Upload a script or pitch bible. Get predictive audience intelligence, commercial forecasting, and a clearer path to greenlight, grounded in real engagement behavior instead of demographic guesswork.
Every output separates what we read from your material from what the model assumes, and labels how confident it is.
Script, pilot, series bible, treatment, or deck. Add project assumptions if you have them. They're optional.
A private model extracts premise, tone, themes, character strength, and cultural specificity, then maps them to taste clusters.
Audience Impact Score, personas, comparable-title read, and three revenue scenarios with stated assumptions and confidence.
A Greenlight Readiness read and prioritized revisions ranked by their likely effect on interest, retention, and ROI.
Sample read of a fictional pilot. The cluster names are the model's 12 taste clusters.
How strongly a project may resonate across selected segments, with predicted awareness, intent, emotional connection, and recommendation likelihood for each.
Whether the material reads as authentic, specific, and differentiated, or where it risks feeling generic. Transparent criteria, never reducing a group to a stereotype.
The films and series your project sits nearest, with audience overlap and the lesson each one carries. Add or remove comps and the forecast updates.
Conservative, base, and breakout: gross, net, ROI, break-even, and payback. Every line labeled as market data, your input, or modeled estimate.
A composite read across demand, creative strength, differentiation, budget efficiency, and distribution fit, with the reasoning behind the number.
The specific changes most likely to move interest, completion, retention, and return, so a "maybe" becomes a fundable "yes."
We don't guess by age and zip code. We map taste (how communities actually consume, connect, and share) and score resonance segment by segment.
Adjust budget, marketing, licensing, audience size, and distribution strategy. The model shows three scenarios and never blurs the line between fact, assumption, and upside.
De-risk development slates. Rank projects by audience opportunity and commercial fit before you commit capital.
Model subscriber acquisition and retention value, not just viewership. Test a title against the communities you're trying to grow.
See the business case beside the creative one: scenario ranges, confidence, and the risks that actually threaten return.
Encrypted uploads, role-based access, project deletion on demand, and no model training on your material without explicit written consent.
Upload a project and see the audience, the scenarios, and the greenlight read in one place. Built to inform human judgment, never to replace it.
The full path — upload, analysis, audience, culture, comps, economics, and greenlight — with the evidence trail intact at every step.
Feature scripts, TV pilots, series bibles, treatments, pitch decks, documentary proposals, unscripted concepts, short-form ideas. PDF, DOCX, TXT, PPTX. Add title, format, genre, budget, target platform, comps, and stage — all optional. Missing inputs become clearly labeled assumptions, not silent guesses.
Premise, genre and subgenre, tone, themes, character strength, protagonist relatability, emotional stakes, structure, pacing, originality, franchise potential, cultural specificity, universal accessibility, entry points, positioning, and candid creative and commercial risks.
An Audience Impact Score from 0–100, with predicted awareness, interest, intent, emotional connection, recommendation likelihood, repeat-viewing, and subscriber acquisition and retention potential — each with a confidence label.
3–6 audience personas per project: demographic and psychographic profile, motivations, objections, the characters and themes that hook them, preferred platform, and the marketing message and channel most likely to convert them.
An assessment against transparent criteria — never reducing a community to a stereotype. Which cultural elements create differentiation, and which need more context, sensitivity, or development before launch.
Relevant comps by genre, tone, audience, budget, cultural positioning, platform, and release strategy — with audience overlap and the specific lesson each carries. Add or remove comps and the forecast recalculates.
Development, production, marketing, talent and rights, contingency against licensing, theatrical, streaming value, international, subscriber value, brand, merch, franchise, and long-tail library — as conservative, base, and breakout scenarios.
A composite score across demand, creative strength, differentiation, budget efficiency, distribution fit, revenue and franchise potential, execution risk, and marketing clarity — plus a plain-language recommendation.
Taste-based segmentation, resonance scoring, personas, and cultural analysis — the layer traditional demographics can't see.
Illustrative read for a fictional TV drama, "Golden Hour." Predicted resonance across selected segments, 0–100.
How resonance is expected to convert down the line — awareness through subscriber value.
Relative size × engagement potential. Bubble size = reachable audience.
Where the material earns differentiation — and where it needs development before launch.
Three scenarios, revenue by channel, break-even, payback, and live sensitivity — with every line traced to fact, assumption, or upside.
Premium limited series · fictional example · all figures illustrative
Modeled gross across windows for the selected scenario.
Move a variable to see net and ROI respond. Illustrative model.
Ranked by impact on modeled net revenue. Tornado read.
A worked example on a fictional premium limited series. Every number is illustrative — included to show the shape of a real report, not to forecast an actual title.
"Golden Hour" is a strong audience opportunity with a financeable base case that needs two creative revisions to de-risk. The Atlanta music-industry world and Afro-Latinx lead give it real differentiation on premium streaming, and the core segment reads exceptionally warm. The commercial case holds in the base scenario but is sensitive to marketing efficiency and international licensing — both improvable. The recommendation is promising with revisions: greenlight-track it, fix the two flagged beats, and re-run before committing full production capital.
Ten dimensions, weighted into the composite read.
Likelihood × impact on return.
Add or remove a comp to see how the audience overlap average shifts.
| Title | Positioning | Overlap | Lesson |
|---|
Ranked by likely effect on interest, retention, and return.
The same analysis answers a different question depending on the seat you're in.
See who your script is for before the room does. Sharpen positioning, spot the beat that's costing you, and enter a pitch with a defensible audience read instead of a logline and hope.
Compare projects on audience opportunity and commercial fit, not taste and instinct alone. Decide what to develop, what to shelve, and what to fix first.
Bring a consistent, transparent read to every greenlight meeting. Model franchise potential and marketing clarity alongside the creative case.
Estimate subscriber acquisition and retention value by community. Test whether a title actually grows the audience you're trying to reach.
Scenario ranges, break-even, payback, and the risks that actually threaten return — with confidence levels you can put in a memo.
Assess cultural fit and resonance before you attach a brand. Avoid the mismatch that reads as performative, and find the partnership that lands.
Comparable-title overlap and international resonance to support licensing conversations and window strategy.
Credible decision support starts with being honest about its own limits. Here's what the platform does, what it can't, and why every output ships with a confidence label.
A private language model extracts creative metadata from your material — premise, tone, themes, character and structural signals, cultural specificity — and turns it into a representation the prediction models can score. Your material is not used to train shared models. Analysis based on your upload is labeled separately from anything the model infers.
The base model is built on real engagement behavior — what audiences actually watched, finished, and shared — validated against historical outcomes, rather than surveys or self-reported preference. The production model maps engagement into 12 taste segments and is scored against multi-million account populations (about 3M in the high-budget ROI case). Training used tens of thousands of account-level viewing histories and 300+ franchise titles with semantic embeddings; holdout Path B performance is roughly MAE 0.10 and correlation ~0.78. We're explicit about coverage: predictions are strongest where the training behavior is densest, and we tell you when a project sits at the edge of what we've seen.
Every score carries a confidence label — high, moderate, or low — driven by how much of the input is evidence versus assumption, how close the project sits to well-observed behavior, and how much the comparable set agrees. Low confidence isn't a failure; it's a signal to gather more before you decide.
Cultural analysis runs against documented criteria and is reviewed to avoid reducing any community to a stereotype. We test for systematic over- or under-prediction across segments and report known gaps rather than papering over them. The platform flags when it's extrapolating beyond represented data.
Can: relative audience resonance, likely personas, comparable-title proximity, scenario-based revenue ranges, and where a project is strong or exposed. Cannot: guarantee a hit, price a specific title's box office to the dollar, or replace the judgment of people who know the room. Individual outcomes in entertainment are high-variance by nature. Use the model to narrow uncertainty and pressure-test assumptions — not to outsource the decision.
The best result is a sharper conversation, not an automated verdict. Executives who use the platform well treat the score as one voice in the room — the one that brings the receipts — and keep final judgment where it belongs: with people accountable for the outcome.
Scripts and bibles are pre-public, high-value, and often contractually restricted. Confidentiality isn't a feature here — it's the precondition for the product existing at all.
Material is encrypted in transit and at rest. Uploads are isolated per project and per account.
You retain full rights to everything you upload. We claim no license to your creative material.
Your material is never used to train shared models without explicit, written, opt-in consent.
Control who on your team sees which projects. Granular permissions for studios and financiers.
Delete a project and its derived analysis. Configurable retention windows for enterprise.
SSO, audit logs, custom security review, and data-residency options for enterprise agreements.