System operational · 4-hour avg turnaround
● Private beta · invitation only

Greenlight decisions, backed by evidence.

ReelPen reads your screenplay end-to-end — not a 13% excerpt — and produces an auditable report grounded in verbatim citations. Fifteen specialty modules read a single source-of-truth extraction, then cross-reference each other to surface insights no single reader catches. Every claim is provenance-tagged. Every quote validated against the original script.

Currently in private beta with a small group of partner studios and agencies. Reports delivered within one hour.

100%
Script coverage
15
Specialty modules
0
Hallucinated quotes
≤1h
Delivery SLA
Trusted by acquisition teams at
Meridian Pictures
Northgate
Halverson & Co.
CIRRUS/MEDIA
Foxglove Lit
Atlas/Stage
01 — What we measure

Fifteen specialty modules. One source of truth.

Every module reads from the same Phase 1 extraction — the same scenes, the same beats, the same characters. Then each one slices what it needs and produces a focused diagnosis with verbatim citations. Eleven modules are live in private beta; four more are in active development for the public launch.
M-01

Greenlight Score

Composite 0–100 health score across five axes: craft, structure, market readiness, execution risk, authenticity. Tier classification (Pass / Develop / Verified / Greenlight-ready) with module-weighted rationale. Synthesizes every other module's verdict.

Tier 2B synthesislive
M-02

Plot Structure

Act balance vs format ideal · beat coverage across Save-the-Cat / Field / Yorke / Aristotle · causal spine integrity (therefore-vs-and-then) · value arc trajectory · structural risks with severity.

Tier 2A specialtylive
M-03

Pacing

Scene length distribution, per-act tempo, long-scene drag clusters, short-scene fragmentation, beat density per page, momentum risks. Format-aware (TV pilot vs feature thresholds).

Tier 2A specialtylive
M-04

Genre Authenticity

Convention compliance against declared genre · signature devices · hybrid signals · tone register check · audience expectation match. Identifies where the script honors its register and where it borrows from another.

Tier 2A specialtylive
M-06

Character Deep Dive

Protagonist arc (want/need/start/end), supporting cast roles, antagonist analysis (or its honest absence), archetype distribution (Jung/Egri taxonomy), character network and isolation pairs.

Tier 2A specialtylive
M-07

Logline Optimizer

Five logline drafts (functional / aspirational / theme-fwd / genre-fwd / safety) with strengths and target market. Recommended logline + short pitch + elevator pitch + positioning axes.

Tier 2A specialtylive
M-08

Synopsis Package

Four-tier synopsis chain: one-line · two-line · short (~150w) · long (~600w) · treatment (~1200w). Plus key selling points with evidence paths, audience pitch, comparable hints.

Tier 2A specialtylive
M-09

Dialogue Analysis

Per-character voice signature, blind-page identifiability test, signature devices, density distribution, subtext register, iconic-line density. Catches voice blends across principals.

Tier 2A specialtylive
M-10

Emotion Map

Per-scene valence and tension trajectory, emotional peaks and troughs, value-shift inventory by axis, per-act emotional balance, audience emotional journey synthesis.

Tier 2A specialtylive
M-13

Revision Notes

Prioritized revision register: top 5 high-leverage edits (with where, why, expected impact, effort) plus structural / character / dialogue / thematic / pilot-specific notes. Includes a protect list — what NOT to change.

Tier 2B synthesislive
M-15

Extrinsic Risk

Seven categories: talent · IP & clearance · legal/regulatory · geopolitical & cultural · content sensitivity · market timing · production extrinsic. Severity-coded with mitigation hints — what would block this from greenlight.

Tier 2B synthesislive
M-11

Comparable Films — in development

Tier-stratified cohort of comparable films from a 50K-film database with DNA-pattern matching against tonal register and structural archetype. Returns 8 anchors with ROI bands.

requires brain layersoon
M-12

Market Risk — in development

Genre saturation analysis, recent-comparable performance trend, audience cycle position, distribution window risk. Forecasts the market environment a script would land into.

requires brain layersoon
M-16

Production Tier — in development

VFX scale · location count · cast tier · stunt density · period setup. Classifies the script into a production complexity band so producers can size crew, budget, and schedule before development closes.

rule-based + brainsoon
M-17

Budget Estimator — in development

Tier-stratified budget bands derived from a 13K-film cost corpus, anchored by production complexity (M-16). Returns a confidence interval, not a point estimate. Honest about within-tier variance.

requires brain layersoon
▲ Pricing available at public launch · beta access is currently complimentary for invited partners
02 — Pricing

Transparent. Per-report. No retainer.

We bill per analysis. No seats, no subscriptions, no minimums. Volume agreements available for studios processing 100+ scripts per quarter — speak with our enterprise team for SLA and on-prem options.
Starter
Single report
$39/ report
4-hour turnaround · No commitment
Enterprise
Full dossier
$179/ report
24-hour turnaround · Volume pricing
Greenlight Score6-dim verdict + DNA scoring
Reader-grade coverageSynopsis · comments · verdict
Structural diagnosticsPlot · pacing · character
Market & ROI briefComparables · budget tier · risk
Page-cited revision notesPrioritized rewrite list
Volume & SSOSAML · audit log · DPA
03 — Why cascade

Coverage AI, re-engineered for trust.

Most coverage tools sample a fraction of your script and ask one model to do everything. The result is shallow, inconsistent, and prone to inventing material that isn't on the page. The cascade architecture takes a different shape — and the difference is auditable.

Traditional AI coverage

Excerpt & summarize
  • Script coverage~13% · only an excerpt is read · later pages invisible to the model
  • Hallucination riskQuotes are paraphrased or invented · no validation against source
  • Module isolationEach tool is a separate prompt · no cross-reference between specialties
  • Provenance"It said it" · no traceable path back to the script
  • SynthesisSingle-pass summary · no recognition of pattern convergences
  • ReproducibilityRe-running the same script gets a different report

▸ ReelPen Cascade

Read once · slice many · synthesize
  • Script coverage100% · every page read end-to-end in Phase 1 · whole-script ground truth
  • Hallucination riskZero hallucinated quotes · every verbatim citation validated against the source before the report ships
  • Module compositionEleven specialties read the same source of truth · cross-module synthesis layer surfaces convergent insights
  • ProvenanceEvery claim cites a Phase 1 path · audit-grade · committee-defensible
  • SynthesisCross-module pattern recognition · explicit convergences AND conflicts · prioritized actions
  • ReproducibilityValidator chain enforces consistency · re-run produces structurally identical output
04 — How the cascade works

Three phases. One source of truth.

Most AI coverage tools read a 13% excerpt of your script and ask one model to summarize everything. We do the opposite: we read every page once, build a verbatim-grounded ground truth, then dispatch fifteen specialized modules that each see only what they need. The result is auditable, reproducible, and free of hallucinated content.
Phase 01

Extraction

The full screenplay is read end-to-end and converted into a structured ground truth: scenes, beats, characters, dialogue, theme statements — all with verbatim citations from the source. This is read once, then never re-read.

single Claude pass · whole script · verbatim grounded
Phase 02

Specialty modules

Eleven specialty modules run against curated slices of the Phase 1 extraction. Plot Structure sees acts and beats. Dialogue sees voice patterns. Each module produces a focused diagnosis with provenance to the source. No module re-reads the script.

eleven modules · parallel · slice-bounded
Phase 03

Cross-module synthesis

A synthesizer reads all module outputs together and surfaces patterns that no single module can see — convergences, conflicts, the dominant structural feature. This is where the cascade earns its name: ground truth flows in, specialty insights flow out, synthesis ties them together.

composer · validator · branded PDF
Validation

Verbatim & provenance check

Every quote in the final report is matched against the original screenplay. Every analytical claim cites the Phase 1 path it came from. If a module hallucinates a line, our validator catches it before the report ships. Zero hallucinations is not aspiration — it's our pre-flight check.

automated · validator chain · pre-delivery
05 — Sample report

See a real cascade output. Not a mock.

The sample below is a full master report from our pilot run on a 22-page TV pilot ("The Kingdom of Future"). Every quote is verbatim from the screenplay. Every claim is tagged to a Phase 1 path. The Greenlight composite, the cross-module synthesis, the prioritized revision register — all generated by the eleven live cascade modules running end-to-end.
▸ KOF Episode 1 · TV Pilot

"Causally tight, dialogue-disciplined premium-cable thesis pilot — recognition without reversal."

The cascade synthesizer's headline diagnosis after reading 11 specialty modules. Every word backed by ≥2 module convergences and tagged to specific scenes in Phase 1.

  • Composite score: 73 / 100 (Tier 2 — Develop / mid-band)
  • 9 cross-module signals identified
  • 2 module conflicts surfaced (production cruxes)
  • 109 verbatim quotes — all validated against source
  • 288 KB branded PDF · 16 sections · 596 lines
▸ Read it yourself

A full 16-section master report.

Executive synthesis · at-a-glance · plot structure · pacing · genre authenticity · character deep dive · logline optimizer · synopsis package · dialogue analysis · emotion map · revision notes · extrinsic risk · combined risk register · module audit. All in one document, all cross-referenced, all auditable.

Open the sample report →
06 — Frequently asked

The cascade, explained.

The questions partner studios and agencies asked us during the pilot. For procurement, security, or data-handling questions not covered here, request our compliance packet.
Q.01
What does "100% script coverage" actually mean?
Phase 1 of the cascade reads every page of your screenplay end-to-end in a single pass and converts it into a structured ground truth — scenes, beats, characters, dialogue, theme statements — with verbatim citations from the source. Most other AI coverage tools sample 10–15% of a feature script (the 30K-character "smart excerpt" pattern) and leave the rest invisible. We don't. The trade-off is honest: a single Phase 1 read costs more upfront, but every downstream module benefits from the same complete extraction without re-reading the script.
Q.02
How do you guarantee "zero hallucinated quotes"?
A pre-delivery validator chain. Every verbatim quote in the final report is matched against the original screenplay text. If a module produces a line that paraphrases or invents content not in the source, the validator flags it and the module's output is marked failed. The report ships only when every quote is grounded. We also enforce a "cross-speaker join forbidden" rule that prevents the model from concatenating dialogue across an intervening character — a subtle hallucination pattern other tools miss.
Q.03
Why three phases? Why not just one model call?
Three reasons. First, attention degrades over long contexts — a single model reading 197 pages and producing 17 specialty analyses simultaneously will drop signal in the middle. Second, modules need different slices: Plot Structure cares about acts and beats but not dialogue voice; Dialogue cares about voice but not act balance. Slicing keeps each module focused. Third, cross-module synthesis is impossible without specialty outputs to compare. Phase 3 surfaces convergences ("Plot says recognition-without-reversal + Emotion says warm-conviction closing → pilot ends without urgency") that no single pass produces.
Q.04
What does a typical report look like?
A branded master report in markdown and PDF — typically 16 sections covering executive synthesis, at-a-glance script facts, per-module diagnostics, combined risk register sorted by severity, and a module audit summary. Every analytical claim cites a Phase 1 path. Every verbatim quote is validated. The KOF v2 sample report (linked in the section above) is 596 lines and 288 KB. Look at it before signing anything.
Q.05
Does this replace our readers?
It is not designed to. The cascade is a triage and structural diagnosis layer, not a substitute for human judgment about story. Most pilot partners use ReelPen to surface the questions worth bringing into the senior reader pass — "Act 2 sags, midpoint is recognition not reversal, no antagonist on the page" is a coverage starting line, not a coverage ending one. Reader hours are redirected, not eliminated.
Q.06
What happens to my script after analysis?
Your screenplay and the cascade artifacts (Phase 1 extraction, module outputs, master report) are stored encrypted on dedicated infrastructure, never used for training, and purged on request. Beta partners receive an explicit data-handling agreement before submission. Public launch includes SOC 2 reporting, SSO, and on-prem deployment options for enterprise.
Q.07
When does pricing become real?
Public launch, later this year. Beta access is currently complimentary for invited studios, agencies, and producers. Pilot partners help us calibrate bundle sizing, module emphasis, and integration patterns — feedback shapes the public launch directly.

Run a real cascade on your script.

Private beta access for studios, agencies, and producers. Submit one screenplay, get a full cascade master report within an hour. Verbatim-grounded, validator-checked, committee-defensible. No payment during beta.

Request beta invite → Sign in (already invited)
Replies within one business day. We're inviting in small batches to keep the pilot honest.