Skip to content

Writing for Audio

Writing for audio and synthetic voices creates new challenges for the writer to make it sound good. This page contains information and tips to make the script sound great.

Limitations

Some platforms have special limits on the audio length.

Alexa:

  • 4 minutes of audio for text
  • 90 seconds of audio for repeat text

The average synthetic voice has a speed of 160 words per minute. In three minutes, it can read 480 words. This is slightly over one standard pocket book page (around 350 words/page).

Tips & tricks

Here are some tips & tricks on how to get the most out of the voice in your audio book.

  • To put emphasis on a word, surround it with 'quotes' (this is currently not supported for Polly voices).
  • If you want to change the pronunciation of a word or passage, try writing it as you would say it rather than how it is actually spelled.
  • Add prepositions to help the listener navigate the scene, and to create a pleasing rhythm.

    After the flood, my basement smelled like mould.

  • Avoid having a lot of dependent clauses.

    I handed in my homework, and got a good grade, but my friend Charlie had not finished the assignment, and skipped school that day.

  • Use the voice as an editor. The voice will sometimes stumble when there is a problem with the structure of sentence. If the voice is not enunciating correctly, try simplifying the sentence and change the wording.
  • Avoid compound nouns with verbs in them. You can remove the noun and use it in a simple sentence instead. Instead of "Lightning strike" write a sentence like "Lightning flashed across the sky".
  • To add a 1-second pause, paste the following snippet into the text: <break time="1s"/>. You can change the duration of the break by replacing the 1 with another number, up to a maximum of 10 seconds. Here is an example with 0.2 seconds break:

    After the flood <break time="0.2s"/> my basement smelled like mould

Fabella Audiobook Builder

To get good results from the voice synthesis it is important that the script is of good quality. Here are some guidelines:

The script should be a simple text document without formatting and only have the text that is supposed to be read out. There should not be any page numbers or similar.

The text should be free from extraneous line-breaks. Only have line-breaks between paragraphs.

Don't

You try to move forward silently, but something crunches underneath your heel.
It causes you to lose your balance and tumble to the floor.
Turning your head, you find yourself face-to-face with a human skull.
A torn cloth is draped over the bones. You can see a mocking grin between the loose threads of the fabric.

Do

You try to move forward silently, but something crunches underneath your heel. It causes you to lose your balance and tumble to the floor. Turning your head, you find yourself face-to-face with a human skull. A torn cloth is draped over the bones. You can see a mocking grin between the loose threads of the fabric.