Docs/Voices & languages

Voices & languages.

Ten distinct personalities, twenty-three languages, all available on every plan at no extra cost. Voices can be switched per request and even within a single WebSocket session.

The voice library

Each voice ships with a tag describing its default register. You can soften, brighten, or slow any voice with the per-request speed and inline expression tags.

idNameRegisterBest for
meeraMeeraWarm, expressiveCustomer support, friendly agents, conversational
arjunArjunAuthoritative baritoneAnnouncements, formal IVR, news
kabirKabirBright, energeticAds, sports, promotional reads
vikramVikramCalm, wiseGuided meditation, mentoring, support
anayaAnayaGentle, lyricalSoft reassurance, audiobooks, lullabies
saanviSaanviPoised, cinematicStorytelling, trailers, long-form narration
noorNoorCrisp, urbaneEditorial reads, podcast hosting
ariaAriaSparkling, youthfulLifestyle apps, social, e-commerce
orionOrionDeep, measuredDocumentaries, premium voiceover
lumaLumaCool, conversationalSales agents, outbound calling

Hear them all on the samples page or in the home-page studio.

Languages

One model, twenty-three languages. The phonemiser, acoustic model, and vocoder were trained jointly across all of them, so accent-mixing in a single sentence works without an extra round-trip.

CodeLanguageQuality tier
enEnglishFirst-class
hiHindiFirst-class
esSpanishFirst-class
frFrenchFirst-class
deGermanFirst-class
itItalianFirst-class
ptPortugueseFirst-class
nlDutchStandard
plPolishStandard
ruRussianStandard
trTurkishStandard
arArabicStandard
koKoreanStandard
jaJapaneseStandard
viVietnameseStandard
idIndonesianStandard
el, bg, da, fi, hr, lt, sl, svEuropean long-tailBeta

Custom voices

Voice cloning from a 30-second sample is available on the Enterprise plan. Clones are private to your workspace, encrypted at rest, and can be deleted (along with all derived audio) on demand for GDPR compliance.

Contact us to enable cloning on your account.