LaperDocs

Quick start: idea to first scene

Create a Script project from a premise, review the AI plan, and reach a clean first scene without losing control.

Start with a useful brief

Open New project and describe the story you want to make. A productive brief gives the assistant four things:

  1. The protagonist and what they want.
  2. The force opposing them.
  3. The intended format or scale, such as a short film or episodic series.
  4. A tonal or thematic constraint you care about.

You may attach DOCX, PDF, TXT, or Markdown reference files. Attachments are context for the opening conversation; they are not automatically copied verbatim into the screenplay.

Let the project open before judging the result

Laper creates the project, opens its collaborative document, then starts the initial AI turn after the document and local cache are ready. The assistant may establish a title, outline, beats, core characters, locations, and opening scenes in stages. The exact order can vary as the writing agent evolves, so evaluate the resulting story structure rather than expecting a rigid wizard.

Review in the right places

  • Read Outline for the overall dramatic promise and sequence.
  • Read Beats for turning points and units of change.
  • Return to Script to judge what is actually on the page.
  • Open Characters and Locations to check the entities derived or created for the project.

If the direction is wrong, correct it early: “Keep the isolated lighthouse, but make the protagonist the missing keeper's daughter. Revise the outline only; do not write new scenes yet.” A bounded correction is safer than asking the AI to redo everything.

Write the first scene deliberately

Ask for a specific scene goal, conflict, and exit condition. For example: “Write the opening scene in the current episode. Establish the storm through filmable action, let Mara hide why she came, and end when the radio says her own name.”

After the scene appears, inspect the scene heading, action, character cues, and dialogue. Fix structural mistakes at the text source. Use version history before a large experiment, and keep requests small enough that you can compare intention with result.