SpokenWords

Narrator's guide

As a narrator, you record the manuscript in the Studio. SpokenWords transcribes each take and aligns it with the text, allowing words to highlight during playback. You can listen back, re-record, replace part of a sentence, trim takes and adjust your microphone.

In narrator view, the recording overview stays at the bottom of the Studio. Playback controls are above the waveform; the record button and recording-mode selector are below it. Leads, producers and administrators can choose Narrator in the role switcher to preview this view.

Your production card

The Productions dashboard shows a card for each production assigned to you. New invitations appear under Incoming, with Accept and Decline buttons. After you respond, the card moves to the appropriate group and its ⋯ menu becomes available.

The ⋯ menu has two items. Production roles lists everyone assigned to the production, their roles, invitation responses, assigned chapters, and deadlines. You can view the crew but cannot change assignments; those controls belong to the lead or an administrator. Production details shows the full description and details of your assignment, including your role, deadline, and assigned part of the publication.

The recording overview & listening

The overview shows a group of waveform bars for each sentence, coloured by status. Use it to navigate the section and control playback.

Play, skip, and change speed

Click a sentence to make it active, then press play (or Space) to listen from there. The play button becomes a pause button while playing; the spoken word highlights as it is read, and the time readout shows your position and how much of the section is left. Use the skip buttons to jump a sentence at a time, and the speed menu to slow down or speed up (e.g. 1.5×) without changing pitch.

While playback runs, ← and → seek 5 seconds back or forward, and ↑ and ↓ select the previous or next sentence, so playback carries on from the start of that sentence. Esc pauses.

Navigate by waveform

Click any bar in the waveform to jump to that sentence. Double-click to zoom to fit, scroll or use the zoom slider to zoom in and out, and Shift+scroll to pan. The hint line under the canvas always documents these gestures.

Chapter or whole-publication scope

The scope toggle switches the overview between the current chapter and the whole manuscript, so you can gauge overall progress. (Trimming is only available in chapter scope.)

Recording

Read on, one sentence after another

A recording starts at the selected sentence and runs on for as long as you read. Press → or N, or click Next sentence to the right of the record button, when you finish a sentence to move on to the next one; at the levels that mark your reading, the marking also follows you by itself. Next sentence shows while the microphone is armed and works while a recording runs. If you stumble, press Backspace: what you read of the sentence is thrown away and you start it again, without stopping the recording. Press Space or R when you want a break. The recording stops, keeps what you read, and selects the next sentence once the take is saved, so the next recording starts where you left off. To stop and stay on the sentence you just read, press ← instead; → then records it again. U, and Record or Record again in a sentence's menu, start recording at that sentence and run on, just as R does. Select a sentence with ↑ / ↓ and play it with Space. When a sentence has focus, Enter opens its menu.

Arm the microphone

Before you record, press Arm microphone to the left of the record button (or Shift+R). The microphone opens and the level meter appears at the top right of the text, so you can check your level before you say a word. The record button and R, →, ← and U work only while the microphone is armed, so a key pressed by mistake, or a word picked up by speech input, never starts a recording. The microphone stays armed between recordings, and every recording starts on a microphone that is already listening. Press the button again, or Shift+R, to disarm it, which also stops a recording in progress and keeps what you read. Esc disarms it too when nothing is playing. Moving to another chapter or reloading the page disarms it as well. Choosing Record or Record again from a sentence's menu arms the microphone by itself.

While the microphone is armed, a large [ opens the selected sentence: that is where your next recording starts. It moves with the selection, also while you record.

Record and stop

Select where you want to start and press the red record button (or R or →). The button starts pulsing and turns into a stop button, and the active sentence's bar shows a live microphone waveform that tracks your reading. Every way of stopping keeps what you read; they differ in which sentence is selected afterwards:

While you record, the book's own links — footnote references, links to another chapter and web addresses — do not open anything. Clicking one counts as clicking its text: in another sentence it stops the take and selects that sentence, and in the sentence you are reading it changes nothing. The links work again as soon as the take stops.

However you stop, play starts from the first sentence of the take you just recorded, so you hear it back. If you select another sentence while the take is saved, play starts there instead. To punch in after Space, press ← to select the sentence you just read, then Space to play it, and R where the new reading should begin. Disarming with Shift+R or the arm button starts play from the first take you recorded since arming.

Point at the cloud in the page header to read the take's progress after you stop: Saving the recording while it is written to this computer. While the microphone stays armed, the cloud is dimmed and says 1 recording uploads when the microphone is closed (with the number of recordings waiting). Once you disarm, it says Uploading 1 recording until the take is on the server, and Everything is uploaded when none is left. At Disabled the sentences read as recorded at once. At the other levels a sentence is tinted yellow-grey, with a shimmer across it and a pulsing status dot, while its words are recognized, and it reads as recorded once the finished transcript is back. While a transcription runs, the statusbar shows Transcribing the recording with the time left.

Watch your level

The level meter has a green target zone, and shows TOO LOUD while the input clips. After you stop a take, the status line warns about its level: a weak-signal warning when your signal stayed too low for a few seconds, and Too loud when the take clipped anywhere, even on a single plosive, which the sentence's status dot shows too. A screen reader reads the warning once. Its Microphone settings button opens the microphone settings at the gain slider, and the warning stays until your next take or until you disarm. Once your reading has reached a good level, pauses do not raise the weak-signal warning. If the microphone is muted, the meter shows MUTED, the take stops, and the microphone stays ready: unmute it and press Record to go on. An unplugged microphone is reported, and you arm it again after plugging it in.

Fixing a take

Record a sentence again

Choose Record again from a sentence's status-dot menu or from a reviewer's note to record from that sentence, or press U to record from the selected sentence. Each starts an ordinary recording at that sentence: stop after it to replace only that sentence, or carry on into the next ones with → or the marking, as in any recording. A reviewer's note sits right after its sentence, so it shows where to stop. Stop the way you stop any recording: the record button, Space or R keep the take and select the next sentence; Esc keeps it and disarms the microphone.

Try a sentence again from the keyboard

While you record, these keys let you try again without reaching for the mouse. No key that stops a recording discards what was recorded.

← →Record the current sentence again
← ← →Record again from the sentence before it
Backspace or UStart the sentence again without stopping
← SpaceStop and listen from the first sentence of the take

Punch in mid-sentence

While listening with the microphone armed, press the record button (or R) to start re-recording from exactly where playback is — so you fix only the tail of a sentence and keep the good start.

Delete a take

Choose Delete recording from the status-dot menu to remove the take immediately, without confirmation. No key deletes a recording. A message confirms that the recording was deleted. The sentence returns to not recorded, or another take becomes active if one is available.

A sentence that came with Pre-recorded audio offers Delete pre-recorded audio in the same menu. Use it to clear the audio before you record the sentence again: the sentence then counts as not recorded and plays nothing. It also acts at once, without confirmation. Deleting a take you recorded over pre-recorded audio clears the sentence as well; the pre-recorded audio does not come back.

Inspect levels & switch takes

The status-dot menu on a recorded sentence shows the take's audio level, notes when a take was recorded by someone else, and — when a sentence has more than one take — lets you switch between them so you keep the best.

Trimming a take

Adjust a take's start and end to remove silence or include more surrounding audio. Both trimming controls edit the same take.

Inline waveform

From a recorded sentence's status-dot menu, choose Show waveform to open a per-sentence waveform right under the sentence, with word boundaries, two draggable trim handles, and a play button. Drag a handle, then Save trim to save.

Overview scissors

Select a recorded sentence and click the scissors in the overview header. The overview zooms to that take and shows two handles plus a sub-bar with the duration change, Save, and Reset (which puts the handles back where they started). Playback stops while you trim and the play button is disabled; Esc, the scissors again, changing section, or starting a recording all leave trim mode.

Microphone setup

Open Microphone with the microphone icon in the rail at the right edge of the studio. There you can pick your microphone device, set the input gain, watch a live level preview, set the key press delay, and see where your voice is recognised. The tab opens during a recording too, so you can change the gain without stopping. In focus mode, which hides the rail, the Microphone settings button of a level warning opens the same settings in a dialog. Auto-calibrate listens while you speak normally for about three seconds and then sets the gain for you; if it hears no voice it says so and leaves the gain alone.

Speech recognition

Open Speech recognition with the head-and-circuit icon in the rail at the right edge of the studio. It lists four speech models, each with its download size, or with CPU or GPU once the app has chosen where the model runs on this computer. The speed estimate under the lists names the model and the CPU or GPU it is for. Nothing is downloaded until you press a model's download button, and its trash button removes it from this computer again.

Recognize words has four levels. Disabled is where you start: nothing listens while you read, and each take is saved without word timings. Press → each time you finish a sentence and start the next. In a continuous recording the studio divides the recording at those presses and saves each sentence as its own take. If you never press →, the whole recording is saved as the take of the sentence you started on. The other three levels use the model chosen under Model for speech feedback to recognise your words: In realtime as you read them, After pauses when you pause after a sentence, and After recording ends once you press Stop. They stay unavailable until a model for speech feedback is downloaded. A level, switch or model you cannot choose says why in a line under it. The one exception is a level that needs the realtime speech model when that model is missing: the reason is given once, in the line under Manual feed, and a screen reader reads that line for the level too.

While you record, the level, Manual feed and Estimate position feedback cannot be changed, and a line says why: a recording keeps the settings it started with. They can be changed again after you press Stop. The same goes for the model for speech feedback. A recording also keeps the choice of where it is decoded and whether final sync runs by itself after Stop as it was when you pressed Stop, so turning those switches afterwards affects the next recording.

The Model for speech feedback list sits in the same panel, above Model for final sync. A model appears there once you have downloaded it in Model for final sync and your computer can run it. The list only says whether a model is too slow to follow your reading. If a model fails to download, start or finish its test, its row under Model for final sync says so. The chosen model decides how quickly and how accurately your words are marked while you read. Changing it does not change the model used for final sync.

Pianissimo is a speech model trained on Swedish. It can give speech feedback while you read and can also be chosen for final sync. It is a large download, and it appears in the list only when your organisation has enabled it. It makes fewer mistakes than the medium KB-Whisper model, but KB-Whisper reads dialects better, so neither replaces the other.

Manual feed is on unless you turn it off. While it is on, the highlighted sentence stays until you press →, at every level. Turn it off to let the highlight move on to the next sentence by itself, as your words are recognised. That needs a level other than Disabled, so turning it off there chooses the fastest level your computer can follow, and → then leaves a continuous recording where it is. Deleting the smallest model turns manual feed on again.

Estimate position feedback is off unless you turn it on. On, the studio estimates which word you are reading from your reading pace and marks it as it goes, at every level; with manual feed on it waits on the sentence's last word until you press →. Off, nothing is guessed: the highlight moves only when you press → or when your words are recognised. In realtime always shows the position, so the switch stays on there.

The radio button beside each downloaded model chooses the model for final sync. The overview header then offers Run final sync on N sentences for the sentences that model would read better; press it to run final sync on them. The work runs in the background, waits while you record, and carries on after you close the browser. Turn on Run final sync automatically after Stop if you want it to start by itself after each recording. When you trim a sentence that has misread or missing words, it is offered for final sync again, so the words at its new edges are heard.

Offline & syncing

SpokenWords saves recordings locally when you are offline and uploads them in the background when you reconnect. The cloud in the page header shows whether everything is uploaded and, when you point at it, how many recordings wait and why. While the microphone is armed, uploads and the final sync wait so that the computer can give the recording its full attention; the cloud says so, and the sentences do not. A sentence only shows that work is running while its words are still being recognized: its background is tinted yellow-grey with a shimmer across it, and its dot pulses, until the words are in. A recording the server refused, or one whose upload failed, does not go up by itself: the cloud then shows no arrow and says 1 recording cannot be uploaded yet, and the statusbar and the notifications bell show the message with Try again. A failed upload is also tried again the next time you open the studio. A recording that arrived after its section had moved on is kept on the server for the producer to decide on, so it counts as uploaded and the cloud can still say Everything is uploaded.

Handing a section to the next role

You finish one section at a time. When the section you have open is recorded, click Mark article complete in the header: that section moves on to the reviewer, who is told it is waiting. Your other sections are untouched, so you can hand on chapter 3 and carry on recording chapter 4.

The handover button appears when your narrator assignment covers the open section and it is at Recording.

You cannot undo this action yourself. It changes the section's stage. If you hand over a section too early, ask the producer to move it back using ⋯ ▸ Article ▸ Change stage.

Keyboard shortcuts

Click the keyboard icon in the rail at the right edge of the studio to open Keyboard shortcuts. It lists the keys available in your current role, grouped by what the studio is doing: Any time, While playing, With the microphone armed, While recording and Article waveform. A key listed under Any time is not listed again where it does the same. ⋯ ▸ Keyboard shortcuts opens the same list. In focus mode, leave the mode with its button first. The same list is shown below, with the keys of the key map you use in this browser.

To change the keys, press Customize in Keyboard shortcuts, select a key and press the new one: × turns a key off and Add another shortcut adds a key. When you have changed something the button turns into Save and Cancel. Standard itself never changes: saving gives your keys a name and keeps them in this browser as a key map of your own, which you can rename, copy or delete from the ⋯ beside the key-map list, and you can switch back to Standard at any time. Keys the browser uses for itself, and Escape, cannot be chosen. Keys that work only while a list or the text has focus are shown with a lock and cannot be changed.

A key listed under Any time does the same in every state, except where a state lists it with another meaning, as While recording does for Space and ↑ / ↓. Every key that stops a recording keeps what you read. While the waveform has focus, its ← and → move between sentences as listed under Article waveform, and Space there does what it does anywhere else in the studio. On a Mac, the key labelled delete is Backspace, and Fn+delete is Delete. Play the selected sentence has no Standard key; you can give it one in a key map of your own.

Focus mode, the notifications bell, locked manuscripts, and other controls shared by every role are described under Common to every role on the overview.