Reader archetypes, not demographics
Four taste archetypes clustered from a public film-preference dataset. Each one describes a genre lean — the model carries no age, gender, region, or income attribute.
AIP
Audience Intelligence Platform
AI-powered audience intelligence
A simulated reading room for your screenplay.
Upload a script and it is read scene by scene by a swarm of reader personas and a craft ensemble. AIP then surfaces where those readers disagree with each other, rather than averaging them into a single verdict.
How it works
A labeled reader simulation, run over the actual text of your script.
Your screenplay is parsed into a scene graph — scenes, beats, characters, dialogue, action. A swarm of reader personas reacts to it in character, while a craft ensemble of structure, dialogue, pacing, and theme specialists works the same graph. Both are held to the page: a reaction that cannot cite the scene it came from does not make it into the report.
Reader archetypes
Taste profilesBlockbuster & Family
Leans toward Adventure, Action, Fantasy, Children
Action & Genre
Leans toward Action, Sci-Fi, Crime, Film-Noir
Arthouse & Dark
Leans toward Drama, Mystery, Horror, Documentary
Eclectic & Horror
Leans toward Comedy, Musical, Western, Horror
Clustered from a public film-preference dataset. An archetype is a taste profile — it stands for no real person and carries no demographic attribute. You can also invent readers of your own in a theater and give them any attributes you like; those are your description of someone, not a measurement of anyone.
Grounded in your own pages
Every reaction cites the scene it came from and quotes the line that caused it. A claim whose citation does not resolve to real text in your script is rejected instead of shown.
Divergence, not consensus
The report keeps the spread: which archetypes lean toward the script, which lean away, and the exact scenes where they part company. Dissent is never averaged out.
Self-hosted inference
Your script is parsed and read entirely on infrastructure we run ourselves. Its contents are never sent to a third-party model API — not for analysis, not for training.
No fabricated certainty
AIP does not hand you a number for real-world reception. Nothing here has passed held-out validation, so the platform ships qualitative signal rather than a figure it cannot stand behind.
Repeatable by design
Deterministic stages are cached by content hash, so the same script bytes yield the same parse and the same scene graph every time. No silent drift between runs.
Get started
Upload a draft and read the disagreement.
Create an account to run a script through the reader swarm and the craft ensemble. You get a divergence-first report tied to your own scenes — not a verdict, and not a number standing in for a real audience.
Our network
Partners
Working together with organizations that advance innovation, technology, and audience intelligence.
