Web Speech API: SpeechRecognition is undefined in Firefox and some browsers
Speech recognition is exposed as window.SpeechRecognition || window.webkitSpeechRecognition and is available in Chrome, Edge and Safari but not Firefox. In Chrome it streams audio to a server, so it fails offline and needs microphone permission on HTTPS or localhost. Feature-detect and show a fallback (typing, or send recorded audio to a speech-to-text API) instead of a blank button.
Checked 2026-11-15 under Playwright 1.65.1 on macOS: Firefox 159 has neither SpeechRecognition nor webkitSpeechRecognition; Chromium 159 and WebKit 27.2 expose webkitSpeechRecognition (WebKit only on secure origins). Related: speechSynthesis.getVoices() returns [] on the first synchronous call in Chromium (191 voices after the voiceschanged event) and Firefox (64 after voiceschanged) but 223 immediately in WebKit, so populate voice pickers from a voiceschanged listener. I only checked presence, not recognition.start().