Structured nodes: description, character, dialogue

Two ways, whichever suits the hand you are free with: press Tab on any script line, or type / at the start of an empty line. The menu floats out of the current line as a row of type chips.

While it is open the line is read-only on purpose—that is what keeps the digits you press in the menu instead of dropping them into the text. Back out with Esc or just keep typing.

The order is fixed, so 1–9 map straight onto it: 1 Description, 2 Character, 3 Dialogue, 4 Comment, 5 Dialogue action, 6 Narration, 7 Transition, 8 Scene setting, 9 Scene. Move the highlight with ↑↓ (or Tab) and confirm with Enter. Clicking works too; number keys just keep you inside the writing rhythm.

The easiest to misuse is Character: it only declares who speaks the lines that follow. Someone entering, looking out of the window, saying nothing—that is a description line. Use Character as an appearance marker and the appearance counts go wrong.

• Enter on a character line creates his dialogue; Enter on a dialogue line stays dialogue, so a run of lines needs no re-picking.
• Enter on any other type returns to a description line.
• Action is an insertion inside dialogue: the legal shape is character → (any number of actions) → one or more dialogues. An action cannot create a speaker.
• A note ends the current speaker—it is words to yourself, not part of the dialogue.
• On plain lines, ↑ ↓ at the first or last row steps into the previous or next node; scene and scene setting nodes are multi-field, so Enter walks the fields first and only starts a new line after the last one.
• Enter on an empty line inside a scene leaves the scene; Backspace on an empty line deletes it and hands focus back.

Once a line has a type, what you can fill differs by type—knowing it beats trial and error:

• Content blocks (Description, Character, Dialogue, Comment, Dialogue action, Narration, Transition): the text itself plus one line of intent. A character line also links to an actual card, and bracketed prop names in a line are recorded on that line.
• Scene: location (picked from the location library, or a new name that files a card), time (free text), a short title, a scene summary, and numbered beats—the beats are what later prose expansion works from. Summaries carry their origin: one produced by maintenance can be refreshed, one you typed is never overwritten.
• Scene setting: location and time, a description, and the in-story timeline index, label and note that hang this moment on the timeline. Its location may only be the scene’s own location or a descendant of it.
• Chapter: title, chapter intent, and a colour. Chapter numbers come from the current order, not from the title—so a fresh chapter showing “Chapter 3” with an empty title is correct behaviour.

Scene and scene setting are multi-field nodes: Enter walks through the fields first and only starts a new line after the last one, so the mouse can stay out of it.

Changing type only swaps the semantic mark on that line; nothing you typed moves. So try freely. What does change is structure: making a line a scene pulls the lines after it into that scene, and deleting it removes the scene and renumbers the rest.

Two ways back: undo is the app-wide project stack (Cmd/Ctrl + Z, up to 30 whole-project steps, with typing inside one second collapsed into a single step), and every edit is written locally first, then queued for sync, so a dropped connection never costs you finished words.

• The menu belongs to script lines only. Tab inside a card field (character name, place description) is native input behaviour and will not open it.
• The legacy comment, parenthetical and caption types are gone with no read-compatibility: that content is written today as Comment, Dialogue action, or Narration.
• Selecting across lines then copying yields plain text — no node types or styling. Copying inside one node keeps your exact range.
• Do not see a type? The row scrolls: ← → or the number keys still reach it.

Full semantics and the page layout are in Script editor overview.