Text to speech with multiple speakers
Write the conversation as a script and start each line with the speaker's name in capitals and a colon. Turn on Multi-voice in the editor, give every name its own voice, and the whole exchange comes back as one MP3 with an SRT caption file, a short pause at each change of speaker, and the voices level-matched to each other. Standard voices work on every plan, including the free plan's 25,000 characters a month, and the names you cast are never read aloud or charged.
Try the live product
Type, choose a voice, then press play.
Type a script, pick a voice, and hear it read back.
Free to try — no sign-up.
118 charactersA complete two-person script looks like this: “MAYA: Did the first test batch go out this morning?” on one line, “LEO: It did, and the replies are already coming in.” on the next, and “MAYA: Then we read them together after lunch.” on the third. Cast MAYA with one voice and LEO with another, and those three lines come back as a single track. A paragraph with no name in front of it is read by the main voice, which suits a narrator who sets the scene between turns.
Interviews, two-host explainers, language-lesson role plays, training scenarios and radio-style sketches all follow the same pattern. On a paid plan the cast can also take premium voices at 4× credits, and one conversation can mix the two, such as a premium host with standard guests: each line is billed at its own speaker's rate, and expression tags written into the script are matched to each voice.
No signup needed: paste a script and hear the first 500 characters.
How it works for multi-voice dialogue
- 1
Write the script with each speaker's name in capitals and a colon at the start of their line, or ask the editor's AI writer for a dialogue and it drafts one with the labels in place.
- 2
Turn on Multi-voice in the script editor and the Cast panel appears, listing the names the script uses. Give each speaker a voice, and play the sample of the voice you picked before you settle on it.
- 3
Generate once: every line is voiced by its own speaker, and the turns are joined into one MP3 with a short pause wherever the speaker changes.
- 4
Download the MP3 and its SRT captions. After an edit, generate again; with standard voices only the lines you changed are voiced and billed a second time.
Built for multi-voice dialogue
Names that pick voices, not words
Once a name is in the cast, the capitalised name before the colon chooses the voice and is left out of the audio, and its characters do not count against your allowance.
One track, one caption file
Every turn lands in a single MP3 with a short pause at each change of speaker, and the SRT follows the same timing, without the speaker names.
Voices levelled against each other
With loudness normalisation on, which is the default, each voice is level-matched to the rest of the cast before the whole track is normalised, so a softer voice is not lost beside a louder one.
Standard and premium in one cast
On paid plans a cast can mix both kinds of voice. Premium lines use 4× credits and standard lines count each character once, every line at its own speaker's rate.
Start from a recording
Transcription can label each speaker in a recording you own or have permission to use, and the labelled script opens in the editor ready to cast.
126 studio voices (100 premium, paid plans) · 20 output languages · SRT captions, bulk mode and audiobooks on every plan
Free to start
Use the free plan first, then move to a monthly plan when your production volume grows.
25,000 characters
Free plan monthly allowance
$7/month
About 5.5 hours on the Creator plan
Questions, answered
How do I format a script with several speakers?
Start a line with the speaker's name in capital letters followed by a colon, such as HOST: or DR. LEE:, then the words they say. A blank line between speakers is optional. Names use the Latin capitals A to Z, and digits can follow as in GUEST 2, even when the lines themselves are in Arabic, Hindi or another script. Headings such as NOTE: or WARNING: stay ordinary text unless you add that exact name to the cast, and a backslash in front of an ALL-CAPS prefix has it read aloud as written. Then switch Multi-voice on in the editor so the names become speakers: with it off, the whole script is read in one voice, names included. A name you have not cast is read aloud along with its line, so give every speaker a voice.
Is it limited to two speakers?
No. Two-person conversations are the common case, but a narrator plus several characters works the same way: each name in the cast gets one voice and keeps it for the whole script, and paragraphs without a name go to the main voice. When a scene has exactly two speakers and both use premium voices, their turns are voiced together in shared passes rather than one line at a time.
Can one conversation mix standard and premium voices?
Yes, on paid plans. Give each speaker whichever voice fits: each line is billed at its own voice's rate, premium at 4× credits, and expression tags are converted to suit each voice. The two kinds of voice can differ in tone, room sound and pace, so listen to the whole conversation before you publish. The free plan casts standard voices only.
How are characters counted in a dialogue?
Only the spoken lines count: the cast names before the colons are stripped out before voicing and are not charged. The free plan includes 25,000 characters a month, and one script can hold up to 100,000 characters however many speakers it has. If a voice-over fails, its characters are refunded automatically.
What files do I get back from a dialogue?
One MP3 of the whole conversation and an SRT caption file timed to it; the captions carry the spoken lines without the speaker names. Free downloads end with a short spoken outro, which paid plans remove.
Can I re-voice a conversation I recorded?
Yes, if you own the recording or have permission from everyone in it. Transcription can label who is speaking; rename the speakers if you like, give each one a voice in the Cast panel, and generate the whole exchange again in the new voices.
When is another tool the better choice?
When you need a particular real person's voice: there is no voice cloning here, and every speaker is a catalog voice. Turns are voiced one after another, so people talking over each other is not possible, and music or sound effects go on in your own editor. For a live, unscripted conversation with an AI, look elsewhere too, because this voices scripts you have written.
Can I hear a two-speaker scene before signing up?
Yes. The demo on the homepage has a Two speakers tab: edit a line of the sample scene and hear it voiced, with no account. Casting a script of your own takes a free account.