Help

What this is

A place to turn text into narration. A project holds numbered passages; each passage is read aloud by a voice you choose, and the sound is kept so you can play it and download it as a .wav file.

Making audio

Press Make audio on a passage, or Make all missing audio on the project. The work is queued and done by a separate process, so the page shows queued and then making and refreshes itself until it is done. Change a passage's text, voice or speed and its audio becomes out of date: the old sound is kept and can still be played until you make the new one.

Voices and speed

A project has a voice and a speed, and a passage may use its own. Speed 1.00 is normal; the range is 0.50 to 2.00.

Pronunciation and expression

The speech engine takes plain text only. It has no phonetic input and no markup for emphasis, so these are what you can do:

• Respell a word on the project's Pronunciation page: every whole word that matches is replaced by the spelling you give (for instance Nginx by engine ex) before the passage is spoken. Your passage keeps its own text. Only the passages that contain the word go out of date.
• Punctuation sets the pauses and the pitch: a comma is a short pause, a full stop a longer one, a question mark lifts the end, and a dash or an ellipsis make a break. Several short passages read better than one long one.
• Speed for a passage that should be slower or quicker than the rest.

Limits

• 50 projects to an account, 500 passages to a project.
• 3000 characters to a passage.
• 500 respellings to a project.
• 100 passages waiting to be made at once.

Voices: Kokoro-82M (Apache-2.0), run on the server's processor through sherpa-onnx.