Inline text markup

Markup is part of the stored text. It round-trips through every format as plain characters, and a consumer that cannot honour a tag should strip it for display rather than rewriting the stored value.

Reference

Markup	Where	Meaning	If you cannot honour it
`\|`	Dialogue Text	Sentence split. `"One.\|Two."` is shown as two successive lines, not one paragraph. Audio files pair with segments by position.	Replace with a space or newline.
`[f]`	Menu Text	Force a menu even when this is the only offered choice.	Ignore for display; honour in traversal.
`[a]`	Menu Text	Action line — the choice is something the player does, not says (Walk away). Desktop renders it in italics.	Strip; optionally style.
`[var=Name]`	Either	Substituted with the variable’s current value at display time.	Substitute; or show `{Name}` as the production exports do.
`[var=?Name]`	Dialogue Text	Input prompt (Desktop’s spelling). Pause, ask the player, store the answer in variable `Name`, remove the token from the displayed line.	Strip, and treat the variable as unset.
`[?Name]`, `[x=?Name]`	Dialogue Text	Same as `[var=?Name]`; anything left of `=?` is ignored.	As above.
`[pic=N]`	Dialogue Text	Picture swap. Show the node’s own `Pictures` entry `N` (1-based) instead of the actor’s portrait, from this point in the line.	Strip.
`[pica=N]`	Dialogue Text	Show the actor’s `Pictures` entry `N`.	Strip.
`[picc=N]`	Dialogue Text	Show the conversant’s `Pictures` entry `N`.	Strip.
`[em1]…[/em1]` … `[em4]…[/em4]`	Either	Emphasis span. Four project-defined styles (colour, bold/italic/underline, label).	Strip the tags, keep the text.
`[gesture:…]`, `[expression:…]`, `[vocal:…]`, `[mood:…]`, `[pose:…]`	Dialogue Text	Performance cue. What a character does at this point in the line. See Performance cues.	Strip.

Tags are case-insensitive. A bare [pic] with no = is not a tag and is shown literally.

Rules that matter to a consumer

Strip, never rewrite. The production exports (voiceover CSV, proofread text, screenplay) strip every tag and show [var=Name] as {Name}; they never save the stripped text back.
Split before you strip. | is a segment boundary; [var=…] and picture tags are per segment. Split first, then process each segment.
Input tokens are consumed. After the player answers [?Name], the displayed line has no trace of the token; to echo the answer, the author writes a separate [var=Name].
[f] and [a] are choice flags, not text. Strip them from menu labels. [f] changes traversal (see runtime semantics); [a] does not.
Emphasis is only honoured when matched. A properly matched, non-nested [emN]…[/emN] pair is a styled run; an open tag with no matching close is treated as plain text. Every tag literal is consumed either way, so tags never leak into displayed text.
| inside a bracketed tag is a bug the Unity preflight warns about: Dialogue System splits on | first and would tear the tag in half. Keep | outside brackets.

Performance cues

A performance cue says what a character does while a line is spoken: a facial expression, a body gesture, a vocal sound, a change of mood or a change of pose. It is written where it should happen.

[channel:name]
[channel:name,intensity:0.7,target:conversant]

Part	Values	Meaning
`channel`	`expression`, `gesture`, `vocal`, `mood`, `pose`	What kind of cue it is. `expression`, `gesture` and `vocal` are momentary. `mood` and `pose` hold until changed.
`name`	a vocabulary name, such as `surprised`, `shrug`, `angry`, `sit_chair`	What to perform.
`intensity`	`0` to `1`, optional	How strongly.
`target`	`conversant` or an actor’s name, optional	Who performs it. The default is the node’s speaker, so a listener can react to what is being said. Quote a name that contains a space or comma.
`at`	milliseconds, optional	Exact time from the start of the segment’s speech. Written by exporters when audio timing is known; do not author it by hand.

= is accepted in place of : after the channel, and tags are case-insensitive.

A cue belongs to the | segment it sits in. Its position is its character offset in that segment’s spoken text, the text left after every tag is stripped. A player that can time cues fires each one as that point in the speech is reached; one that cannot fires it when the segment starts.

Dialogue Text:  I never [gesture:shrug]agreed to that.|[expression:angry,intensity:0.8]You know I didn't.
Displayed:      I never agreed to that.
                You know I didn't.
Cues:           segment 0, offset 8  → shrug, as "agreed" is spoken
                segment 1, offset 0  → angry, as the second line starts

The same performance can be set for a whole line with the node’s custom fields Expression, Expression_intensity, Gesture, Mood and Pose, and a character’s baseline with the actor’s Mood and Pose. Fields set the default; inline cues add beats inside the line.

Rules for a consumer:

Never speak or display a cue. Strip it before text reaches the screen or a voice provider.
Unknown names are a no-op, never an error. A player performs what it can and ignores the rest.
Cues in Menu Text do not fire. Put the cue on the line that is spoken.
A cue never outlives its segment. Stop a momentary cue when the segment’s speech ends.

Emphasis styles

The four emphasis slots are defined per project — colour, bold / italic / underline, and a label (“Stage direction”, “Shouted”). In JSON they live in _editor.emphasis; in XML they are the EmphasisColorN, EmphasisStyleN and EmphasisColorNLabel attributes on the root element, exactly as Desktop writes them. The style string is three characters, one per flag, e.g. b-- for bold only, bi- for bold italic.

A consumer that does not render emphasis loses nothing but the styling.

Examples

Dialogue Text:  Last crossing of the day.|Coming aboard, or staying with the rain?
Displayed:      Last crossing of the day.
                Coming aboard, or staying with the rain?

Dialogue Text:  What do they call you?[?PlayerName]
Displayed:      What do they call you?        ← then an input box; answer → Variable["PlayerName"]
Next line:      Welcome aboard, [var=PlayerName].
Displayed:      Welcome aboard, Ines.

Menu Text:      [f]Continue
→ a one-button menu even though it is the only option

Menu Text:      [a]Walk away
→ label "Walk away", flagged as an action

Dialogue Text:  I said [em2]no[/em2].[pic=2] Get off my boat.
Displayed:      I said no. Get off my boat.   (with "no" in emphasis style 2; the node's second picture shown from "Get off")