Speech to Text

This tool itself never uploads your audio anywhere — it uses the Web Speech API's SpeechRecognition interface, which is a standard, built-in browser feature. What happens to the audio internally depends entirely on your specific browser's own implementation: Chrome's built-in speech recognition, for instance, sends audio to Google's speech recognition servers for processing (the same infrastructure behind Google Assistant), since accurate real-time transcription currently requires substantial server-side processing power that isn't practical to run locally in every browser. This is standard behavior for every website using the Web Speech API, not something specific to this tool — if audio privacy is a strict requirement, check your specific browser's documentation for exactly how it implements speech recognition.

The Web Speech API's SpeechRecognition interface has inconsistent browser support — it works well in Chrome, Edge, and other Chromium-based browsers, and is supported in Safari, but Firefox has historically not implemented it at all. If you see an 'unsupported' message, try switching to Chrome, Edge, or Safari; there's no workaround within an unsupported browser since this tool relies entirely on the browser's own native implementation rather than a JavaScript library that could otherwise fill the gap.

That's the interim results feature working as intended — as you speak, the recognition engine continuously refines its best guess at what you just said, sometimes revising earlier words as more audio context arrives (the same way your phone's voice-to-text often 'corrects itself' mid-sentence). This tool distinguishes interim (still-being-refined, shown in a lighter style) from final (locked-in, confirmed) text — only interim text can change or disappear; anything that has been finalized stays in the transcript permanently.

Both stop the microphone from actively listening, and neither one erases your transcript so far — the difference is purely about what you intend to do next. Pause is meant for a brief interruption where you plan to click Start again shortly and keep adding to the same transcript. Stop is meant to signal you're finished with this recording session; clicking Start after Stop begins a fresh listening session that still appends onto your existing transcript, so nothing is actually lost either way — use whichever button matches your intent.

Browsers automatically end a speech recognition session after detecting an extended period of silence, as a built-in safeguard against a microphone being left open indefinitely. This tool detects that automatic stop and, as long as you haven't clicked Pause or Stop yourself, immediately restarts listening behind the scenes so you can keep talking without needing to click Start again — you may occasionally notice a very brief gap in transcription right as this happens.

Punctuation and capitalization quality depends entirely on the underlying browser's recognition engine, not on this tool — modern implementations (particularly Chrome's) do a reasonably good job inferring sentence breaks and capitalizing the start of sentences and proper nouns from natural speech patterns and pauses, but it's not perfect and won't match a human transcriber's judgment. For anything requiring precise formatting, treat the transcript as a strong first draft and review it afterward rather than a final, publish-ready document.

The language selector is locked while actively listening specifically to avoid an inconsistent mid-session state, since a recognition session is bound to a single language for its duration. Click Pause or Stop first, then change the language, then click Start again — the new session will use the newly selected language from that point forward.

This tool itself imposes no time limit, and the continuous auto-restart behavior described above means a session can, in principle, run indefinitely as long as your browser tab stays open and active. In practice, some browsers or operating systems may still impose their own limits or pause microphone access for background/inactive tabs — keep the tab focused and active for the most reliable long-running transcription.