Speech to Text

Speech to Text lets you speak into your microphone, shows what was recognized as text, and lets you copy it with one click. It doesn't contain its own recognition engine; it calls the browser's built-in SpeechRecognition interface, so whether it works and how accurate it is depend largely on your browser. One thing needs saying up front: browser speech recognition is usually not processed locally. The recorded audio is sent to the browser vendor's cloud servers, and the text comes back. That makes the tool fine for jotting down ideas or dictating a rough draft, and a poor choice for reading out anything sensitive. It cannot transcribe uploaded audio files.

How to use

1. Open the page in a Chromium-based browser such as Chrome or Edge. If your browser doesn't support speech recognition, the page shows a single red notice saying so and suggesting those browsers.
2. Pick the recognition language at the top: 'Chinese' (zh-CN) or English (en-US). Choose before you start, because the language is read at the moment you click the start button.
3. Click 'Start recording'. The first time, the browser asks for microphone permission, which you need to allow.
4. Once running, the button turns into 'Stop', and a pulsing red dot with 'Listening…' appears. Speak normally into the microphone.
5. Whenever the browser decides a phrase is finished, its text is appended to the 'Recognized text' area. Click 'Stop' when you're done.
6. To keep the text, click 'Copy' at the top right; it shows 'Copied'. 'Clear' deletes everything recognized so far.

How results are displayed

From the code, the tool only keeps results the browser marks as final. Interim results are turned off, so you won't see words flickering and being revised as you speak; text appears in one piece after a phrase ends. Recognition runs in continuous mode, so you can keep talking without pressing start again. New text is joined directly onto the existing text. The tool itself adds no spaces, line breaks or punctuation, so punctuation and phrasing come entirely from the browser's engine. The results area is a read-only display, so you can't edit inside it; copy the text elsewhere to make changes.

Supported range and known limits

• Only two recognition languages: Chinese (zh-CN) and English (en-US). There's no Cantonese, Japanese or others.
• Live microphone input only. You can't upload audio or video files.
• Works only in browsers that provide SpeechRecognition, mainly Chromium-based ones. Whether it works and which cloud service is used is the browser's decision, not the tool's.
• Errors are not reported. If you deny microphone permission, the network drops or no sound is detected, recognition simply stops and the page returns to 'Start recording' without saying why. If clicking seems to do nothing, check the microphone permission in your address bar and your connection.
• Recognition can also end on its own without you clicking Stop (for instance after a long silence). The page then returns to the idle state, and you start again when needed.
• Accuracy depends on your accent, background noise, microphone and the browser's engine. The tool makes no accuracy guarantee, and names, jargon and numbers are especially error-prone, so always proofread.
• Copying uses the browser clipboard API and fails silently as well.
• Results live only on the page. Reloading or closing it loses them, so copy what you need.

FAQ: The page says my browser isn't supported.

Try the latest Chrome or Edge. Some browsers don't offer the speech recognition interface or ship with it disabled. Note that even in a supported browser, some embedded web views (such as the built-in browsers of certain apps) may not provide it.

FAQ: I clicked Start but no text appears.

Check in this order: (1) has the browser allowed this site to use the microphone; (2) is the right input device selected in your system, with enough volume; (3) is your network working, since recognition needs an internet connection; (4) does the selected language match what you're speaking? Speaking English in Chinese mode gives poor results. Because the page doesn't show an error reason, troubleshooting is the only way to find out.

FAQ: Why is there no punctuation?

Punctuation is decided by the browser's recognition engine; the tool doesn't insert any. Some engines output it and some don't. If it's missing, paste the text into a document and add punctuation by hand.

FAQ: Does it work offline?

Generally no. The browser interface this tool relies on usually sends audio to the cloud for recognition, so without a connection recognition fails and stops silently.

Privacy and processing

The page itself doesn't send your voice or the recognized text to this site's servers. However, the actual recognition is done by the browser's speech interface, and the page states this clearly: recorded audio is usually sent to the browser vendor's cloud servers for processing and the text is returned. That processing follows the browser vendor's own privacy policy and is outside this tool's control. If what you plan to say is confidential, concerns other people's privacy, or includes passwords, medical or financial details, don't use it. When you leave this page, the tool stops recognition automatically so the microphone is released.