Voice typing in Hindi and English: how to turn speech into text for free, fix a blocked mic, and get cleaner results

Typing a long message on a phone is slow, and typing in Hindi on an English keyboard is slower still. Voice typing flips that: you talk, the text appears, and you fix the few words it got wrong. This guide explains how the Speech to Text tool works, what to do when the mic refuses to start, and how to get text you hardly need to edit.

How to use it, step by step

1. Open Speech to Text in Chrome, Edge or Safari.
2. Under Language you will speak, choose the language. Choose it before you start, because changing it while you are listening stops the session.
3. Press Start speaking and allow the microphone when the browser asks.
4. Talk at a normal pace. You will see ● Listening… with grey words trailing along as the browser guesses.
5. Press ■ Stop when you are done. The text stays in the box, and you can click into it and edit by hand.
6. Use Copy text to paste it into WhatsApp, a document or an email, or Download .txt to save it as speech.txt.

Two small behaviours are worth knowing. First, if the box already has text, new speech is added after it, with a space in between, so you can dictate in several rounds. Second, the tool keeps listening on its own. Browsers tend to end a listening session after a pause, and the tool quietly starts it again until you press Stop, so a short silence will not cut you off.

Which languages can you pick

The language you choose tells the browser which speech model to use. Pick the language you will actually speak, not the one you want written. A Hindi sentence spoken into English (US) mode comes out as odd English words.

Why speaking beats typing, and by how much

People generally speak faster than they type on a phone. A commonly quoted pace for relaxed speech is around 130 words a minute, while many people type 40 words a minute or fewer on a phone keyboard. These are rough figures, not a promise, but they show why dictation feels so fast for a long message.

Where voice typing helps most is the first draft: a long WhatsApp message, an email, notes after a meeting, a school essay outline, or a post in a language whose keyboard you find awkward. Where it helps least is anything full of numbers, names, codes or formatting, because those are the words a speech service gets wrong most often.

Mic blocked or not starting: the fixes, in order

Most failures come down to a handful of causes. Work through them from the top, because the first two fix the majority of cases.

Getting cleaner text

Accuracy depends far more on how you speak than on which button you press. These habits make the biggest difference:

• Hold the phone close and sit somewhere quiet. A fan or a busy road costs more accuracy than an accent does.
• Speak in whole phrases at a steady pace. Words dropped one by one, with long gaps, are harder to read than a natural sentence.
• Do not rush the last word. The end of a sentence is where words are most often clipped.
• Expect to add punctuation yourself. The tool does not add commas or full stops for you, so plan a quick pass to add them after you stop.
• Say names and numbers slowly, then check them. Phone numbers, amounts and spellings are worth a second look every time.
• Dictate in short rounds. Stop, read, fix, then start again. It is faster than correcting one huge block at the end.

After dictating, the text is plain, so tidying it is easy. Fix capital letters with the Case Converter (see the guide to changing case) and check the length with the Word Counter, which the article on word and character limits explains.

What this tool cannot do

This matters because many people search for “audio to text” when they really mean a recording. Playing a recording near the microphone is a poor workaround: the sound is degraded, and echo makes the result worse. Be honest with yourself about which job you have. For a live voice, use this tool. For a saved file, pull the audio out with the video to MP3 guide and use a service built for file transcription.

If your text is in a picture instead, such as a screenshot or a photographed page, that is a different problem, and copying text from an image with OCR is the right route. And to hear text rather than say it, see the guide to text to speech.

Privacy: what actually leaves your device

Be clear about this before you dictate anything sensitive. Toolcloak does not record, upload or store your voice. But the recognition itself is done by your browser’s speech service, and in Chrome and Edge that is run by Google or Microsoft. So your audio may be sent to them, under their policies, not ours.

• Fine for: shopping lists, drafts, notes, casual messages, practice essays.
• Think twice for: passwords, bank details, medical or legal details, and anything you would not say on a phone call to a company.
• Tip: if you must dictate something private, say a placeholder (“account number here”) and type the real number by hand.