Multi-Voice Generator

Write a script and label each line with a speaker name, like "Narrator:", "Alice:", or "Bob:". Assign a distinct voice to each speaker, then generate. We render every speaker's lines in their own voice and stitch them into one seamless audio file — perfect for audiobooks with character voices, dialogue scenes, or scripted podcasts.

Put the speaker name followed by a colon at the start of each line, e.g. "Narrator: It was a dark and stormy night." Lines without a speaker label continue the previous speaker. You can also use [S1]/[S2] tags. Click "Detect speakers" and we'll pull out every speaker so you can map each to a voice.

Yes — that's the whole point. Every speaker can use any voice in our catalog, including different models and your own cloned voices. Give your narrator a calm voice, your hero a bold one, and your villain something sinister.

Stitch mode (the default) renders each speaker separately with their own voice, then joins the clips — this is how you get genuinely different voices per character. Native dialog mode uses a model like Dia or MOSS-TTSD to generate the whole conversation in one pass for very natural turn-taking, but it uses a single model voice rather than per-speaker voices.

The Podcast Generator writes the script for you with AI from a topic, limited to two hosts. Multi-voice generation takes a script you provide — any number of speakers — and gives you full control over who says what and in which voice. Use it for audiobooks, plays, dialogue, or podcasts where you want to write the words yourself.

Multi-voice generation is billed by characters, exactly like normal text-to-speech — the total is the sum of every speaker's lines. Your daily free character bucket is used first, then your account balance. Premium-model voices cost more per character than free models.

Yes. Clone your voice with our Voice Cloning tool, then pick it as the voice for any speaker. You can voice the narrator yourself while AI voices the other characters.

Yes. Everything you generate is yours to publish, sell, and distribute. No royalties, attribution, or licensing fees.

Scripts are limited to 200 speaker turns per render, with the same per-turn character limit as normal TTS. For full-length books, use the Audiobook workspace, which renders chapter by chapter and supports multi-voice chapters.