Live captions are real-time subtitles: the words appear on screen as someone speaks, during a video, a call, a live stream, or a face-to-face conversation. That makes them an accessibility feature and a following-along feature, not a production feature — and that distinction is the thing people trip on. Live captions are generated on the fly and, in almost every tool, thrown away the moment the audio stops. They help the person in the room right now; they do not leave you with a file.
This guide covers the three places you actually turn live captions on: your device (phone, tablet, or computer, which can caption any audio playing — including a conversation held up to the mic), your video-call and meeting apps (Google Meet, Zoom, Microsoft Teams), and your live stream (YouTube Live). It also covers the one step everyone skips — setting the spoken language so the captions are accurate — and what to do when you need the words to survive the call instead of vanishing with it.
The steps
Decide first: live captions, or captions you keep?. These are two different jobs. Live captions run in real time so a viewer or participant can follow along, and they disappear when the audio ends. Captions on a finished, published video are burned into the pixels or shipped as an SRT file and stay forever. If your goal is accessibility during a call or stream, follow the live-caption steps below. If your goal is a captioned clip to post, that is a separate workflow — see the companion guide on adding captions to a video.
Turn on system-wide live captions on your device. Both major platforms can caption any audio the device plays — a video, a call, or a conversation you hold a phone up to. On Android, open Settings, then Accessibility (on Samsung, Hearing enhancements), and turn on Live Caption; you can also toggle it from the volume-bar menu. On iPhone, iPad, or a recent Mac, go to Settings (or System Settings), then Accessibility, then Live Captions, and turn it on. Both run on-device, so the captions work without a connection and the audio never leaves the device.
Switch on live captions inside your video-call app. In Google Meet, open the meeting's three-dot menu, then Captions, and turn them on; Zoom and Microsoft Teams both expose captions in the in-call toolbar (in Teams it is More actions, then Language and speech, then Show live captions). These are per-meeting and per-participant — turning them on shows them on your screen only, so each person enables their own. None of these save the captions by default; that is a separate record-and-transcribe step.
Set the spoken language before you rely on the captions. Accuracy collapses when the caption engine is listening for the wrong language. In every tool there is a spoken-language setting — set it to what is actually being said in the room, not just your interface language. If your audience needs another language, Google Meet and Teams both offer translated live captions (each viewer picks their own target language), but each gates it differently: Meet needs a Business Standard/Plus or Enterprise Workspace edition (or a Gemini add-on), while Teams needs the meeting organizer to hold a Teams Premium (or Microsoft 365 Copilot) license — an ordinary Microsoft 365 Business plan gets same-language captions but not translation. Confirm your specific license before promising it to an audience.
Enable live captions on a live stream. If you are broadcasting, the captions have to be switched on at the source. On YouTube Live, go to Create, then Go live, then Stream, turn on Closed captions in the stream settings, and choose Automatic captions as the source; YouTube's own help page notes the rollout is still gradual and may not have reached every channel yet, and it runs English speech recognition only, in real time. Most streaming software and webinar platforms expose a similar per-stream caption toggle. Turn it on before you go live — you cannot retroactively add live captions to a broadcast that already aired.
Caption an in-person conversation. For a face-to-face conversation, the device-level live caption from step 2 works if you point the microphone at the speaker. For a dedicated experience, a real-time transcription app (such as Google's Live Transcribe) shows speech as large on-screen text and is built for conversations rather than media playback; the same Android app bundles a separate always-on Sound Notifications feature that alerts you to things like doorbells and alarms, independent of the live transcription. These are accessibility aids for the moment — treat accuracy as approximate and never rely on them for anything high-stakes or medical.
Keep the words if you need them later. Live captions are not a transcript. If you want the words after the fact — to reuse, quote, or caption a recording — you have to capture them deliberately: start transcription in your meeting app (Meet, Zoom, and Teams each have a separate Record or Transcribe control that produces a downloadable file), or record the session and run it through a caption or transcription tool afterward. Decide this before the call starts, because there is no way to recover live captions once they have scrolled off screen.
Common gotchas
Live captions are almost never saved. When the call, video, or stream ends, they are gone — if you need the words, start a transcription or record the session instead.
On-device live caption accuracy varies with audio quality, accents, and crosstalk. Apple's own note says Live Captions should not be relied on in high-risk or emergency situations, and the same caution applies to every real-time engine.
Device and language gating is real: Apple Live Captions needs a recent device with the system language set to English (U.S.) or English (Canada), and Android's Live Caption and call-captioning support varies by manufacturer and model. Check your specific device.
Translated live captions are gated differently per tool: Google Meet needs a Business Standard/Plus or Enterprise Workspace edition (or a Gemini add-on); Microsoft Teams needs the meeting organizer on Teams Premium or Microsoft 365 Copilot — a standard Microsoft 365 Business license only gets same-language captions. Free accounts on either tool get captions in the spoken language only. Confirm the specific license before you promise multilingual captions to an audience.
A live caption is not the same as the caption on a published clip. Turning on live captions during recording does not put captions on the exported video — that still needs a burn-in or SRT step afterward.
Live stream captions must be enabled before you go live. There is no way to add real-time captions to a broadcast after it has aired; only the recorded replay can be captioned later.
Legal note
Captioning or transcribing a conversation can carry consent obligations. Apps may show a notice when transcription starts, but a banner is not consent — some regions require every party to agree before a conversation is captured, and using someone's words in published content is a separate permission. Announce it, and get explicit sign-off before you turn a captured conversation into public content.
Where Kompozy fits
Live captions solve the live moment, and only the live moment. They let a deaf viewer follow your stream, let a participant read a noisy call, let you caption a conversation held up to a phone — and then they scroll off and are gone. The instant you want that moment to become something people can find, watch on mute, and share, live captions have nothing to hand you: no file, no styling, no track on the exported clip. That hand-off from the real-time moment to a kept, accessible, published asset is the gap [Kompozy](/) is built to close, and it closes it by making captions automatic rather than a feature you remember to turn on.
Kompozy is a full AI content generation and multi-platform publishing engine, so the caption is not a step you perform — it is baked into every video the engine ships. Drop a recording of that stream or call in and [Clipped Shorts](/glossary/clipped-short) cuts it to vertical with burned-in, word-synced captions, while [Persona Shorts](/glossary/persona-shorts) and the other video formats render their own caption track through the same pipeline — so every clip is sound-off ready and screen-reader friendly without a per-video decision. From one source it also fans out into formats a caption tool will never touch: [Quote Graphics](/glossary/output-buckets) of the sharpest lines, a Carousel, a recap Blog Article, and an Email Newsletter, all held to one [Persona Brief](/glossary/persona-brief) so the voice stays yours. Then [Autopilot](/glossary/autopilot) schedules the captioned set across the eight social platforms plus blog and email behind a per-post review gate. The result is accessibility at scale: instead of flipping live captions on for one viewer in one meeting, every published piece ships captioned by default. Starter ($199/mo for 5,500 credits) fits a solo creator captioning a weekly stream; Pro ($499/mo for 18,000 credits) suits a team turning every call and broadcast into a captioned content set on autopilot; Enterprise is custom for agencies running captioned video across many brands.
Frequently asked questions
What is the difference between live captions and regular captions?
Live captions are generated in real time as someone speaks — during a call, video, or conversation — and are discarded when the audio ends. Regular captions are added to a finished video and stay with it permanently, either burned into the pixels or uploaded as an SRT subtitle file.
Are live captions saved anywhere?
Usually not. Device live captions and the default in-call captions in Meet, Zoom, and Teams are ephemeral and vanish when the audio stops. To keep the words you have to start a separate transcription or record the session, which produces a downloadable file.
How do I turn on live captions on my phone?
On Android, open Settings, then Accessibility (or Hearing enhancements on Samsung) and turn on Live Caption — you can also reach it from the volume-bar menu. On iPhone or iPad, go to Settings, then Accessibility, then Live Captions and turn it on. Both caption any audio the device plays, on-device.
Can I get live captions in another language?
Yes, through translated live captions. Google Meet and Microsoft Teams both let a viewer read captions translated into their own chosen language in real time, but the licensing differs: Meet needs a Business Standard/Plus or Enterprise Workspace edition (or a Gemini add-on), while Teams needs the organizer on Teams Premium or Microsoft 365 Copilot — an ordinary Business plan is not enough for either. Check your specific plan first.
Do live captions work for in-person conversations?
Yes. Point your phone's microphone at the speaker with device-level Live Caption on, or use a dedicated real-time transcription app like Google's Live Transcribe, which is built for conversations and shows large on-screen text. Treat the accuracy as approximate.
How accurate are live captions?
They are good on clear audio and a correctly set spoken language, but accuracy drops with accents, background noise, overlapping speakers, and technical jargon. Because they run instantly with no human review, do not rely on them for high-stakes, legal, or medical situations.