AI script breakdown.

What a breakdown is, and why it is the first real document

A screenplay is written to be read. A breakdown is the same screenplay re-written to be shot. For every scene it lists what the scene needs: which characters, how many extras, where it is set and whether that is interior or exterior, day or night, which props are handled, what the cast wears, which vehicles and animals appear, what has to be rigged or added later, how long the scene is in pages. The first assistant director makes it, traditionally with a ruler and a set of coloured highlighters, one colour per department.

It matters because everything downstream reads it. The schedule groups scenes by location and cast so the company moves as few times as possible. The budget prices each department's list. The art department builds from the props column; wardrobe from the costume column; the stunt coordinator from the effects column. A scene missed in the breakdown is a scene the schedule does not know about, and a prop missed is a runner sent to a shop on the morning of the shoot. The breakdown is the first document in a production that is a plan rather than an intention, which is why it is the first thing worth automating and the last thing worth automating badly.

How an automatic breakdown works, step by step

Parsing first. Before anything can be extracted, the screenplay has to be read as structure: sluglines become scenes, scenes carry their interior-or-exterior and day-or-night, character cues become cast, action lines are separated from dialogue. A PDF makes this harder than it sounds — page breaks fall inside scenes, and a scanned script has no structure at all — so the parser is a real piece of engineering rather than a prompt. Script reads a PDF, a Fountain file or plain text into scenes before any breakdown is attempted; a scene that was split or merged on import is fixed here, not later.

Then extraction, per scene, against a taxonomy. Each scene is read for its elements, and the elements land in typed categories: cast, props, locations, effects, vehicles, animals, costumes. The taxonomy is the point. A list of nouns is not a breakdown; a breakdown says which department each noun belongs to, because that is how a budget reads it. Script's breakdown extracts into that typed list and keeps it editable, so a wrong category is a drag rather than a re-run.

Then the questions a producer asks. How many days? Script's shooting-days planner groups scenes by location and cast — the two things a company moves for — to estimate a day count. Which scenes share a set? Which character is in the most scenes, and could their days be blocked together? These are the first conversations of pre-production, and a breakdown that can answer them on the day the script arrives changes what that first meeting is about.

Then the handoff. A breakdown is not an end in itself. In PREVIZ it is the input to the shot list — coverage is proposed per scene from what the scene contains and who is in it — and the cast and location lists become the identities a production will lock. The shot list page picks up where this one stops.

What goes wrong, and how to check a breakdown

Implied elements. "She pulls up outside the house" needs a car; the scene never says the word. "Rain streaks the window" needs a rain rig or a visual effect. An automatic breakdown reads the text and is good at explicit elements and inconsistent at implied ones. Check every exterior for weather and vehicles, every scene that begins mid-action for what the action required.

Invented elements. The opposite failure: an engine reads "a cluttered desk" and lists a lamp, a typewriter and a stack of letters that the writer never specified. Each is a prop someone might source. A good breakdown distinguishes what the script says from what the art department might decide, and lets you delete the second kind in one gesture.

Characters that are the same person. "MARA" in scene one and "MARA (V.O.)" in scene nine and "YOUNG MARA" in the flashback are one, two or three cast members depending on the script, and a parser cannot know which. Merge or split them deliberately. This matters twice over in PREVIZ, because a cast member becomes a locked identity — the face every frame of the production is graded against — and two entries for one person means two locks.

Scene boundaries. A slugline missing its INT./EXT., a montage written as one scene that is really six locations, a scene number repeated. These are the script's problems, and the breakdown inherits them unless they are fixed at the parse. Look at the scene list before the element list.

What a good output looks like, then: every scene present once with the right time of day; every named character present once; the explicit props and the implied vehicles and effects both listed and marked which is which; a day estimate with the grouping it assumed visible; and the whole thing editable, because the first assistant director's judgement is not being replaced, it is being given a head start.

Where PREVIZ does it

Script is the room. Import or write the screenplay; scenes separate; the breakdown extracts into the typed, editable list; the shooting-days planner estimates the count; the shot list builds at the density you set; and the result hands to Storyboard with the breakdown attached, so the frames know which scene they are in and who is in them. The whole production can leave at the end as a production package — scene index, breakdowns, cast and continuity sheet, shot list, boards, schedule — as a PDF or a link a crew reads.

What Script does not do is the full stripboard with crew, equipment and union constraints; that is a scheduling job, and the day estimate is an input to it rather than a substitute. It also does not decide what the art department will put on the desk. It tells you what the writer put there. The rooms page describes Script beside the five rooms that follow it, and pricing is by credits, with every room open on every plan.

Questions people ask