Chapter 24 · Formats
The Table Read
A page you have read forty times stops telling you anything. Hearing it spoken by someone other than the voice in your head puts the clumsy line back in front of you. Storskarvia casts a voice per character and reads your pages out loud, highlighting each block as it goes. This chapter covers what it speaks, what it deliberately does not, and the places where the cast list and your script disagree.
Where the button is
Open the Review tab of the ribbon. The third group holds Research, Insights, and the read button, which is labelled Table Read in a screenplay and Read Aloud in every other format. Clicking it opens a small dialog with the cast at the top, a Pace slider under that, and the transport buttons along the bottom.
Nothing is sent anywhere. The speaking is done by your browser's own speech engine, on your machine, which is why the available voices depend on the computer you are sitting at rather than on your account.
Casting the voices
The dialog lists one row per role, each with a dropdown of the English voices your browser exposes. The first row is always NARRATOR. In a screenplay, the rows below it are your characters, up to fourteen of them, ordered by how many cues each one has. The busiest part in the document is first.
Storskarvia pre-fills the dropdowns by walking down your voice list and handing out a different voice to each role until the list runs out, then starting again from the top. That is a starting point rather than a suggestion, and on a machine with three voices and nine characters it will repeat. Change any dropdown and the choice is saved to that document straight away, so the casting is waiting for you next time you open it. If the save fails, the status bar reads That setting did not save.
Character names are matched with their extension removed and their case flattened, so
MARIA, Maria, and MARIA (V.O.) are one role with one voice.
That is deliberate: a character on the phone is still the same actor. Renaming a character is
covered in Chapter 18, and a rename moves the cues
rather than the casting, so re-open the dialog afterwards and set the new name's voice.
What gets spoken, and in whose voice
Pressing Play rebuilds the reading queue from the document as it stands, then works from the top of the page down. In a screenplay, each block is treated like this:
| Block | What happens |
|---|---|
| Character cue | Not spoken. It sets who speaks the dialogue that follows. |
| Dialogue | Spoken in that character's voice. |
| Parenthetical | Skipped entirely. |
| Scene heading | Spoken by the narrator, announced as "Scene." followed by the heading. The speaker then resets, so a stray dialogue block after a new heading is read by the narrator rather than by whoever spoke last. |
| Action, transitions, lyrics, everything else | Spoken by the narrator. |
| Empty blocks | Skipped. |
The queue is built fresh on every Play. Fix a line, press Play again, and the new line is the one you hear. There is no cached recording to invalidate.
What is skipped
Two omissions are on purpose. Parentheticals are directions to an actor, so speaking them would add words to the scene that nobody in the scene says. Character cues are labels rather than lines, so they steer the voice instead of being read out.
One omission is a limitation rather than a decision. A dual-dialogue block, the side-by-side pair covered in Chapter 16, is one block as far as the reader is concerned. Its two columns are read together, in the narrator's voice, in the order they appear. Two people talking over each other is exactly the thing a synthesised read cannot do, and the honest workaround is to split the pair back into ordinary cues while you listen, then rejoin them.
Pace, play, pause, stop
The Pace slider runs from 0.7 to 1.3 in steps of 0.05, and starts at 0.95, a shade under conversational. Slower is worth trying on dialogue you suspect of being overwritten, because a line that survives 0.7 is genuinely clean.
Below it are four buttons:
- ▶ Play starts from the top of the document.
- ⏸ Pause holds the read where it is and becomes ▶ Resume. The button label tells you which state you are in.
- ⏹ Stop ends the read and re-enables Play.
- Close stops the read and closes the dialog. Closing never leaves a voice talking behind a shut dialog, and neither does navigating away from the editor.
Pace is read at the moment each block starts speaking, so moving the slider mid-read takes effect on the next block rather than the current one.
Following along on the page
As each item starts, Storskarvia selects the block being spoken, scrolls it to the middle of the page, and flashes it briefly. The selection covers the first 200 characters of a long block rather than the whole thing, so a page-long paragraph highlights its opening rather than sweeping the screen.
The practical consequence: the read moves your cursor. If you want to fix a line you just heard, press Stop first. Typing into the manuscript while the read is running is possible, and it means the queue you are hearing and the text you are editing have drifted apart until the next Play.
Outside screenplays
Every format gets the button, and outside screenplays it reads Read Aloud and casts a single voice.
| Format | What you hear |
|---|---|
| Novel | Every paragraph and heading in the narrator's voice, in order. A list, a blockquote, or a table is one item rather than several, so its lines run together without a pause between them. Footnote text is not read, and images and page breaks are passed over. |
| Stage play | The whole page in one voice, character cues included. The cue is a block like any other here, so you hear the name announced before its speech. |
| Audio drama | One voice throughout, and the sound and music cues are read out along with the dialogue. Useful for timing, blunt for performance. The spoken-word runtime figure in Chapter 20 is the better tool for length. |
| Video script | In the one-lane layout, a visual block is announced with the word "Visual:" before its text, so you can tell picture from voice by ear. In the two-column layout, each row's picture column and audio column are read together as a single run. Chapter 21 covers the two layouts. |
The cast list still appears in these formats, holding the single NARRATOR row. Setting that voice is worth thirty seconds, because it is the voice you will hear for the entire document.
When it does not sound right
- The dropdowns are empty. Your browser is exposing no English voices. The read still plays, in whatever voice your system uses by default. On most machines the fix lives in the operating system's speech or accessibility settings rather than in the browser or in Storskarvia.
- The list was short and then filled in. Browsers load voices asynchronously. If you open the dialog the instant the page is ready, the cast can rebuild itself a moment later with the full list. Re-check your casting if it appears to have changed.
- A character is read in the wrong voice. Only the fourteen busiest speakers get a row. Anyone below that line, and anyone who only speaks inside a dual-dialogue block, has no assignment and falls back to the default voice.
- A block was silent. If your browser's speech engine fails on one item, the read moves to the next one instead of stopping. A silent block in the middle of an otherwise working read is that, rather than a missing line in your script.
- The casting is different on another machine. The choices are stored with the document, and the voices are not. A voice your other computer does not have cannot be used, so that role falls back to the default there.
The read is a listening tool rather than a proof. For questions the ear cannot answer, whether a character's eye colour changed between chapters or whether a claim in the dialogue is true, the staff in Part VI are the ones to ask.
Updated 2026-08-31 · Full contents · Something wrong or missing?