Live Captioning Software:
Three Delivery Paths, One Choice

Choosing live captioning software starts with a delivery question: where must people read the captions? A public webcast needs text in its player. A private meeting may already have captions in its meeting app. An attendee in a room may prefer text on a personal device. These are useful buying paths, though a product can support more than one.

This guide compares the integration work and limits of each path. For a service procurement, including human CART, use our live captioning services checklist. For a personal setup tutorial, see the live captions guide.

Three Ways to Deliver Live Captions

These paths describe where viewers receive text. They are not mutually exclusive product categories: an event may send one caption feed to a webcast and another display to people in the room.

1. Captions in the video player

A streaming team sends live captions through its production and delivery pipeline. Depending on the system, this may involve an encoder, an ingest service, a caption feed, or timed text tracks such as WebVTT. The team must check the player, language selection, synchronization, and archived output. A caption feed reaching an encoder does not by itself prove that the final recording has usable captions.

This path fits public webcasts and other video deliverables. The key procurement questions are who controls the stream and player, which caption formats they accept, and how the archive will be checked. Apple's HLS authoring guidance, for example, specifies how WebVTT subtitle tracks and accessibility labels should be declared for Apple devices.

2. Captions in a meeting platform

Zoom, Google Meet, and Microsoft Teams can show captions inside a call. Availability, language choices, translated captions, and saved transcripts vary by plan, settings, and participant permissions. Check the platform's current controls before buying another tool. Zoom now separates ephemeral meeting captions from saved meeting transcripts; saying that platform text can never be exported would be misleading.

3. Captions on a separate display

A separate browser tab, phone, or room display can show text without changing the video stream. MirrorCaption uses desktop browser-tab audio capture for supported browser meetings, and a microphone for in-person conversation. It shows source speech and translation in a live session and can save a transcript. It does not add a caption track to a meeting recording or public video player.

For an accommodation request, ask the participant what display works for them. A separate display can help someone follow a conversation, but may be awkward to use alongside video. Human CART is another delivery option when accuracy and individual access needs call for it; see how to request CART.

What Accessibility Standards Change

WCAG 2.1 Success Criterion 1.2.4 calls for captions for live audio in synchronized media. The criterion does not set a percentage accuracy score, but W3C explains that captions should communicate speech, speaker identity, and meaningful non-speech sound in real time. Its guidance says this criterion was intended for broadcasts and does not automatically require every two-way web call to be captioned regardless of participants' needs.

Do not treat delivery architecture as a compliance certificate. Test the actual stream and player, caption timing and completeness, and whether the requested accommodation is effective. A separate text display may be valuable in a meeting; do not assume it makes a public video or its archived version conform to a live or prerecorded media requirement.

The US 47 CFR §79.4 governs certain IP-delivered video programming, including defined programming previously shown on television with captions. It is not a general rule for every workplace meeting or webcast, nor merely a file-format rule. Teams with covered programming should assess the rule's scope and caption quality and delivery duties with their accessibility specialists.

Measure Delay Where the Viewer Sees It

Latency can accumulate at audio capture, recognition or human captioning, synchronization, distribution, and the viewer's player. An engine's response time does not equal the viewer's caption delay. Ask a vendor where its measurement starts and ends, whether it measures partial or final text, and how it reports slower sessions.

Run a pilot with the actual microphones, network, event software, and viewing device. Watch a difficult passage with names, interruptions, and relevant non-speech sounds. Compare the caption to the audio at the viewer's screen. Different streaming configurations may change delay; no one protocol guarantees a specific result for every implementation.

Check Language Support End to End

A language count for speech recognition is only one part of the answer. Confirm the spoken source language, the desired caption language, the translation feature, the caption format, and whether the player exposes the track. Older CEA-608 workflows have limited character support; modern timed-text delivery can provide separate tracks, subject to the player and workflow. Apple's HLS specification requires language metadata for subtitle tracks and a timestamp map for WebVTT.

MirrorCaption offers more than 50 selectable languages in the app, but that does not mean one session can show every language at once. Select and test the required language pair, audio source, and export. If each attendee needs a different target language, plan how each will receive a separate session or display. For spoken translation, see the voice translation guide.

A Short Buying Checklist

1. Identify the viewer and surface. Is this a public player, an internal call, or an attendee's personal screen?
2. Specify the later deliverable. If you publish a recording, check its separate caption or transcript workflow.
3. Ask about access needs. Test readability, timing, speaker identification, and meaningful sounds with the people who will use the captions.
4. Test the exact language pair and devices. Recognition, translation, player tracks, and permissions can each be limiting factors.
5. Compare full costs. Include platform licenses, caption service time, production integration, and any archive editing.

Illustrative example: a city clerk streaming a public meeting needs the player and later recording checked. A separate room display may help attendees but does not create a captioned archive. An internal bilingual meeting might start with its platform's caption controls; a separate display is useful if the needed language or layout is unavailable there. These workflows are examples, not customer accounts or legal determinations.

Frequently Asked Questions

What is live captioning software?

It converts live speech to readable text and delivers that text to a viewer. The main choice is whether the viewer reads it in a video player, inside a meeting app, or on a separate display.

Does live captioning software meet WCAG requirements?

WCAG 2.1 SC 1.2.4 covers live audio in synchronized media. Evaluate the actual video experience, including timing, speech, and meaningful non-speech audio. A separate caption display may help participants, but do not assume it makes a streamed video or its recording conform. W3C says this criterion was intended for broadcasts, not to require every two-way web call to be captioned regardless of user needs.

Why do captions lag behind the audio?

Delay can come from audio capture, recognition or human captioning, synchronization, distribution, and the viewer's player. Ask vendors where they start and stop the measurement, then test the complete viewing path.

Can live captioning software show more than one language at once?

It depends on recognition and translation support, the caption format, the player, and available audio and text tracks. Legacy CEA-608 has limited character support; newer delivery paths can offer multiple language tracks. Test the exact player and language pair rather than relying on an engine's language count.

Do I need a caption encoder or just a captioning app?

Ask where viewers must read captions now and whether a later recording must be captioned. A live encoder may be needed for an external stream, while a meeting platform can also create a separate recording transcript or caption track. Verify the archive output; a live overlay alone does not caption the recording.