Script Generation and Revision

Learning objectives

Explain how to use generation for options while revising action dialogue pacing and coherence and identify the relevant quality criteria.
Apply the lesson method to the student's own film and produce a usable versioned final script.
Evaluate the result, document a revision decision, and connect the evidence to the final capstone.

Section 2: Generative Scripting and Scene Design · Lesson 4 of 5 (9 of 45) · 60 minutes · Evidence: Versioned final script

Warm-Up ~5 min

A text generator can write a full script in ten seconds. That is exactly why the script it writes is rarely good. It averages thousands of stories, so it tends toward the safe, familiar version. The difference between a student who uses a generator and a filmmaker who uses a generator is revision: noticing what is weak, deciding what to keep, and rewriting with purpose. That judgment is the skill employers pay for.

Production journal: write your answer below.

What must a filmmaker understand before they can use generation for options while revising action dialogue pacing and coherence?

Instruction ~10 min

A draft is a starting point

A draft is a version of the script that is complete enough to read and criticize, but not finished. Your final script will be draft three or four, not draft one. Every draft gets a version label (v1, v2, v3) and a short note saying what changed and why. Version notes protect you: they show which choices were yours.

The three parts of a screen script

Action: what we see and hear, written in present tense and in plain, visible terms. “Mara wipes the lens with her scarf” is action. “Mara feels lonely” is not; you cannot film a feeling directly, so show the behavior that carries it.
Dialogue: what characters say, or what a narrator says. Good dialogue is short, sounds like a person, and carries a goal. People rarely say exactly what they mean.
Scene headings: a line naming where and when each scene takes place, such as INT. LIGHTHOUSE KITCHEN, EVENING.

Prompting a generator as a collaborator

Give the tool your beat sheet, your logline, your audience, tone and limits, and ask for a specific job: draft one scene, suggest three ways to say a line, find where the pacing drags. Smaller, specific requests give better and more useful results than “write my script.” Always ask it not to invent new characters or facts. After the draft arrives, you read it as an editor.

Revise for four things

Pacing: the speed at which story information arrives. Cut sentences that repeat what the picture already shows. Read aloud with a timer.
Coherence: does everything hold together? Does each scene follow from the last? Do names, objects and facts stay the same? Is there a cause for each effect?
Dialogue: would a real person say this? Could a line be cut and the scene still work?
Action: can every sentence be shown on screen and generated or filmed with what you have?

Keep, revise or reject

Mark every generated line with one of three decisions. Keep means it already works. Revise means the idea is useful but the wording is not yours. Reject means it does not serve the story. Write one short reason for each.

Worked example

The tool wrote: “Mara: I have always felt so alone in this lighthouse, Tomas, and your arrival fills me with hope.” Decision: reject. It states feelings outright and no one talks that way. Revision: Mara sets two cups on the table without being asked, then takes one away. Cause and effect improved: the viewer learns she expected him to stay, and learns she decided not to say so.

Common mistakes

Accepting the first draft because it looks finished.
Letting the tool rename characters or add new ones, which breaks the asset plan.
Dialogue in which characters explain their feelings in full sentences.
Overwriting old versions so you cannot show what you changed.

Workflow Study ~10 min

Writers who use generative tools describe the same cycle again and again: request, read, judge, revise, save. The tool is fast at the first and last steps. The middle two steps are where the quality comes from, and they cannot be handed off.

A creator typically asks for one scene at a time, with clear context. They paste the result into their own document, then mark it up, deleting filler and replacing generic lines with specific ones. They read it aloud to hear where it sounds false. They also check facts and names, because generators confidently invent details.

Experienced creators save every version before editing it, and they write a one-line note: “v2: cut the opening narration, because the picture already shows the storm.” That line is worth more than the draft, because it records a human decision. Without it, nobody can tell a considered edit from a lucky accident.

As you study, record each decision point in your production journal.

Decision point	What the filmmaker does	What the tool automates	What still needs human review
Request a draft	Writes a specific prompt with the beat, tone, limits and a rule against new characters	Generates a draft scene or script	Whether the draft follows the beat sheet and stays inside the limits
Mark up the draft	Marks each line keep, revise or reject with a reason	Nothing; this is editorial judgment	Whether the reasons are about the story, not just taste
Revise by hand	Rewrites the weak lines and fixes cause and effect	Can offer alternative wordings if asked	Whether the result sounds human and every action can be filmed or generated
Verify and save	Checks facts and names, saves a labeled version with a note	Can reformat the text	Whether any claims, names or private details are wrong or should not be included

Guided Studio ~15 min

Studio task: Mark generated lines as keep revise or reject and explain the decision.

Use the sample generated scene from your teacher, or the one below. It is a 40-second scene from a film about a lighthouse keeper.

“Mara stood on the lighthouse balcony. ‘The sea is so vast and I am so small,’ she said sadly. Suddenly, a mysterious ship appeared. ‘Oh, a ship!’ she cried. Tomas waved from the deck. ‘Hello, Mara! I have brought your supplies. Also, I have decided to tell you a secret that I have kept hidden for many years.'”

Split the scene into its lines and actions. Number them.
Mark each one keep, revise or reject. Write a one-sentence reason for each decision.
Choose one line you marked revise and rewrite it, so that it shows instead of explains.
Identify one place where cause and effect is weak. Write a sentence repairing it.
Compare decisions with a partner. Discuss any line you marked differently.

Coaching tips. If you reject nearly everything, check that you are rejecting for story reasons, not because you are tired of the style. If you keep nearly everything, read each line aloud and ask what the viewer would lose if it disappeared. “Suddenly” and “mysterious” are good words to challenge, since they tell instead of showing.

Keep / revise / reject log (copy into your production journal)

Script version: v___   Scene: ______

Line or action | Keep / Revise / Reject | Reason (one sentence)
1. ______ | ______ | ______
2. ______ | ______ | ______
3. ______ | ______ | ______

My rewrite of one line: ______
Cause-and-effect repair: ______
Version note: v___ changed ______ because ______

Knowledge check

+50 XP

Which sentence is written as filmable action?

Myth or Fact?

A final script can simply be the first draft a generator produces, as long as it reads well.

Fill in the blank

Checking that scenes follow from one another and names and facts stay consistent is checking the script’s .

+50 XP

You asked a generator for a scene and it added a talking fox you never mentioned. What should you do?

Myth or Fact?

Saving every version with a short note about what changed helps show which decisions were yours.

Describe one generated line from your own script that you rejected or rewrote, and explain how the change improved cause and effect.

1 / 6

Independent Studio ~15 min

Studio task: Produce a draft then complete a substantive human revision with tracked version notes.

Use your timed beat sheet from Lesson 2.3. Produce a versioned script that fits your target running time.

Request a first draft from your school-approved text generation tool, one scene at a time. Include your logline, the beat for that scene, tone, audience and a rule against new characters, locations or invented facts. Save the result as v1 exactly as received.
Mark up v1 using the keep / revise / reject log. Aim to revise or reject at least a third of the generated lines.
Rewrite by hand. Fix pacing (read aloud with a timer), tighten dialogue, make every action filmable and repair at least one cause-and-effect gap. Save as v2 with a version note.
Ask a peer to read v2 aloud. Mark where they stumble or look confused. Make a final pass and save v3 with its note.
Check every fact, name and place in the script. Remove any private information.

Model prompt (edit before using): “Write a 45-second scene for my short film. Beat: [paste]. Audience: classmates aged 14 to 17. Tone: [tone]. Characters allowed: [names only]. Location: [one place]. Use short lines of dialogue and describe only what can be seen or heard. Do not add characters, places or facts.”

Checklist before you submit: v1, v2 and v3 are saved; each version has a note; the final script follows the beat sheet and fits the target time; the keep / revise / reject log is attached; my file follows the course naming convention.

Exit Evidence ~5 min

Cite one change that improved cause and effect.

Student Deliverable & Assessment

Deliverable: Versioned final script.

Submit the versioned final script. The file must use the course naming convention and include enough process evidence to show the student's decisions and revisions.

File name (course naming convention): AIF_S02-L04_versioned-final-script_v01_YOURINITIALS

Criterion	Advanced	Proficient	Developing	Beginning
The work addresses the stated story or production purpose.	The final script follows the beat sheet, runs inside the target time and clearly serves the logline, with strong, filmable action and natural dialogue.	The final script follows the beat sheet, fits the target time and has usable action and dialogue.	The script drifts from the beat sheet or has several lines that cannot be filmed.	The script is incomplete or ignores the plan.
The student applies the required method accurately and produces usable capstone evidence.	Three versions with labeled notes, a complete keep / revise / reject log and a read-aloud timing prove a real revision process.	Versions v1 to v3 with notes and a completed log are present.	Versions exist but notes or the log are thin.	A single version, or no revision evidence.
The student identifies at least one quality issue and records a meaningful revision decision.	Names a specific cause-and-effect or pacing problem and shows a rewrite that fixes it, tested by a peer read.	Identifies one quality issue and records a meaningful revision decision.	Notes a problem but the revision is vague.	No revision decision recorded.
The work complies with consent rights attribution privacy disclosure and school policy.	Facts, names and private details are verified; all generated material is logged and disclosed; no imitation of living creators.	Facts are checked and generator use is documented.	Some claims unchecked or logs incomplete.	Contains unverified claims, private information or undisclosed generation.

Responsible Production

Requirement: Do not include unverified claims or private information in generated scripts.

A generator can sound sure of facts that are wrong. If your script mentions a real place, event, job, scientific idea or historical detail, check it against a reliable source and keep a note of where you checked. Remove any real names, addresses, phone numbers or personal stories that do not belong to you or that someone has not agreed to share. Do not ask the tool to write dialogue in the voice of a living person. If AI generation shaped your final script in a material way, say so in your credits or project notes. Your teacher can provide a non-generative path, such as writing from a partner’s oral suggestions, with the same revision requirements.

Up NextComplete & Continue →

AI Filmmaker: From Script to Screen

Curriculum