In short: Write the script with lines that start with A: and B:, generate it in the two-speaker dialogue mode so both hosts are in one MP3, create an intro with the podcast intro generator and music with the Music Studio, then join them in any audio editor and upload to your podcast host.
What kind of podcast works best
AI voices are great for formats where the script matters more than spontaneous talk: daily news summaries, explainers, book and film reviews, language lessons, company updates and short stories. Two hosts make these formats livelier, because one can ask the questions listeners would ask.
Be open with your audience: say in the show notes that the voices are AI-generated. Listeners appreciate honesty, and it is required on some platforms.
Step by step
- Plan a short episodeStart with 5 to 10 minutes: a hook, three points, and a closing line. Shorter episodes are easier to finish and to listen to.
- Write the script as a dialogueStart every line with
A:orB:(or two names). Let host A explain and host B react, ask and summarise. See how to write a dialogue script. - Generate both hosts in one fileUse the two-speaker dialogue mode in Text to Speech, pick a different voice for each speaker, and generate one MP3 with the whole conversation.
- Create an introWrite a 10-second intro with the podcast intro generator and make a short theme with the intro music generator.
- Join and level the audioIn Audacity, GarageBand or any editor, place intro, music bed and dialogue, keep music about 20 dB quieter than the voices, and export MP3.
- PublishUpload to your podcast host (for example Spotify for Creators or Buzzsprout), add show notes, and submit the feed to the podcast apps.
A simple episode template
| Part | Length | What happens |
|---|---|---|
| Cold open | 10 to 15 s | Host B asks the episode's big question |
| Intro | 10 s | Theme music and show name |
| Point 1 | 1 to 2 min | Host A explains, B asks a follow-up |
| Point 2 | 1 to 2 min | An example or story |
| Point 3 | 1 to 2 min | What it means for the listener |
| Wrap-up | 30 s | B summarises in one sentence, A gives the call to action |
Tips for a natural conversation
- Choose two clearly different voices, for example a female and a male voice, or two accents, so listeners never confuse the hosts.
- Keep turns short. Real conversations switch every few sentences.
- Add reactions, like "Wait, really?" or "That makes sense.", to break up long explanations.
- Use pauses such as
[break 1s]between topics. See how to add pauses. - Keep the same voices every episode so the show has a recognisable sound.
Frequently asked questions
Can I monetize a podcast made with AI voices?
Generated audio can be used commercially; see the terms. Each podcast platform has its own rules, so check theirs and disclose that the voices are AI.
Can the two hosts speak different languages?
Yes. Pick a voice in each language; this works well for language-learning shows.
How long can one dialogue be?
Long episodes can be split into several dialogue parts and joined in an editor.
Do I need to sign up?
You use your Speechma Pro account.
The Speechma newsletter
Make better voiceovers, faster
- New voices and features the day they launch
- Script templates and tricks that make AI voices sound human
- Creator stories, community picks and early access