How to Choose AI Voices for a Two-Host Podcast
DIALØGUE uses Fish Audio for podcast speech. Choose two hosts you can distinguish, test the language and names in your actual script, and check the result before sharing.
What should you listen for?
Start with a short passage containing a name, a number, a question, and a reply. Use the same passage for each candidate so differences in writing do not decide the comparison. Clear words and comfortable pacing matter more than a dramatic first impression.
How do you pair two hosts?
Give one host the explanation and the other a question that tests it. Listen without looking at the script: can you tell who is speaking? Contrast can come from pitch, rhythm, or delivery; gender alone does not establish useful contrast. A catalog preview helps narrow the choice, but it cannot guarantee your episode will sound the same.
Which engine does DIALØGUE use now?
Speech generation and voice previews use Fish Audio. Older posts describe earlier ElevenLabs and Gemini workflows; those are historical references. The available catalog is curated by language. A larger global catalog does not mean every voice is a suitable choice for every language.
What should you review before spending credits?
Check proper names, abbreviations, dates, and changes of speaker in the script. Keep delivery cues such as [laughing] when they are intentional. After generation, compare the audio against the approved text and ask a fluent listener about names you cannot judge yourself.
Review the creation workflow or choose voices for your episode.
Written by
Chandler NguyenAd exec turned AI builder. Full-stack engineer behind DIALØGUE and other production AI platforms. 18 years in tech, 4 books, still learning.
Related Articles
Ready to create your own podcast?
Turn any topic or document into a professional podcast — with outline and script review before audio.
Create a Podcast

