Where an AI voice earns its place
Podcasts are built from two kinds of audio. There is the conversation, which only you and your guests can make. And there is the packaging around it, which is scripted, repetitive and time-consuming to record well. An AI voice is a good fit for the second kind.
- Intros and outros that stay identical while one line changes each week.
- Sponsor reads you can produce the moment the copy is approved.
- Recaps and “coming up” teasers written after the edit, without setting the microphone up again.
- Narrated shows, where the whole episode is a script.
- Corrections, generated and dropped in without re-recording.
Writing for the ear
Listeners cannot glance back at a sentence, so podcast copy has to land first time.
- Put the subject first. “Our guest this week is a beekeeper” works better than a long lead-in.
- Repeat names. In audio, “she” loses its owner quickly.
- Round your numbers. “Almost a third” is easier to hear than “31.6 per cent”.
- Read it aloud before you generate. If you run out of breath, the sentence is too long.
Timing guide
| Segment | Typical length | Words |
|---|---|---|
| Cold-open teaser | 10 seconds | 25 |
| Show intro | 15 to 20 seconds | 40 to 50 |
| Sponsor read | 30 seconds | 75 |
| Longer sponsor read | 60 seconds | 150 |
| Outro and credits | 20 to 30 seconds | 50 to 75 |
Settings
Use Natural for intros and sponsor reads, which should sound like a person talking. Use Narration for long scripted episodes, where an even tone matters over ten or twenty minutes. If a sponsor read sounds too flat, try Expressive or lower Stability a little.
Whatever you choose, keep it the same every episode. The announcer’s voice is part of the show’s identity, like the theme music.
