Screenplay import

The Screenplay page. The Screenplay button on the rail opens it as a tab beside your boards, and leaves the list on the left as it was. It brings a written script in and turns it into boards.

Which formats parse

Two, each with its own parser. There is no third-party screenplay library involved.

Format	Extensions	How it reads
Fountain	.fountain, .spmd	Heuristically — the format has no explicit element types, so structure is inferred
Final Draft	.fdx	Structurally, from the file’s own element types

The file picker accepts exactly .fountain, .fdx and .spmd, and dispatch is by extension, never by the MIME type the browser reports — browsers call a .fountain file plain text.

All eight of Final Draft’s paragraph types are mapped. An element type that is not recognised is dropped and reported, never dropped silently.

What a scene becomes

Element in the script	What it becomes
Scene heading	The board’s name — a node only if you opt in
Character cue	Not a node. It names the speaker of the line below it.
Dialogue	A node, with a speaker
Parenthetical	Folded into the speech beneath it, in italic
Transition	A node, in bold
Action, centered, shot, general	A node, plain

What the import writes

• A folder named after the script — only when there is more than one board.
• One board per group, described by the writer’s own scene summary where Final Draft carries one, and Scene N otherwise.
• The nodes, and edges chaining them in order.
• Entity items for the characters, created first, after which the boards are re-planned against them.

How a speaker becomes an entity

Every speaker is matched against every entity in the project. Not against one collection you nominate, and not against whichever kind of node you routed the lines to — a script’s cast is a property of the script.

Matching is by name, in this order: an exact name, then the entity’s code, then a single content word. A tie is not a match: as soon as two entities qualify on that last rule, the cue is left unmatched and reported, rather than resolved to whichever came first.

Where it looks	Every entity, in every collection
What it decides	Which entity a line’s speaker refers to, what the preview draws, and which characters add the missing ones would create
What it does not decide	Whether the match is written as a reference or as the literal name — that is the routing’s job

Entities are optional. A speaker can go to an entity reference field or to a plain or rich-text field holding the literal name; requiring an entity per character was considered and rejected. With the name going to a text field the cast is still matched — the preview still names the speaker, and the colours still land — the line simply carries text instead of a reference.

The controls

One surface, not a wizard.

Empty state — Bring a script in, with Every scene becomes its own board, named after its slug line. Drop a .fdx or .fountain file here, or choose one. The whole panel is a drop target.

Loaded — Choose another script and Clear, plus a line reading Final Draft · N scenes or Fountain · N scenes.

Left column: two tabs

Scenes. Select all / Select none, a Only scenes with dialogue filter with a count, and a Table / Cards toggle. Every scene is selected by default.

Board grouping. How scenes are gathered into boards:

Option
Every 2 / 3 / 5 / 10 pages
Every 2 / 3 / 5 / 10 scenes
The writer’s sections
One board per scene

Plus Merge boards under and Split boards over thresholds. The default is every 5 pages, chosen from a corpus of over twelve thousand scripts rather than by taste.

Routing. Each part of a script is routed to a node type and a field:

Scene names · Sequence notes · Spoken lines · Description · Transitions · Shots · Other lines

The routing is derived automatically from your project’s own node types, and a routing you like can be saved as a named preset in the project’s settings.

Right column: a live preview

It mounts the real canvas and runs the real planner and writer into a throwaway document. It is not a mockup of the result; it is the result, discarded.

The action bar

A Layout choice — Vertical or Horizontal — and then Import N boards.

Afterwards: a result line, Open the first scene, how many characters were created, and any warnings.

Nothing is lost silently

Before you press the button, the panel lists what it could not use: what the file named that TaleShard cannot read — in the file’s own spelling — what you chose not to route, and every unrouted piece of data.

What is not read

Fountain: synopses (=), page breaks (===), and dual dialogue — the marker is stripped and both speakers appear in sequence. Sections (#) are not elements; they survive only as the input to the writer’s sections grouping mode.

Final Draft: Singing (reported by name rather than guessed at), revision marks, colours, fonts, strikeout, dual dialogue, and the cast list. That last one is deliberate: the file’s cast list is autocomplete history, not a roster — in the reference script it holds 66 entries against 81 real character cues, polluted with transition text and a typo.

Two other doors

• The Documents section’s importer accepts these extensions too — but produces a prose document, not boards.
• The AI board generator recognises a Fountain or Final Draft document and, when you leave its instructions empty, converts it deterministically, with no model call — so that path costs no AI credits.

Limits

Nothing caps scenes, boards or characters. The only gate is the Pro plan requirement.

What breaks

Situation	What happens
A .pdf	Not accepted. There is no PDF path.
A screenplay imported as a Document	Plain body text, no scenes and no boards
An element type the parser does not know	Dropped and reported in the pre-import summary
A part you did not route	Listed before you import, with what it holds
Dual dialogue	Both speakers appear in sequence, not side by side
A board that cannot be created	Reported, not thrown — the rest of the import proceeds
You want it back out as a script	Not possible. There is no export.