Sona for iOS is coming·Join the waitlist →
SONA ENGINE API

The dictation engine, in your product

The same streaming speech-to-text that powers Sona — clean punctuation, live corrections, technical vocabulary — behind one WebSocket.

Get an API keyRead the docs
Free tier for prototyping · usage-based after
transcribe.js
const sona = new SonaEngine({ apiKey: env.SONA_KEY });

const stream = sona.transcribe.stream({
  language: 'auto',
  vocabulary: ['kubectl', 'PyTorch', 'Sona'],
  punctuation: true,
});

stream.on('partial', (t) => editor.preview(t.text));
stream.on('final', (t) => editor.insert(t.text));
01 — STREAMING

Partials over WebSocket

Low-latency partial results while the user is still talking, finalized text on pause — the same pipeline the Sona app runs on.

02 — PRECISION

Word-level detail

Timestamps and confidence per word, speaker turns, and clean punctuation — ready for captions, editors, and agents.

03 — CONTROL

Vocabulary per request

Pass domain terms with each session — product names, tickers, drug names — and the engine biases toward them instantly.

BENCHMARK

Swap in Sona with two lines

The API is drop-in compatible with common STT client shapes. Most teams switch in an afternoon and keep their existing audio plumbing.

Start benchmarking →
Word-level accuracy, long-form dictation
Sona Engine97.2%
Big-cloud STT93.4%
Open-source model88.9%
Internal long-form dictation benchmark, May 2026.

Ship voice input this sprint

Free tier, no card. Keys are issued instantly.