Playground

Voice settings

Every control in the Text to Speech settings panel, with its label, range, default and what it changes.

This page is the lookup table for the Options panel on the right of the Text to Speech page, and for the Tags row and the Generate button under the writing area. The panel holds the voice, the three controls Speed, Guidance and Quality, and Reset to defaults. Every value here is what the panel accepts; a generation always uses the values shown at the moment you click Generate.

Selected voice

Control What it shows or does
Voice The voice's name and its dialect.
Change voice Opens the Choose a voice dialog with the Explore and My Saved Voices tabs, search, the Type menu, the dialect chips and Recently used. Click a voice, then Select.
Choose a voice Shown in place of the name when no voice is selected.

Under the writing area, the Dialect chips, the round voice buttons and All voices pick a voice the same way. A voice fixes the language and the dialect. There is no separate language or dialect control; pick a different voice to change either.

Options

Label Control Range Default What it changes
Speed Slider 0.5 to 2.0 in steps of 0.05 1.0 The reading pace, as a multiple of the voice's normal pace. 1.0 is normal; lower is slower, higher is faster.
Guidance Slider 1.0 to 5.0 in steps of 0.1 2.0 How closely the voice follows the text. Keep the default unless you are comparing outputs in your own listening tests.
Quality Five buttons 16, 32, 48, 64 or 128 48 How many synthesis steps are run. More steps are slower and cleaner. Only these five values exist.

Speed shows its value to two decimal places with a multiplication sign, for example 1.00×. Guidance shows one decimal place, for example 2.0. The value you see is the value sent.

Reset to defaults, at the top of the panel, puts Speed, Guidance and Quality back to their defaults and leaves the selected voice as it is.

Tags

The Tags row, under the writing area, holds five buttons. Each one drops a cue in square brackets into the text, for the voice to act out at that point. The cue appears as a small removable chip.

Button Text added
[laugh] [laugh]
[cough] [cough]
[clear throat] [clear throat]
[sigh] [sigh]
[chuckle] [chuckle]

A cue lands at the cursor, or at the end of the text when the cursor is not in the writing area. Click the chip's minus to remove it, or type the same cue by hand anywhere in the text. Cues count toward the 5,000-character limit and are charged like any other characters.

Generate button

Part What it shows
Generate Starts a generation.
Generating... Replaces the label, with a spinner, while a generation runs.
about N credits An approximate estimate before generation and the returned cost afterward. It is not your balance.

The button is unavailable while the writing area holds only whitespace, while the voice list is loading, and while a generation runs.

Call it well

Change one setting at a time and listen to a short sentence after each change, because every generation is charged.

Recommended

  • Generate one sentence at the defaults, then move Speed only, then listen again. Touch Guidance and Quality last, one at a time.

Less effective

  • Move all three controls before the first generation and try to hear which one did what.

Worked example

You want a slower, steadier read of a news paragraph.

  1. Click the Fusha chip and click a voice.
  2. Drag Speed from 1.00× to 0.50×. Leave Guidance at 2.0 and Quality at 48.
  3. Click Generate and listen.
  4. To compare, click Reset to defaults: Speed returns to 1.00×, the voice stays, and a second click on Generate gives you the normal pace for the same charge.

FAQ

Do the settings persist between visits?

No. Each visit starts with Guidance 2.0, Speed 1.0 and Quality 48, and the default voice.

Can I type a value the slider does not offer, such as speed 1.3?

Not in the portal; the speed slider moves in steps of 0.05. Over the API, speed and cfg accept any number inside their ranges. See Text to Speech voices and options.

Do the settings change the price?

No. The charge depends on the characters in the text only. A cue from Tags adds characters, so it adds to the charge.

Where do I choose the language?

Through the voice. Pick an English voice for English and an Arabic voice for Arabic; the dialect comes with the voice as well.

Which setting should I try first?

Speed. Keep Guidance and Quality at their defaults until you are comparing outputs on purpose.