Choosing a Korean AI Podcast Host Pair: A Test Script

What should the test dialogue contain?

Use a short exchange with a figure, an explanation and a follow-up. This invented update provides all three:

Host 1: 이번 달 교육 신청자는 백이십 명입니다. 지난달보다 스무 명 늘었습니다.

Host 2: 그럼 지난달에는 백 명이 신청한 거네요. 이번 달 교육 일정도 같나요?

The figures are consistent: 120 now, 20 more, therefore 100 previously. The last question asks about the schedule instead of inventing an answer. It gives you a useful place to hear whether the second speaker sounds like a question rather than another announcement.

How do you compare candidate pairs?

Give the first host the explanation and the second the clarification. Keep text and settings fixed when swapping candidates. Listen without the script, then with it. Use this worksheet; the blank results are yours to fill, not measured scores from us.

How do names and formality change the choice?

Use the level of politeness your audience expects and keep it consistent unless a deliberate role difference requires otherwise. Add a real name only after confirming its pronunciation. Do not use a global voice count as a Korean quality ranking. DIALØGUE uses Fish Audio; catalog previews can help shortlist voices but cannot establish how this dialogue will sound.

What would justify choosing neither pair?

If both blur a number or make the hosts difficult to distinguish, fix the script and try again before choosing a winner. A calm explainer and a livelier questioner are roles to test, not universal prescriptions. This guide provides an editorial worksheet, not a native-listener study or an engine benchmark.

See the general voice guide, create a Korean podcast, or test your own script.

The worked examples are synthetic editorial examples, not a vendor benchmark or measured pronunciation result. Verify the generated audio before sharing.