demo · v150 · developer trial

Transcript Post-Processor

Paste a raw speech transcript (no punctuation) and see how a rule-based processor adds commas and sentence-ending punctuation — simulating what Chrome 150's unspokenPunctuation: true mode infers automatically from prosody and grammar. A diff view highlights exactly which punctuation was inserted and where.

Chrome 150's unspokenPunctuation uses on-device neural models on prosody cues. This demo uses a deterministic linguistic ruleset — sentence boundaries, question words, conjunctions — to show what kinds of inferences the engine makes. The output will not match the neural model exactly.
Preset: View:

Raw transcript (no punctuation)

Words: 0 Sentences inferred: 0 Commas added: 0 Questions: 0

Diff view — added punctuation highlighted

Inference rules applied

    // Chrome 150: enable unspoken punctuation
const recognition = new SpeechRecognition();
recognition.unspokenPunctuation = true;  // ← new in Chrome 150
recognition.continuous = true;
recognition.interimResults = true;
recognition.lang = 'en-US';

recognition.onresult = (event) => {
  const transcript = event.results[event.results.length - 1][0].transcript;
  // transcript now contains: "Hello, how are you? I'm fine, thanks."
  // without unspokenPunctuation:
  // transcript would be: "hello how are you im fine thanks"
};

    see also

    implementation reference

    Need the exact API surface, compatibility boundaries, errors, lifecycle, and source links? Read the matching gendn reference ↗