles, milestones, delivery

hard

Learn with your AI

Open this lesson in your favourite AI. It'll walk you through the why, explain the demo, and quiz you on the try-it list.

Open in Claude Open in ChatGPT

Why this matters

Making a 3-minute film solo with AI tools is possible; making a 10-minute one solo is brutal; making a 20-minute one requires a team. The roles — writer, director, generation-artist, editor, colorist, sound designer, composer — start to separate out even if one person wears many hats. This task introduces the production pipeline as a series of milestones with clear deliverables, so you can scale from solo to team without drowning in 400 MP4 files in a folder called 'Desktop.'

Demo

Six milestones: (1) Logline approved, (2) Treatment signed off, (3) Shot list finalised, (4) Pre-viz cut locked, (5) Final generations complete, (6) Picture lock (edit) + delivery. Each milestone has a deliverable that the next milestone assumes. Skipping any milestone — say, starting final generation before the pre-viz cut is locked — is the root cause of 90% of AI-film production disasters.

# Production schedule — 3-minute AI narrative short
 
Week 1
  MON: Logline + reference gathering (2 hrs)
  TUE: Treatment draft 1 (3 hrs)
  WED: Treatment review / revisions (2 hrs)
  THU: Script + shot list v1 (4 hrs)
  FRI: Character reference sheet(s) (3 hrs)
 
Week 2
  MON–TUE: Pre-viz all shots (8 hrs)
  WED: Rough edit, music placeholder (4 hrs)
  THU: Review with collaborators, notes, revise plan
  FRI: Pre-viz locked; final shot plan fixed
 
Week 3
  MON–WED: Final generations (20 hrs across the week)
  THU: Picture edit lock (4 hrs)
  FRI: Color + sound + music + delivery (6 hrs)
 
Deliverables per milestone:
  ☐ Logline approved
  ☐ Treatment signed off
  ☐ Shot list + reference sheets frozen
  ☐ Pre-viz cut locked (no more script changes)
  ☐ Final generations complete
  ☐ Picture lock + delivery
 
File system:
  /project-last-window/
    01-script/         treatment.md, script.md, shot-list.csv
    02-refs/           character-sheets/, mood-board/
    03-previz/         shot-1A-previz.mp4, ...
    04-final/          shot-1A-final.mp4, ...
    05-edit/           project.prproj or .fcpxml
    06-deliverable/    last-window-v1.mp4, last-window-v1-h264.mp4

Try it yourself

Commit to a calendar with dates for each milestone. If you can't assign dates, break the milestones into smaller ones. Dates force reality into the plan.

Build the file-system layout above for a real project. Populate it as you go. Find it BEFORE you have 200 files named 'clip1.mp4', 'clip1 final.mp4', 'clip1 final final.mp4'.

For a team of 3, assign milestone ownership. Writer owns logline-through-script; director owns shot-list-through-picture-lock; editor owns picture-lock-through-delivery.

Track generation hours across the project. Budget was 20 hours for Week 3. If you're at 40, you're overshooting and need to cut shots, not push harder.

Deliver. Don't keep iterating. Ship v1. Then maybe do v2 if the project warrants it. Shipping is a discipline — most AI films never ship because the maker keeps tweaking.

Prompt your AI

Use these three in order. Each builds on the one before.

1. Basics & terminology

In one paragraph, explain the six milestones of an AI-film production schedule and why skipping any of them tends to cause the project to implode.

2. Why it works (the mechanism)

Walk me through a production-pipeline bottleneck: what signals am I looking at that indicate a milestone is behind schedule, what are the typical recovery moves (cut shots, downgrade ambition, ship partial), and how do I know when to abandon a project?

3. Advanced — application & what's next

I'm producing a 15-minute AI documentary with 4 collaborators. Walk me through how to structure weekly stand-ups, review cadences, storage/version control (Frame.io, Dropbox, git-lfs), and how to handle the 'too many opinions' problem at the edit-lock stage.