Voices

The Voices section is where a project gets its sound: voice-over for every line a character speaks, and music and sound effects for the scenes around them. It has no Ctrl number; open it from the activity bar.

Voice-over has no presence on a board or a node — no row, no chip, no icon. Everything happens in this section. The one per-field setting is Spoken aloud in the node builder, which decides which text gets a voice-over take (see Translatable and spoken aloud).

The list and its four pages

The section’s list, titled Voices, is the voice library. Its title row carries four buttons — Cast · Lines · Music · Sound FX, named when you rest the pointer on them — and each opens that page as a tab beside your boards. The button whose tab is in front is lit.

Where	What it holds	Layout
Voices, the list	The voice library — one entry per voice, in folders. Opening one shows its design page.	Each voice opens as a tab
Cast	Every entity with the voice it speaks with, and the project’s pronunciation dictionary	A tab
Lines	One board at a time: every voiced line, its audio and its takes, and the bulk actions	A tab, with a board picker on its left and the lines on its right
Music	Describe music and generate several takes at once	A tab
Sound FX	The same, for sound effects	A tab

Closing one of these tabs does not stop it: a batch of lines or takes still being generated carries on, and reopening the tab shows it where you left it.

The three voice-over places are one chain. A voice is designed in Voices, given to a character in Cast, and speaks that character’s lines in Lines. A line is always spoken by its speaker’s voice — there is no per-line voice picker. The how-to pages follow the same order: designing a voice, casting, producing the lines, working with a studio, and music and sound effects.

What counts as a line

A line is one field of one node, in one language: any Text or Rich text field marked Spoken aloud, which is on by default for both types. Turn it off on a field nobody should hear.

The line’s speaker is the node’s first Entity field. A line on a node with no Entity field has no speaker, and cannot be voiced until it has one.

A board appears on the Lines page once it holds at least one voiced line.

A voice

Part	What it is
Name	The display name. The voice’s code is made from it when the voice is created.
Colour	Assigned when the voice is created; the dot beside it in the drawer, on its page and in Cast
Who speaks	The engine that speaks this voice
How they sound	The description a voice is designed from — one for the voice itself, and one for each language cast separately
How they perform	Steady, Natural or Volatile, where the engine offers it
Reference take	The audition you kept. It is what the play button in Cast sounds.
Details	The engine’s own controls, and a read-only record of how the voice was made

A voice that has never been auditioned or given a ready-made voice still does a job: its page reads casting only — no audio is generated for this voice, and it is the casting a studio’s recording script carries. That is a complete voice, not a broken one.

A take

Every line holds a history of takes, per language, and one of them plays.

State	What it means	Made by
AI draft	Audio generated in TaleShard	Generate missing, Regenerate, a line’s own regenerate button, a re-cast
delivered	A file you supplied	Import recordings, Match from library, or picking a clip on the line
approved	Signed off with Approve	The take’s own button
was approved	A signed-off take that another take has replaced as the one that plays	Use or Approve on another take, a re-cast with Re-record the approved takes too ticked, or a ticked import

A line has at most one approved take; a replaced sign-off is kept, marked, rather than deleted.

Two markers sit beside the state: plays names the take the player sounds, and text changed appears when the line’s words were edited after its take was made.

An imported or attached file is never marked as AI audio, and that cannot be changed later.

What is protected

Approved and delivered audio is never touched by Generate or by a re-cast. A replaced take is not deleted either — it stays in the line’s take history.

Action	AI drafts	Approved takes	Delivered recordings
Generate missing	Lines with audio are skipped	Skipped	Skipped
Regenerate selected, or one line’s regenerate button	Replaced; the old take stays in the history	Left alone	Left alone
Re-cast	Replaced in the new voice	Left alone, unless you tick Re-record the approved takes too	Never touched, in any mode
Match from library	Lines that already have audio are skipped	Skipped	Skipped
Import recordings	Skipped unless you tick them	Skipped unless you tick them	Skipped unless you tick them
Clear board	Removed	Removed	Removed

Clear board is the one action that reaches signed-off audio. It takes the audio off every line of the board in the language you are looking at. It is an ordinary edit, so Ctrl+Z brings it back, and the files themselves stay in Media.

Removing a single take from a line’s history removes the take, not the audio file.

Languages

A take belongs to one language. The language control in the project header decides which language the Lines page works in, which language a voice’s page edits, and which language an import fills.

There is no fallback to another language. A line with no take in the active language plays nothing, and a line not yet translated into it shows not translated rather than borrowing the original’s audio. Translation itself is Localization.

Where the audio goes

Everything this section makes or imports is an ordinary file in Media.

Audio	Media folder
Generated and re-recorded lines	Voiceovers / «board name»
Imported recordings	Voiceovers / «board name»
Music takes	Audio / Music
Sound FX takes	Audio / Sound FX

The folders are created the first time something lands in them, and reused afterwards.

Credits

Generating audio spends credits, the same balance as every other AI action — see how credits work.

Spends credits	Free
Audition, Re-design… and a Remix — three voices for one charge	Keep this one — keeping an audition
Speak this and Hear it — charged for the audio; nothing is saved	Use this voice — keeping a ready-made voice
Start from this voice — one short clip, which is kept	Casting, and the pronunciation dictionary
Generate missing, Regenerate, a re-cast and its preview	Approve, Use, Reopen
Suggest direction	Match from library and Import recordings
Music and Sound FX takes, and Improve	The recording script and the AI audio report

Every button that spends shows the figure before you press it — Generate missing · ~N credits. A ~ in front means it is an estimate rather than an exact figure. Speech is charged by the length of the text, so a long line costs more than a short one.

A line or take that fails is not charged. A run you stop is charged only for the lines it had already finished. After a run, a note under the toolbar says what it spent, and each take in a line’s history shows its own cost (12 cr.).

Plans

Voice-over, music and sound effects are on every plan; credits are the meter.

The recording script export needs the Pro plan or above. Below that, its button stays visible with the plan’s badge, and pressing it explains the gate. The AI audio report beside it is free on every plan.

Hearing it in play mode

The player’s settings carry two voice-over controls, beside Text speed and Accessibility (see Player settings):

Control	What it does
Play voice-over	On by default. Remembered per project, in this browser — the only player setting that outlives the visit.
Volume	0–100%, in steps of 5. Shown only while Play voice-over is on, and remembered on its own, so turning voice-over off and on again keeps your level.

Both are offered to a visitor on a share link as well as to the author.

Each line plays its current take in the active language. Audio never holds up the story: if a clip cannot play — it is missing, or the browser refuses to start it — the line still goes on.

Limits

Limit	Value	Enforced by
Takes per music or sound batch	Each engine’s own maximum — the Takes control offers no more	The server refuses a larger batch
A designing passage	Each engine’s own minimum and maximum, shown as a character count under the passage	The Audition button, and the server
A music or sound description	Each engine’s own minimum and maximum	The server
An imported recording	Your plan’s per-file upload limit, and the project’s storage	The upload — see Media
A re-cast preview	5 lines	The dialog

What breaks

A line that cannot be voiced says why on its own row in Lines, and is skipped — never sent, never charged.

On the row	Why	The fix
not translated	The line has no text in the active language	Translate it, or switch language
nothing written	The field is empty	Write the line
speaker not cast	The speaker has no voice — or the node has no Entity field, so there is no speaker	Cast them in Cast, or give the node its speaker
voice missing	The speaker is cast to a voice that no longer exists	Cast them again
voice not minted	The voice — or this language’s separate cast — has no audio of its own yet	Audition and keep one, or pick a ready-made voice
failed — retry	The last generation of this line failed. Nothing was charged.	Generate it again
text changed	The words were edited after this take was made	Regenerate it, or have it re-recorded

Situation	What you see
A character is cast to a deleted voice	voice “«code»” is gone on their Cast row. The record is kept, not cleared, so you can see who they were.
Out of credits	Insufficient AI credits. Please purchase more credits or wait for your monthly reset.
The engine cannot speak right now	«Engine» cannot speak a line right now. Pick another engine for this voice.
A line too long for the engine	That line is too long to speak.
A music or sound description out of bounds	«Engine» needs a description of at least N characters, or takes a description of at most N characters
A preview the server could not render	The preview could not be rendered. — nothing was written