Short fiction
Flash fiction and short stories come alive when each character has a voice. Paste the story and it is cast in seconds.
Multi-speaker TTS
Multi-Voice Text to Speech That Knows Who Is Talking
Paste a scene, a short story or a dialogue, and this multi-voice text to speech tool works out who says each line, gives every speaker a different AI voice and reads the narration in another. Fix any line on its own, then download the audio as MP3 or a chaptered M4B.
AI audiobook samples
View allTurn pasted text into multi-speaker audio
- 1
Paste the text
Drop in a scene or a whole story. Lines such as “Chapter 2” or Markdown headings start new chapters; otherwise it stays as one.
- 2
Check who says what
Quoted lines go to speakers from their tags, script lines such as “MAYA: Not here.” to the name in front, and the rest to the narrator. Reassign any line from its speaker menu.
- 3
Pick the voices
Keep the suggested voice for each speaker or swap it. Preview voices side by side so the main pair sound clearly different.
- 4
Generate and download
Generate the lines, play them back in order and download one MP3 for the chapter, a ZIP or an M4B with chapter markers.
How to format text for multi speaker text to speech
The casting reads dialogue the way novels and scripts are written. A minute of tidying before you paste saves most manual fixes.
Do this
- Put every spoken line inside quotation marks, straight or curly
- Tag a speaker the first time they talk: “Maya said” or “said the captain”
- Keep one speaker per paragraph, and start a new paragraph when the speaker changes
- Use the same name for a character throughout, rather than switching between a nickname and a full name
- Prefer common speech verbs such as “said”, “asked”, “replied” or “whispered”, which are recognized more reliably than rare ones
- Or write it as a script, one speech per line: “MAYA: Not here.” works on its own, and “Maya: Not here.” once a paragraph holds two such lines
What trips it up
- Single letter labels such as “Q:” and “A:”, which stay with the narrator; use the speakers’ names instead
- Long untagged exchanges between three or more people
- Dialogue set in italics or dashes instead of quotation marks
What each voice model does with dialogue
All four models give each speaker a separate voice. They differ in how far you can direct a line and what it costs.
| Model | Directions like [whisper] | Inline tags like [laugh] | Credits per 1,000 characters |
|---|---|---|---|
| Standard | Yes | Yes | 9 |
| Lite | Yes | Yes | 6 |
| Premium (Turbo 2.5) | No, brackets are removed | No | 13 |
| Expressive (v3) | Yes, the widest range | Yes | 34 |
The full tag list and examples are on the AI voice emotion tags page.
What people make with multiple voices text to speech
Language practice
Two-person dialogues for learners, voiced by native-sounding speakers. See language learning audiobooks for graded readers.
Explainer dialogues
A question-and-answer script between two hosts is easier to follow than one voice reading both sides.
Proof listening a scene
Hear whether a conversation flows before you send it to an editor. Writers use text to speech for writers the same way.
Bedtime stories
A gentle narrator plus playful voices for the fox, the owl and the child in the story.
Game and quest text
NPC dialogue and quest scenes read by distinct voices, for a prototype or a tabletop recap.
Listen to text to speech with different voices
The player includes a two-voice conversation and a full-cast story, each generated in this studio with a separate voice for every speaker.
Full-cast AI audiobook
Alice's Adventures in Wonderland
Transcript
0:000:41
Playlist
[say with hyped, high energy like a YouTube intro]
Stop scrolling. In the next sixty seconds, I'm going to show you the trick that doubled my channel.
Direct each speaker, not just the scene
Every line is its own clip, so each one can carry its own direction. Add [nervous] to one speaker and [bored] to the other and the same exchange plays completely differently. Inline tags such as [laugh] or [sigh] land exactly where you type them on Standard, Lite and Expressive.
When a line still sounds wrong, change it and regenerate only that line; the rest of the conversation keeps its audio. A project-wide delivery setting, from stable to creative, controls how much the voices vary from take to take.
Text to speech with multiple voices versus one voice
Why one voice flattens dialogue. With a single voice, listeners have to track every “she said” to know who is talking, and a fast argument turns into a blur. Giving each speaker a voice moves that work from the listener’s head into the audio.
What it takes by hand. With single-voice TTS you would split the text by speaker, generate each piece in a different voice and stitch the clips together in an audio editor. In the AI audiobook creator the split, the casting and the joining happen per line, automatically, and the result plays in order.
What it costs. Credits are charged per 1,000 characters, whichever voice reads them. A 1,500 word scene is about 8,500 characters, or roughly 77 credits with Standard voices, under $2 at plan prices. Sign up with Google and you get 100 gift credits (valid for 7 days, about 12 minutes on Standard; usage varies by text and voice tier) to try it.
When to upload instead. For a full book, upload the file rather than pasting: EPUB, Word, PDF, TXT and Markdown keep their chapters, and the multi-voice audiobook page explains casting across a whole novel. For plain text without dialogue, text to audiobook is the simpler route.
Languages. The voice library covers 13 languages, from Spanish and German to Japanese and Arabic, with English the most polished. Paste text in the language you want to hear, and preview a voice on a few lines before you generate a long piece.
Privacy. Only the lines you generate are sent to the speech model, and nothing you paste is used to train AI models.
Multi voice text to speech FAQ
Can text to speech use multiple voices in one file?
Yes. Each line is read by the voice assigned to its speaker, and the lines are joined in order into one track per chapter. Paste a single scene and you download it as one MP3 with every voice in it.
Is there a limit on speakers in one text?
Up to ten named characters get their own voices, plus the narrator and an “Other voices” role shared by minor speakers. For most scenes and short stories, that covers every character who talks.
Can I paste a screenplay with speaker names?
Yes. Script lines such as “MAYA: Not here.” or “HAMLET. To be, or not to be” are cast by the name in front, and a parenthetical like “MAYA (quietly):” becomes a direction. Screenplay cues on their own line, with the speech underneath, work too. Mixed case names such as “Maya:” need two or more such lines in a paragraph, as in a chat transcript.
Is multi voice text to speech free to try?
Yes. Pasting text and casting the speakers cost nothing, and signing up with Google gets you free credits to generate a short sample with any voices, no card needed. Longer projects use credits from a plan.
Why are two of my characters read by the same voice?
Either one of them was not tagged, so their lines went to “Other voices” or the other speaker, or both roles got similar voices. Reassign the lines from the speaker menu, or give one character a voice that differs in gender, age or pace.
Does it work for a two-host podcast script?
Yes, if it is written as quoted dialogue. Tag each host once at the start; after that, untagged lines in a two-person exchange alternate between them, so the narrator hardly speaks. The chapter exports as one MP3, and directions such as [casual, amused] help it sound unscripted.
Can I edit a line after the audio is made?
Yes. Click the line, change the words or the direction and regenerate just that line. Nothing else in the conversation is regenerated, so small fixes cost only a few credits.
Paste a scene for multi-voice text to speech
Hear every speaker in a different voice in under a minute, with free credits to start if you sign up with Google.