Narration is the one job where a voice has to hold up for hours rather than seconds. 163 of the 244 voices here carry the narration tag, the largest group in the catalog, across English, Hindi and nine other Indic languages.
VOICES FOR THIS USE CASE
What narration asks of a voice
A voice that is pleasant across thirty seconds is not necessarily tolerable across nine hours. Narration rewards consistency over character: an even pace, predictable emphasis, and no mannerism that becomes noticeable on the fortieth repetition. That is a different casting problem from advertising or dialogue, and it is why the narration tag exists separately.
Consistency across a long script
Requests cap at 250 characters, so a chapter is assembled from many calls rather than one. Keeping voice, speed and sample rate fixed across the whole run removes most variation. Generate a full page and listen across the joins before committing to a book, because that is where drift becomes audible.
The largest group in the catalog
163 voices carry the narration tag, roughly two thirds of the catalog. That breadth means you can cast for register rather than availability, which is rarely true of the more specialised tags. It spans every accent group and all 31 languages the model supports.
Regional language narration
Nine Indic languages carry recommended voices. Narration in Tamil, Telugu, Kannada, Marathi, Gujarati, Punjabi, Odia and Bengali is thinly served across the industry, and 144 voices here carry an Indic recommendation. For audio publishing aimed at Indian readers, that depth is unusual.
SPECIFICATION
Sample rate
44.1 kHz native, resampled to 8 kHz for telephony
Latency
~200 ms to first byte (p50, warm region)
Output formats
ulaw, alaw, PCM 16-bit, WAV, MP3
Streaming transports
WebSocket, HTTP chunked transfer
Speed range
0.5× – 2.0×, set per request
Explore Voice Similar to
AI Narrator Voices
FAQs