Cast the transcript by typing
The transcript understands one piece of syntax on top of plain text: a character's name in brackets at the start of a line. It's the fast-typing way to do what the Character box and the Characters menu do, and Auto-split can turn a whole tagged transcript into clips at once. This page covers what a tag looks up, where the FX half works, and what happens when a name doesn't match.
01 · SyntaxA name in brackets, then a colon
A tag sits at the start of a line. Everything after it, up to the next tag or the end of the text, belongs to that character.
[Marta]: Spoken by the character named Marta. [Marta:Radio]: Marta again, with the FX named Radio.
The name is a character, not a voice ID like af_heart. It's matched against the project's characters (Edit > Characters...) by name, ignoring case. A character carries the voice, the engine and the FX, so one tag picks all three.
02 · Typing a tagFinish the line and it's assigned
Type [Marta]: Who's there? and press Enter. If the project has a character named Marta and the line isn't already part of a clip, the line becomes Marta's: it's tinted in her color, the gutter names her, and it lands on her track. It's one undo step, like picking Marta from the Character box.
A line that already belongs to a clip isn't reassigned when you revisit it, so editing an old tag doesn't silently move text between characters. Use the Character box or the Characters menu (right-click a selection) to change it.
03 · Auto-splitA whole transcript in one action
Paste a tagged transcript and choose Generate > Auto-split then generate. Every tagged block becomes a clip for its character, on that character's track, and then every stale clip generates.
With Split by paragraph on (in the same menu), each block is also cut at blank lines, so a [Narrator]: block of three paragraphs becomes three clips. Untagged text only becomes a clip when the project has exactly one character; otherwise it's left unassigned.
Within each clip, text is still generated in pieces of about 40 words, cut at paragraphs, sentence ends or pauses (Settings > Target words per segment). Tags decide who speaks; the splitter decides where the files break.
04 · Worked exampleFour lines, three characters
Assume the project has characters named Narrator, Marta and Tomas.
[Narrator]: The house at the end of the lane had been empty for years. [Marta]: Then why is the kitchen light on? [Narrator]: She pushed the door. It didn't creak. [pause:0.8] [Tomas]: Marta, come in. Marta?
Auto-split makes four clips: two on the Narrator track, one each on Marta's and Tomas's. Tomas's clip starts 0.8 s after the one before it instead of the usual gap.
05 · The FX halfWhere [Name:FX] applies
The part after the second colon names an FX saved from the Audio FX tab with its "Save FX Preset..." button. Today it only takes effect when the project has no clips yet and Generate runs the whole document at once (see below). When a tag assigns a clip, by typing or by Auto-split, only the character is used.
To give one line an effect in a project with clips, put the caret in the line and pick the FX from the FX box above the transcript. That sets it for that clip alone, and the gutter shows FX: Radio under the speaker's name.
06 · When a name doesn't matchA typo skips a block, it doesn't stop the run
Typing a tag with an unknown name does nothing: the line stays unassigned until you create the character or fix the name.
Auto-split skips every block whose name matches no character and tells you which names it couldn't find, then splits and generates the rest.
07 · Pauses[pause:1.5] adds silence before the next clip
Put [pause:1.5] on its own line or inside a line, with the seconds you want. When Auto-split runs, the marker stays in the transcript as plain, unassigned text, so it's never read aloud, and the next clip starts that many seconds after the one before it. A marker in the middle of a tagged block splits the block into two clips there.
Without a marker, clips placed one after another still get a gap: 0.35 s by default, 0.9 s across a blank line. Both are in Options > Settings… > Project (Gap, Paragraph gap), and a clip's own Gap before overrides them. A clip locked in time ignores gaps.
08 · Projects with no clipsThe whole-document path reads preset files
Before a project has any clips, Generate (and JIT streaming, when it's on in the Options menu) runs the whole transcript through the engine in one go. That older path resolves tags differently:
| Tag | What it does |
|---|---|
[Name]: | Loads presets/Name.json over the Settings tab's project values for that block. Blocks don't inherit from each other: each starts from the project values again. |
[Name:FX]: | The same, plus the FX named FX, from the project's own fx/ first, then presets/fx/. |
[:FX]: | No preset file, only the FX: the project's voice with an effect on it. |
| A missing preset | That block uses the project values and the status line says Warning: Preset 'Name' not found. A missing FX says the same and the block plays dry. |
[pause:x] | Dropped, and no silence is added. |
Characters and presets/*.json share names on a fresh install, because the first launch copies each preset file into the character library. Once you rename or edit characters, the two can drift, which is one more reason to Auto-split into clips.
The Lexicon applies on both paths, to every block, whoever speaks it.