Finds who is speaking
Narration is separated from dialogue and every quoted line is given to a character. Lines the model is unsure about are flagged for you.
- Quotes
- Dashes
- Speech tags
Upload an EPUB, DOCX, PDF or TXT. Voixa works out who is speaking in every line, gives each character a voice, and reads each chapter as one audio file with the whole cast.
Splitting chapters and finding the speakers is free. Tokens are used when voices are designed and chapters are read.
Real Voixa output, not a mock-up: each line is read in its character’s voice and the text follows the audio.
Features
The same voices, tokens and Studio as the rest of Voixa.
Narration is separated from dialogue and every quoted line is given to a character. Lines the model is unsure about are flagged for you.
Voices are suggested from each character’s gender, age and how the story describes them. Main roles never share a voice.
Voixa can create a new voice for a character from their description and keep the take whose pitch fits their age and gender. Every character voice on this page was made that way.
Use a cloned voice, such as yours or a parent’s, as the narrator. Clones need the speaker’s consent and stay in your account.
Names, old words and dialect: set how a word is read once and every chapter follows.
Chapters are split by their headings, such as “Chapter 12” or “Chương 12”. Add new chapters to the same story as you write them.
Every chapter plays with its text highlighted line by line, like the player above.
Invite people to the story’s project. They can fix lines and voices; the owner pays and decides what gets read.
Real example
Nobody marked who says what. This is how Voixa read the opening of “Block D”.
The lights in Block D went out at nine, the way they did every Thursday, and the stairwell filled with the sound of doors opening. “Not again,” Mara groaned, holding her phone up like a lantern. “I had three pages left on my essay.” Down on the landing, old Mr. Okafor was already sitting on his folding chair, a candle in a jam jar at his feet. “Thirty years in this building,”
The narrator reads the narration and each character reads their own lines, even when a speech tag sits in between.
Nicknames, titles and short forms are grouped under one character. When two entries may be the same person, Voixa suggests a merge.
When the narrator is a character (“I said”), Voixa notices and gives the narration that character’s voice.
Sample stories
Pick one to play it in the player above.
The folk tales are retellings written for Voixa and “Block D” is an original story. Durations are the real length of each file.
How it works
Tokens are only used when Voixa designs a voice or reads a chapter.
EPUB, DOCX, PDF, TXT or Markdown, up to 30 MB. A short story or a whole series.
FreeVoixa splits chapters by their headings and points out what is worth a look: front matter, very long chapters, repeats.
FreeEvery character gets a suggested voice. Listen to their real lines in each voice and pick one.
FreeDesign voices for the main roles if you like, then read one chapter or many. Each chapter becomes one audio file.
Use cases
FAQ
It turns a written story into an audiobook with a cast. Voixa splits the book into chapters, separates narration from dialogue, works out which character says each line, suggests or designs a voice for every role, and reads each chapter as one audio file.
EPUB, DOCX, PDF with selectable text, TXT and Markdown, up to 30 MB per file. Scanned PDFs without a text layer are not supported. You can add more chapters to a story later.
Vietnamese and English are fully supported, with a separate voice for every character. Stories in the other languages Voixa reads can be voiced too, with the voices available in that language.
In our tests on a Vietnamese short story and a Sherlock Holmes chapter, every line we had labelled by hand went to the right speaker. Voixa still marks the lines it is unsure about, and you can give any line to another character.
Yes. Every character gets suggested voices based on age, gender and personality. You can also pick any voice from the library, one of your cloned voices, or design a new voice for that character.
Characters with only a few lines share a small set of alternating voices, so two people talking to each other never sound the same. Give any of them their own voice when they become important.
Add it to your pronunciation dictionary with the way it should be read. The rule applies to every chapter of every story in your account.
One audio file per chapter, with the narrator and every character. Play it in Voixa with the text following along, or download it.
Only if you wrote the story or hold the rights to adapt it into audio. Listening privately is always fine. Voixa asks you to confirm your rights before a chapter is read.
Uploading, splitting chapters and detecting speakers are free. Reading uses tokens by the number of characters, about one token per character in Vietnamese and English, and designing a voice uses a fixed number of tokens. Every account gets free tokens every day.
Upload a story and see which characters Voixa finds. Finding them is free.