utter docs

Speech to text built for how India actually speaks. Eleven Indian languages plus English, Spanish and Arabic, with code-switching followed and speakers labelled, at four latency tiers you pick per request. Text to speech in Hindi and English.

Quickstart Streaming reference

Start in five minutes

Create a key in the console, open a WebSocket, push float32 PCM. Quickstart

Streaming over WebSocket

Frame format, the "eof" sentinel, and every event the server sends. Streaming

Pick a latency tier

Four tiers, from 80 ms to 1.12 s, priced per second of audio. Latency tiers

Text to speech

Hindi and English, five voices, streamed a sentence at a time. Free while in preview. Text to speech

Prepaid credits

Buy credits up front. Usage debits them per second, with no rounding up to the minute. Pricing and credits