Text to speech for audiobooks, stories and podcasts

An audiobook is judged at minute forty, not minute one. A voice that sounds fine in a demo but flat over an hour loses the listener. In Chariot's own evaluation on English and Hindi, naturalness (NISQA) scored 4.91 in English and 4.97 in Hindi, tied for first in English and first in Hindi among the systems tested.

The clip below is Meera, one of 12 native Hindi voices, reading the opening of a Hindi story. The voices also carry mood: a whisper, a delighted line, a deadpan one, which is what a story needs when the narrator moves between characters.

There are more than 50 voices to choose from: 12 native Hindi, 8 British English, 6 American English, 5 Indian-accent voices (Assamese, Nepali, Sindhi) and 2 native voices each for Urdu, Bengali, Marathi, Telugu, Tamil, Gujarati, Kannada, Malayalam, Odia and Punjabi. Speed is adjustable from 0.5x to 1.5x.

बहुत समय पहले, नदी किनारे के एक छोटे से गाँव में एक बूढ़ा मछुआरा रहता था। हर सुबह, सूरज निकलने से पहले, वह अपनी पुरानी नाव लेकर नदी में निकल पड़ता।

PlayPauseLoadingBufferingReplayRetry

0:13

Meera · Hindi voice

Chariot reads 15+ languages, including English, Hindi, Urdu, Bengali, Marathi, Telugu, Tamil, Gujarati, Kannada, Malayalam, Odia, Punjabi, German and Chinese. Each Indian language has native voices, and every voice can read every supported language, so one narrator can read the Hindi edition and the English edition.

Modern Indian fiction is rarely in one language. A Hindi novel has English dialogue, brand names and job titles inside Hindi sentences. Chariot reads Hinglish and other mixes of an Indian language with English as one sentence, so the narration does not stumble at every switch. It also reads Hindi written in Roman letters.

In the same evaluation, Chariot had the lowest Hindi character error rate (7.73%) of the seven systems tested in Hindi. For a long book, that rate is the difference between a few wrong words in the book and a few wrong words on every page.

A history book is full of dates. A health book is full of doses. A business book is full of rupee amounts and percentages. Chariot reads written text the way a person says it: numbers, dates, times, currency, units and abbreviations come out as speech. "Dr. Rao" is read as doctor, "500 mg" as a dose, "₹2,50,000" as an amount.

In Chariot's own evaluation, the model read hard written text (dates, currencies, URLs, units) correctly 94.0% of the time, first of the nine systems tested.

The clip below is Neha reading a line with a doctor's title and a blood group, two things a less careful voice reads as letters.

Um, excuse me sir, can you tell Doctor Rao that my blood group is B positive. I'm really worried.

PlayPauseLoadingBufferingReplayRetry

0:05

Neha · Hindi voice

Yes, if the author records 10 to 60 seconds of speech (about 30 seconds works best). The cloned voice is ready in about 10 to 20 seconds and speaks every language Chariot speaks, so an author who recorded in Hindi can also narrate the English and Marathi editions.

Cloned voices are included on paid plans: 3 on Starter, 10 on Creator, 50 on Startup and 500 on Scale. Consent is required, and you may only clone voices you have the right to use. For a publisher, that means a signed agreement with the author or narrator before the clone is made.Cloned voices are included on paid plans: 3 on Starter, 50 on Startup and 500 on Scale. Consent is required, and you may only clone voices you have the right to use. For a publisher, that means a signed agreement with the author or narrator before the clone is made.

Credits count generated audio: about 1,000 credits per minute. A six hour audiobook is about 360 minutes, so about 360,000 credits. A 20 minute podcast episode is about 20,000 credits.

Starter at ₹499 a month gives 100,000 credits, about 100 minutes, with commercial usage rights. Creator at ₹999 gives 250,000 credits, about 250 minutes, and 10 cloned voices. Startup at ₹4,999 gives 1,500,000 credits, about 1,500 minutes or 25 hours, which is about four six hour books a month. Scale at ₹29,999 gives 10,000,000 credits, about 10,000 minutes, for a catalogue. Annual billing gives 2 months free.Starter at $5 a month gives 100,000 credits, about 100 minutes, with commercial usage rights. Startup at $49 gives 1,500,000 credits, about 1,500 minutes or 25 hours, which is about four six hour books a month. Scale at $299 gives 10,000,000 credits, about 10,000 minutes, for a catalogue. Annual billing gives 2 months free.

The Free plan gives 10,000 credits, about 10 minutes, with every voice and no card. That is enough to generate a sample chapter in three voices and decide before you spend anything.

Commercial usage rights come with every paid plan, so you can sell the audiobook, run ads against the podcast or license the audio. Audiobook stores set their own rules: some accept AI narration if it is labelled as such, and some do not accept it at all. Check the store you plan to sell on before you produce the whole book, not after.

For production, the API (POST https://api.chariot.in/v1/tts) returns a WAV file and takes up to 1,000 characters per request, so a book is generated a few paragraphs at a time and joined. To work through a long script by hand, use Studio. The web app at platform.chariot.in downloads MP3 or WAV for a short piece or a podcast episode.

Text and audio are processed in India.