Workflow 2
Beta · by invite

Script Review & Continuation: submit anything, get a report that turns into revisions

Bring any asset — an outline, a synopsis, a half-finished draft, or a full script. Get a five-dimension review report in about 9 minutes, fix issues point by point on a Word-style tracked-changes desk, and hand off to the creation line whenever you're ready.

For: writers with a back catalog, chief editors and story editors who need one review standard across the company

How it works

Done in 4 steps

  1. 1

    Submit

    Upload an outline, synopsis, or any set of episodes. One screen, zero configuration.

  2. 2

    Read the report

    Automatic intake plus review: an editor's synopsis card, five score bars, and issue triples — about 9 minutes for a full script.

  3. 3

    Fix point by point

    Drill into each issue on the revision desk; AI proposes diffs, you accept or reject. Incremental re-review in about 70 seconds, with score changes marked.

  4. 4

    Keep writing

    Once it clears the bar, hand off to 0-to-1 Creation in one click to continue episodes or export the script.

Core capabilities

Why it's different

A five-dimension review engine

Throughline & characters / pay points & structure / hooks / open loops & setups / dialogue & shootability. Two of the five run as deterministic machine checks against beat maps and ledgers — zero tokens, fully reproducible.

Every issue ships as a triple — no hand-waving

Each finding must include all three: quoted evidence (locatable in your text), supporting data (e.g. "2 dialogue lines this episode vs. a series average of 14"), and a sentence-level model fix.

A tracked-changes revision desk

Issue map on the left, an inline del/ins revision scroll in the middle, AI chat on the right. Generate a dimension's revisions in one click, accept or reject them one by one; incremental re-review returns a comparison in 70 seconds.

Review is generation run in reverse

Review and creation share one data contract — every conclusion in the report links straight back into the creation line, instead of dying as a read-once PDF.

Inside the review

Five dimensions — two of them free, deterministic machine checks

The rubric was validated with paired human-vs-AI reviews across 13 manuscripts and 9 genres, then back-tested against a production-grade script that went through four rounds of chief-editor iteration.

① Throughline & characters

A chief editor's read-through: does the premise hold, do the arcs and emotional line land — with per-episode summaries as evidence.

② Pay points & structure

A machine check against beat maps: is each pay point cut on the eve of an act climax? Zero tokens, fully reproducible.

③ Hooks

Per-episode deep review of opening and closing hooks — type and strength, with a write-in path wherever no hook is detected.

④ Open loops & setups

A setup/payoff ledger heuristic: planted but never paid, paid but never planted — reconciled line by line. Zero-token machine check.

⑤ Dialogue & shootability

Episode length and dialogue density checks: in a one-minute episode, every line has to earn its shot.

Every point is recomputable: deduction rules are stated explicitly — the report never hands you a score based on vibes.

FAQ

About Script Review & Continuation

How is this different from pasting my script into ChatGPT?
Three ways. First, the scoring dimensions are bound to short-drama paywall structure (pay points, hooks, open loops), not generic literary critique. Second, every issue is forced to carry quoted evidence and supporting data you can verify and locate. Third, the report is wired into the revision desk and the creation line, so conclusions become edit actions immediately.
Can I trust the scores?
The rubric was validated with paired human-vs-AI reviews across 13 manuscripts and 9 genres, then back-tested against a production-grade script that went through four rounds of chief-editor iteration; the two machine-checked dimensions are deterministic and reproducible.
Can I submit a half-finished script?
Yes. Any combination of outline, synopsis, and partial episodes works. The report explicitly labels what was scored on written episodes versus what remains plan-level — unwritten content never pads the score.

Run your next script through the data first

The invite-only beta is open. Tell us about your team and use case — we'll get you set up shortly.

Every shipped capability is free to use during the beta