Live Captions and Speech-to-Text: Turning Speech Into Words on Your Screen

Not so long ago, real-time captions mostly meant subtitles on television. Today, the phone in your pocket or the laptop on your desk may be able to turn speech into text almost as it's spoken. For people with hearing changes, and for the family, friends and colleagues who talk with them, that can turn a frustrating conversation into a manageable one. It helps to know what these features do well, where they stumble, and what to check before relying on them.

What are live captions?

Live captions, sometimes called speech-to-text or automatic transcription, use speech recognition software to listen to spoken words and display them as text on a screen. Unlike prepared subtitles, they are generated in the moment, so they appear with a slight delay and are not checked by a person.

They don't replace hearing, and they don't make speech louder. What they offer is a second channel: you can read what you might have missed. Many people use captions alongside listening, glancing at the text to confirm a word or catch a name, rather than reading every line. That combination of hearing and reading can make conversation noticeably less tiring, particularly in the situations described in hearing but not understanding.

Where do you find them on your device?

On many phones, tablets and computers, live captions are built into the operating system. They are usually found in the accessibility section of the settings, often under a heading related to hearing, captions or subtitles. Once switched on, a caption box may appear over whatever is playing or being said.

Separate transcription apps also exist. Some are designed for in-person conversation, showing a running transcript you can place on a table between you and the speaker. Video-calling programs often have their own caption option inside the call controls. Menu names and locations change between versions, so if you can't find a feature, your device's help pages or user guide are the place to look.

Calls, videos and in-person talk: how do they differ?

Captions behave a little differently depending on what you use them for:

• Videos and recorded media: captions can run over clips, podcasts and streamed programmes, which is often the easiest place to start because the audio is usually clear.
• Phone and video calls: some devices can caption the other person's voice during a call. In video calls, captions are often shown below the speaker's image, which pairs well with lip reading.
• In-person conversation: a phone or tablet listens to the room and shows a transcript. This works well when the device is close to the speaker and the room is fairly quiet.

Where do captions struggle, and what about privacy?

Speech recognition has improved a great deal, but it isn't perfect. Accuracy tends to drop when there's background noise, several people speak at once, the speaker has an accent the system handles less well, speech is fast, or the conversation is full of names and specialist terms. Errors can be small, or they can change the meaning of a sentence. For important information such as appointment details or instructions, it's wise to confirm key points in writing or by asking. This matters even more when you're living in a language you learned later; see hearing loss in a second language.

Privacy is worth a moment of attention too. Some features process speech entirely on the device; others send audio to online services. Some keep transcripts, and some don't. Before using captions for medical, legal, financial or other sensitive conversations, read the privacy information for the specific feature or app, and let the other person know you're using captions.

Key takeaways

• Live captions turn speech into on-screen text in real time, adding a reading channel to listening.
• They are often built into phone and computer operating systems, usually in accessibility settings.
• They can be used for media, calls and in-person conversation, with different strengths in each.
• Accuracy drops with noise, overlapping voices, accents, fast speech and specialist words.
• Check how a feature handles audio and transcripts before using it for sensitive conversations.

Keep learning: if you're supporting someone close to you, signs a loved one is struggling to hear covers how to start that conversation.