10-lesson course · about 20 hours

Build agents that follow script changes, and keep the call with your team

For Filmmakers, producers, editors and production teams coordinating a story across development, planning and post-production

Everything for Filmmakers

Set up a named role such as Sequence Desk in OpenAI Dots or Claude Duet. It follows approved revisions and prepares breakdowns, shot requests and post handoffs for human review.

Try lesson 1 freeSee what you’ll learn

What you’ll be able to do

By the end, you can…

Write a brief with clear edgesGive a named role such as Sequence Desk a written brief. It says what wakes the agent, what it may change, what it must never touch and whom it asks. It also says whether the work is a single run or a continuing brief, which connections are approved, what the agreed reports are and how the work is paused or retired.
Keep the approved baseline safeBuild a revision agent that compares approved script versions on an event or a schedule and explains what changed, scene by scene. It ignores duplicates and pauses on misdated or unconfirmed drafts. Only a named person can make a version the planning baseline.
Update breakdowns without inventing anythingHave a breakdown agent change only the scene records an accepted revision affects, and cite that revision on each edit. Scheduling, staffing, cost and safety conflicts go to the director, producer or department lead. The agent never makes up availability, costs or operating instructions.
Keep shot tests tied to their scenePrepare shot requests that cite the current revision. Register returned Higgsfield or Runway tests in order, set a resubmission limit and keep rejected versions. The agent pauses when a test contradicts the plate or another reference.
Hand post a packet that survives revisionsBuild a versioned post packet from a supplied media manifest and review notes. The agent flags missing media instead of substituting another file. It holds old notes back from a replacement shot until someone reviews them.
Test the workflow under change, then pause or retire itRun the connected agents on a small sequence against a new revision, a missing shot and a delayed review, and record each stop with its reason. Altered frames stay blocked until the performer approves them. You also practise pausing, resuming and retiring the agents.

How it works

A tutor that meets you where you are

  1. Lessons that adapt to your levelTen self-paced lessons follow the five modules, from script revisions through to readiness checks. The material adjusts to what you already know, so a script supervisor and a post coordinator can each start from familiar ground.
  2. An AI tutor in place of a live trainerAsk the tutor anything, typed or out loud. It helps you draft agent briefs, check your routing tables and work out why a test run stopped or failed to stop. You can use the course's scene forty-two or a small, non-confidential sequence from your own production.
  3. Quizzes and flashcards for each lessonShort quizzes and flashcards are tied to each lesson. They check the parts that matter in practice, such as who confirms a version, when an agent must pause and what it may never touch.
  4. Progress tracking and a certificateTrack your progress through the modules and pick up where you left off. You receive a certificate of completion when you finish the course.

Curriculum

The curriculum

The course follows a script change through the sequence: from the revision itself, to the breakdowns, the shot tests, the post handoff and a final test of the whole workflow under change. Each lesson builds one agent or one control. Together they make a set you can test, pause and retire.

Module 1Track what a script revision changes.2 lessons

  1. 1Compare approved script versions
    • Write an agent brief for a named role that states what wakes it, what it may change, what it must never touch and whom it asks
    • Choose between OpenAI Dots and Claude Duet, and between a single run and a continuing brief, and give reasons that fit your production
    • Build a revision agent that compares two approved script versions on an event or a schedule and reports the changes by scene
    • Explain why every agent needs a remit and a line it must never cross
  2. 2Preserve the approved baseline
    • Configure the revision agent to recognize duplicate revisions, ignore them and note this in its report
    • Set the agent to pause on misdated or unconfirmed drafts and send a clear question to the script supervisor
    • Test the agent against the Friday-night blue pages and record each stop with its reason
    • Explain why only a named person can make a version the planning baseline

Module 2Turn changes into department questions.2 lessons

  1. 3Update the affected breakdowns
    • Build a breakdown agent that wakes only on an accepted entry in the revision log
    • Update only the scene records an accepted change affects, citing the revision on each edit
    • Separate what the script states from what the agent infers, so plausible additions don't become production requirements
    • Check the agent's edits against the pages before accepting them
  2. 4Route planning conflicts
    • Identify the planning conflicts an accepted script change creates across takes, locations and departments
    • Build a routing table that sends each kind of question to the director, producer or department lead
    • Set the planning agent to hold dependent work until a decision is recorded
    • Write agent questions that state what is known and unknown without inventing availability, costs or operating instructions

Module 3Keep shot tests tied to their scene.2 lessons

  1. 5Prepare the next shot requests
    • Build a shot agent that drafts requests citing the current script revision, scene and breakdown entry
    • Choose between a supported connection and a reviewed export for Higgsfield or Runway
    • Set the agent to mark requests based on a superseded revision as stale
    • Explain why the agent prepares shot requests but never decides final picture
  2. 6Track returned tests and revisions
    • Build a test register that records each returned version in order with its request, revision and status
    • Set a resubmission limit and the stop that happens when it is reached
    • Configure the agent to pause when a returned test contradicts the plate or another reference
    • Check every returned test for whose face and voice appear before it can be selected

Module 4Prepare an edit handoff that survives revisions.2 lessons

  1. 7Assemble the post packet
    • Build an editorial coordinator agent that assembles a post packet only from the supplied manifest and notes
    • Configure the agent to flag missing media and stop retries instead of substituting another file
    • Version the packet so each change to an approved shot produces a new, traceable packet
    • Explain how one error in the packet spreads through conform, grade, mix, trailer, dub and delivery
  2. 8Close the feedback loop
    • Route time-referenced review notes to the person who owns each one
    • Configure the agent to hold notes from a replaced shot for review instead of carrying them forward
    • Track note status from open to closed in the agreed report
    • Test the loop by swapping in a replacement shot and checking that no note moved without review

Module 5Test the sequence workflow under change.2 lessons

  1. 9Run the sequence checks
    • Connect the agents so every handoff names the current approved script revision
    • Build a checking agent that produces a readiness report and continuity checks for human review
    • Test the workflow against a new script revision, a missing shot and a delayed review, and record each stop with its reason
    • Keep altered frames blocked until the performer has approved them
  2. 10Pause and resume the workflow
    • Pause a continuing brief in Dots or Claude Duet and record what each agent is holding
    • Resume the workflow only after the named owner records a decision and every agent holds the same revision
    • Tighten the agent brief where a test showed a silent failure or an unclear stop
    • Retire the agents by closing connections, revoking access and archiving their reports

Who it’s for

Made for you if…

Made for: Filmmakers, producers, editors and production teams coordinating a story across development, planning and post-production

Script supervisors and coordinators

You're the person who has to say which version is current. You'll set up agents that ask you before anything becomes the baseline, rather than guessing.

Producers and department leads

You want the repetitive paperwork handled without decisions on schedules, staffing, cost or safety leaving your hands. The agents send those questions to you and hold dependent work until you decide.

Editors and post coordinators

You keep getting old notes on new cuts and media that quietly goes missing. You'll build a versioned handoff that flags gaps and holds notes back until someone reviews them.

Filmmakers who already prompt AI

You can already judge a single AI result against the scene. Now you want recurring agents that keep departments working from the same script, with clear stops.

Questions

Good to know

What do I need before I start?

You should be able to write a clear prompt for a single production task and judge the result against the scene, either from the Film AI Intro or from your own work. You also need working knowledge of script revisions, breakdowns, shot lists and editorial notes, and access to OpenAI Dots or Claude Duet. It helps to know who on your production approves script versions, schedules, safety and performer likeness. If you don't, you can name those roles for the exercise.

Should I use OpenAI Dots or Claude Duet?

The course treats them as equal options. In the first lesson you choose one, and choose between a single run and a continuing brief, with reasons that fit your production.

Can I practise on my own production?

Yes, as long as you use a small, non-confidential sequence. Otherwise use the course's illustrative scene forty-two. You connect only the apps and folders the exercises need. The course also covers how to protect unreleased story material, performer information and private crew records within agent connections.

Will the agents make scheduling, budget or creative decisions?

No. The agents prepare records and questions for review. Scheduling, staffing, cost, safety and final picture stay with the director, producer and department leads. Altered frames stay blocked until the performer approves them.

Do I need to write code?

No. You work only with what Dots or Claude Duet support. Writing custom agent software, APIs or pipeline code isn't part of the course.

Does the course explain union rules on AI and likeness?

The course treats performer consent as a gate in every workflow. It doesn't give legal advice on SAG-AFTRA or other union contracts. For those, it points you to the current agreement text.

How long does it take, and does it cover multi-team agent setups?

Allow about 20 hours, self-paced. The course covers individual agents connected around one sequence. The multi-team Factory with a shared scene record is part of the Capstone, which this course doesn't include.

Film AI Intensive: proactive agents for sequence preparation

$397

One payment · lifetime access

  • 10 self-paced lessons in 5 modules, from script revisions to readiness checks
  • Hands-on agent building in OpenAI Dots or Claude Duet, your choice
  • Practice on the course's scene forty-two or a small sequence of your own
  • An AI tutor you can ask anything, typed or out loud
  • Quizzes and flashcards for each lesson, plus progress tracking
  • Certificate of completion

Secure checkout by Stripe. Have a promotion code? Enter it at checkout.

Refunds: Full refund if you ask before you open your first lesson. Once you’ve started, purchases aren’t refundable, except where the law gives you that right.

Already bought it? Sign in