Regular text to speech reads everything in one voice. A podcast, interview or sketch needs more: two or three people who answer each other, interrupt and laugh. Stitching separate lines together by hand takes ages, and the pauses still sound off. The Dialogue tool voices the whole script at once, as a single recording.
Write each line on its own row: "Anna: Hi! What are we talking about today?". The name before the colon is the speaker, and you assign each speaker a voice. Lines that start with a dash alternate between two voices. No script yet? Describe a topic — "two friends argue whether to move to another city" — pick a length from 1 to 5 minutes, and the AI writes a two-person dialogue for free that you can edit.
Gemini 3.1 Flash TTS by Google voices two speakers affordably and with natural pauses, and you can set the mood of the conversation in words. Eleven v3 Dialogue by ElevenLabs takes up to three voices with the most expressive delivery: tags like [laughs], [whispers], [interrupting] right inside the lines. You pay only for the spoken text — speaker names aren't counted.