A comprehensive quality-assurance workflow for AI-generated screenplays and teleplays. Covers
format compliance, narrative structure, character consistency, dialogue quality, scene craft,
and format-specific rules for all major screenplay types.

## Key Points

- **The screenplay** — .fountain, .fdx, .pdf, .txt, .md, or plain paste
- **Format** — which of the 11 formats above (ask if unclear)
- **Beat sheet / outline** (optional but strongly recommended)
- **Character list / bible** (optional)
- **Series bible / show bible** (optional, for episodic formats)
- **Episode number** (for series/limited)
1. **Page count**: Is the total within target range? Flag if over/under by >10%.
2. **Scene heading (slugline) format**: Must start with INT. or EXT., include location and time of day. Flag malformed headings — AI commonly writes sluglines in sentence case or omits time of day.
3. **Action line style**: Present tense, visual, concise. Flag past tense, embedded camera directions (unless shooting script), non-visual inner monologue in action blocks.
4. **Character cue formatting**: Names above dialogue must be ALL CAPS. Flag lowercase cues.
5. **Parentheticals**: Brief and sparingly used. Flag: longer than one line, directing actor emotions in detail (AI overuses these), used as substitute for action lines.
6. **Transitions**: CUT TO:, FADE TO: — flag if excessive. Modern specs rarely use transitions except at act breaks.

Screenplay Audit Skill

A comprehensive quality-assurance workflow for AI-generated screenplays and teleplays. Covers format compliance, narrative structure, character consistency, dialogue quality, scene craft, and format-specific rules for all major screenplay types.

Step 0 — Identify the Format

Before running any module, determine which format applies. This changes structural expectations, page-count targets, act structure rules, and format-specific checks.

Ask the user if not already stated. Accepted formats:

Format	Abbrev	Typical Page/Length Target
Feature Film	FEAT	90–120 pages
TV Limited Series (6–8 eps)	LIMITED	45–60 min/ep (drama); 22–30 min (comedy)
TV Series (ongoing)	SERIES	42–58 min (1hr drama); 20–24 min (½hr comedy)
Adult Animated Series	ANIM-A	22–24 min/ep
Kids Animated Series	ANIM-K	11–22 min/ep
Animated Film	ANIM-F	75–100 pages
Stop-Motion Film	STOP	75–95 pages
CG Animated Film	CG	75–100 pages
Documentary	DOC	75–110 pages (feature); 45–60 min (TV)
Short Film	SHORT	1–40 pages (target stated by user)
Web Series	WEB	5–15 min/ep typically

Input Requirements

Collect before auditing:

• The screenplay — .fountain, .fdx, .pdf, .txt, .md, or plain paste
• Format — which of the 11 formats above (ask if unclear)
• Beat sheet / outline (optional but strongly recommended)
• Character list / bible (optional)
• Series bible / show bible (optional, for episodic formats)
• Episode number (for series/limited)

If supporting documents are missing, proceed with screenplay-only auditing and note degraded checks.

Audit Modules

Run all applicable modules unless the user says otherwise. Modules marked (episodic only) apply to LIMITED, SERIES, ANIM-A, ANIM-K, WEB. Modules marked (animation only) apply to ANIM-A, ANIM-K, ANIM-F, STOP, CG.

Module 1 — Format & Page Compliance

Goal: Verify the script is correctly formatted and hits industry-standard page targets.

Steps:

1. Page count: Is the total within target range? Flag if over/under by >10%.
2. Scene heading (slugline) format: Must start with INT. or EXT., include location and time of day. Flag malformed headings — AI commonly writes sluglines in sentence case or omits time of day.
3. Action line style: Present tense, visual, concise. Flag past tense, embedded camera directions (unless shooting script), non-visual inner monologue in action blocks.
4. Character cue formatting: Names above dialogue must be ALL CAPS. Flag lowercase cues.
5. Parentheticals: Brief and sparingly used. Flag: longer than one line, directing actor emotions in detail (AI overuses these), used as substitute for action lines.
6. Transitions: CUT TO:, FADE TO: — flag if excessive. Modern specs rarely use transitions except at act breaks.
7. Dual dialogue: Flag if needed but not formatted correctly.
8. Scene numbers: Present only in production drafts. Flag inconsistent/skipped/duplicated.
9. Title page: Check for title, author, draft date, contact info.
10. (Animation only) — Check for correct use of ANIMATIC NOTE: or visual description blocks.

Output: table of format violations with page/scene reference and severity.

Module 2 — Character & Name Consistency

Goal: Ensure every character is named and cued consistently throughout the script.

Steps:

1. Extract all character cues (ALL CAPS names above dialogue). Build character registry.
2. Flag:
   • Spelling drift in cues (DETECTIVE MARSH vs DET. MARSH vs MARSH)
   • Silent name changes for minor characters between scenes
   • Pronoun inconsistency in action lines
   • V.O. vs O.S. misuse: (V.O.) = voice-over narration, not present; (O.S.) = off-screen but in location. AI frequently swaps these.
   • CONT'D misuse: Missing on dialogue continuing after page break, or used incorrectly mid-scene.
3. Cross-reference against character bible if provided.
4. (Episodic only) — Flag series regular name/cue changes between episodes.

Output: table of Character | Canonical Cue | Variants Found | Scenes Affected | Severity

Module 3 — Story Structure & Beat Sheet Compliance

Goal: Verify the script hits proper structural beats for its format.

Feature Film (3-act):

• Act 1: ~pp. 1–25 | Inciting incident ~p. 10–12 | Act 1 break ~p. 25–30
• Midpoint: ~p. 55–60
• Act 2B / dark night: ~p. 75–85
• Act 2 break / climax build: ~p. 85–90
• Act 3 resolution: pp. 90–110+

TV Drama (4-act or 5-act + teaser):

• Cold open/teaser: pp. 1–5
• Act breaks at roughly equal intervals; cliffhanger required at each break
• Tag scene optional (eps 42–58 min)

TV Comedy (2-act):

• Cold open / teaser
• Act 1 break ~p. 15
• Act 2 resolution ~p. 28–32

Animated (episodic) — 2-act or 3-act compressed. Short Film — single-act or compressed 3-act. Web Series — highly variable; verify against user-stated episode targets.

Steps:

1. Beat sheet mapping (if provided): Confirm each beat lands in correct page range. Flag missing, misplaced, underdeveloped beats.
2. Act break detection: Clear act breaks? Dramatically compelling buttons?
3. Inciting incident: Present and placed appropriately? AI often delays or buries it.
4. Midpoint: Meaningful shift at structural midpoint?
5. Dark moment / all-is-lost: Low point before climax? AI frequently skips this.
6. Climax & resolution: Earned climax? Resolution addresses central question?
7. Subplot tracking: All subplots introduced — flag any opened but unresolved.
8. Plot hole detection: Impossible character knowledge, unmotivated actions, contradictory consequences.
9. Timeline coherence: Time-of-day contradictions, impossible travel, date errors.

Output: structure map + issues list with scene/page citations.

Module 4 — Scene Craft & Action Line Quality

Goal: Catch weak or AI-inflated scene writing.

Flag:

• Over-written action blocks: >4 lines unbroken (industry standard: 3–4 max, then white space)
• Unfilmable directions: "She feels a sudden pang of regret" — inner states without visual translation
• Camera directions in spec scripts: CLOSE ON, WE SEE, PUSH IN — inappropriate in specs (Exception: shooting drafts)
• Passive voice in action: Action should be active and immediate
• Redundant scene headings: Two consecutive sluglines for same location with no content
• Scenes that serve no function: Nothing changes — no new info, no status shift, no tonal shift
• Missing establishing context: Jumping to close action in new location without grounding
• Repeated scene descriptions: Same location described from scratch every time (AI context failure)
• (Animation only) — Action lines describe non-stageable live-action behavior
• (Documentary only) — Narration/VO editorializes instead of describes; interview setups without visual direction

Output: flagged scenes with page reference and issue description.

Module 5 — Dialogue Quality & AI Tell Detection

Goal: Surface AI-generated, expository, or out-of-character dialogue.

Flag:

• On-the-nose dialogue: Characters saying exactly what they mean with no subtext
• Exposition dumps: "As you know, Bob..." backstory speeches
• Uniform voice: All characters sound identical in vocabulary, rhythm, register
• Excessive parentheticals: Flag if on >30% of dialogue lines
• Monologue length: Single speech >8–10 lines in contemporary spec
• Repeated dialogue beats: Same argument/declaration in multiple scenes without escalation
• Dialogue that recaps action: Characters describing what the audience just saw
• Anachronistic/wrong-register: Period-inappropriate language, child speaking like adult
• (Kids animation) — Too complex, scary, or adult in tone
• (Adult animation) — Generic sitcom voice vs. show's established voice
• (Documentary) — Editorializing, leading, or unverifiable narration

Output: flagged lines with scene/page reference and issue type.

Module 6 — Continuity & World Consistency

Goal: Catch internal continuity errors and AI hallucinations.

Steps:

1. Prop continuity: Objects established in scene remain consistent
2. Costume/appearance continuity: Physical descriptions stay consistent unless change scripted
3. Location continuity: Room layout, geography, travel times
4. Time-of-day continuity: CONTINUOUS scenes maintain correct time; impossible jumps flagged
5. Real-world facts: Flag suspected hallucinations — wrong dates, invented institutions
6. Anachronisms: Technology, language, cultural references inconsistent with era
7. (Series) — Flag contradictions with series bible or prior episode canon
8. (Animation) — Flag actions inconsistent with established animation physics

Output: continuity issues with scene/page reference.

Module 7 — Duplicate & Recycled Content

Goal: Find copy-pasted, near-duplicate, or recycled scenes.

Steps:

1. Exact duplicate scenes: Verbatim or near-verbatim repeats
2. Structural recycling: Same scene type repeated (protagonist gets bad news three times, reacts identically)
3. Repeated dialogue phrases: Stock phrases appearing 3+ times
4. Scene-function doubles: Two scenes accomplishing same narrative purpose with no escalation
5. (Series/Limited) — Cross-episode recycling without intentional callback structure

Output: duplicates with scene/page references.

Module 8 — Episodic Consistency (episodic formats only)

Goal: Ensure consistency across episodes.

Steps:

1. Series regular behavior: Consistent voice, motivation, knowledge state across episodes
2. Serialized continuity: Prior episode events referenced correctly
3. Cold open style: Format-consistent across series
4. Teaser/tag conventions: Consistent use across episodes
5. B-story/C-story presence: Appropriate storyline count per format
6. Episode length variance: Flag significant over/under vs. target
7. Series arc tracking (serialized): Season-long arc beats landing in correct episodes

Output: per-episode summary + cross-episode issues list.

Module 9 — Format-Specific Checks

Kids Animated (ANIM-K):

• Age-appropriate content check
• Educational/prosocial element (if required)
• Correct act structure for 11-min vs 22-min
• Character voice appropriate for child audience

Adult Animated (ANIM-A):

• Written for voice performance, not live action
• Overuse of cutaway gags vs. plot-integrated humor
• Cold open/act structure matches show format

Documentary (DOC):

• Interview setups include subject identification
• Narration is factual, attributable
• B-roll directions present and specific
• Archival footage cues labeled (ARCHIVAL FOOTAGE:)
• No fabricated quotes attributed to real people

Stop-Motion (STOP) / CG:

• Action within physical/digital medium constraints
• No staging requiring impossible puppet/rig counts
• Visual gag staging specific enough for storyboard

Short Film (SHORT):

• Concept achievable within page/budget constraints
• No subplots that can't resolve in page count
• Single/minimal location honored if user specified

Web Series (WEB):

• Clear hook in first 30 seconds (~first page)
• Episode endings drive re-watch / next episode
• Episode length consistency across series

Output: format-specific issues list.

Report Format

# Screenplay Audit Report
**Title**:
**Format**: [e.g., TV Limited Series — Drama]
**Episode / Draft**:
**Date**:
**Page count**: [N] (target: [range])
**Supporting documents**: Outline: yes/no | Character bible: yes/no | Series bible: yes/no

---

## Executive Summary
[3-5 sentences: overall quality, biggest risks, recommended priority fixes.]

## Severity Legend
🔴 CRITICAL — Industry dealbreaker or story-breaking error; must fix
🟠 MAJOR — Noticeable error most readers will catch
🟡 MINOR — Small inconsistency; fix in polish pass
🔵 STYLE — Craft suggestion; optional

---

## Module 1: Format & Page Compliance       [findings]
## Module 2: Character & Name Consistency   [findings]
## Module 3: Story Structure & Beat Sheet   [findings]
## Module 4: Scene Craft & Action Lines     [findings]
## Module 5: Dialogue Quality               [findings]
## Module 6: Continuity & World Consistency [findings]
## Module 7: Duplicate & Recycled Content   [findings]
## Module 8: Episodic Consistency           [findings — episodic only]
## Module 9: Format-Specific Checks         [findings]

---

## Master Issues List
[All findings sorted by severity, with scene/page citations]

## Recommended Fix Order
[Prioritized: fix X before Y because...]

Workflow for Long Scripts / Multi-Episode Work

Single long feature (90+ pages):

• Pass 1 — Structural read: build scene list, character registry, timeline
• Pass 2 — Deep module audit using Pass 1 as anchor
• Pass 3 — Cross-check findings against script text before reporting

Multi-episode series (3+ episodes):

• Pass 1 — Per-episode structural read; build cross-episode character registry and event log
• Pass 2 — Per-episode module audit (Modules 1–7 + 9)
• Pass 3 — Cross-episode audit (Module 8)
• Pass 4 — Aggregate report with per-episode summaries + series-level master issues list

Handling Missing Inputs

Missing input	Degraded behavior
No outline / beat sheet	Skip structure compliance mapping; audit structure from script only
No character bible	Build registry from script; flag uncertain name cases
No series bible	Skip series canon checks in Module 8
Format not specified	Ask before proceeding; do not guess
Multiple episodes, no numbering	Process in order received; note numbering assumption

Anti-Patterns

Guessing the format. Always ask. A TV drama and a feature film have completely different structural expectations. Auditing a pilot as if it were a feature produces wrong findings.

Flagging camera directions in a shooting draft. Spec scripts and shooting scripts have different rules. Camera directions are errors in a spec; expected in a shooting draft. Ask if unclear.

Rewriting instead of auditing. The audit report identifies issues; it does not rewrite the script. Offer a rewrite pass separately if the user asks.

Ignoring format compliance. A structurally sound script with format errors reads as amateur to industry readers. Prioritize Module 1 findings in the recommended fix order.

Vague findings without citations. Every finding must include scene heading or page number and a brief quote so the writer can locate it instantly. Vague findings are useless.