Creating audio and podcasts from text

Two features

Task menu ▸ Analyze

• Create Audio – reads the text aloud in one voice. An audiobook.
• Create Podcast – turns it into a conversation between two voices, host and guest.

If the entries are missing or named differently

These features need voices that your account provides. If none are available:

• Create Audio is not visible at all,
• and Create Podcast is called Create Podcast Script instead – you then get the written dialogue but no recording.

That is not a fault. If the add-in reports that no voices were received for your account, sign in again; if it persists, your administrator sets the voices up.

Create Audio

Select the text. With nothing selected the add-in asks whether the whole document should be voiced or whether you would rather pick a file.

Then you choose the voice; a sample sentence can be played first. The add-in tells you what the audio file will be saved as – and warns you if that will replace an existing file.

During generation, paragraphs are skipped that are empty or entirely inside curly, square or round brackets. If there is nothing at all to voice, the add-in says so with exactly that explanation.

At the end you can play the file straight away. If no MP3 encoder is available on the system, a WAV file is saved instead; you are told about that too.

Create Podcast

A settings window collects everything:

• Podcast language – determines which voices you can choose from.
• Host voice and Guest voice. Note the hint in the window: the voices are not filtered by language. Pick voices that can actually speak the language you want – otherwise the result comes out with an accent nobody wanted.
• Audio file – location and format, MP3 or WAV.
• Show and edit script before voicing – tick this. You then see the generated dialogue and can change it before it is voiced. Changing it afterwards means generating everything again.
• Under Show advanced: Target audience, Context, background info, Target length and Extra instructions.

If only one voice is available for your account, the add-in points out that host and guest will then sound the same and asks whether to continue.

During voicing

The text is voiced section by section; the display shows "Segment N of M". If a section fails, the add-in says which – the rest are still merged. The operation can be cancelled.

What to consider first

The podcast dialogue is written by the model. It does not merely summarise your text; it adds transitions, follow-up questions and framing of its own. Read the script before you have it voiced – something may have ended up in there that was not in your source text.