Live Caption AI.Open app
← Blog

November 20, 2026 · 14 min read

Caption Readability: Size, Speed, Lines and Timing

What makes live captions easy to read: line length, speed, type size, contrast, timing and speaker labels, with the U.S. guidance behind each number.

Readable captions are short, steady and in sync. Use no more than two lines, about 45 characters per line, and high-contrast sans-serif text. Keep each caption on screen long enough to read, and keep the pace of speech within what a viewer can follow. For a live event, most of that comes down to how the speaker talks, how the vocabulary is prepared and how people see the text. Live Caption AI puts live captions on every phone in the room through a QR code, so you can set up those conditions in minutes.

This guide is for church tech teams, event organizers and AV crews who run live captions and want the audience to actually read them. It uses published U.S. guidance, explains each number in plain words and shows how to apply it on the day.

Live Caption AI
Make Live Captions Easier to Follow
Live Caption AI sends captions to attendees’ phones by QR code, with Custom Terms to help prepare names and specialized vocabulary.
Visit

TL;DR

  • Lines and length: Section508.gov recommends two lines at most and no more than 45 characters per line.
  • Speed: speech above roughly 180 words per minute may be too fast for captions to keep up, so coach speakers on pace.
  • Look: use sans-serif type, strong contrast and mixed case. Guidance ranges from 18-point to 22-point text.
  • Timing and labels: captions should match the audio, stay up long enough to read, and name speakers when it is not obvious who is talking.
  • Live prep: load names and terms before you start. Live Caption AI supports up to 100 Custom Terms per session.

Table of Contents

What “readable” means for captions

Readability is not one setting. It is the sum of five things: how much text appears at once, how long it stays, how it looks, how well it matches the sound, and how easy it is to tell who is speaking.

Five factors that make captions readable

Section508.gov is the federal government’s accessibility guidance site. It says readable captions synchronize with the audio, include all dialogue and important sounds, use correct spelling, grammar and punctuation, and remain on screen long enough to read. It also asks for a consistent style for speakers, sound effects and music, and no distracting effects.

Notice what is missing: there is no single legal formula. The FCC’s caption rules in 47 CFR Part 79 apply to covered video programming, such as TV. They are qualitative. They require captions that are accurate, synchronous, complete and appropriately placed, and that display at a speed viewers can read. They do not set a fixed character count or reading speed. The numbers below come from accessibility offices and professional bodies, and they are the practical yardsticks most teams use.

Line length and line count

The most repeated rule is the simplest: two lines, short lines.

Source Lines Characters per line Type
Section508.gov 2 at most 45 at most, fewer is better 18-point sans-serif
Washington State University Fewer than 3 Fewer than 32 At least 22-point, mixed case
University of Colorado Boulder 2 at most Not specified Centered block, left-aligned lines
DCMP Captioning Key Generally 2 at most Not specified Left-aligned when possible

The sources differ, and that is fine. A 45-character line suits a wide video frame. A 32-character line suits a narrow screen. A phone is narrow, so shorter lines usually read better there.

Why two lines is the ceiling

A viewer reads captions while also watching a speaker, a slide or a stage. Every extra line pulls the eyes away longer. Two lines can be read in a glance. Four lines force the reader to choose between the text and the room.

Where lines should break

Breaks should follow the grammar. UCOP’s captioning best practices advise keeping related grammatical units together and avoiding segments that mix the end of one sentence with the start of the next. Section508.gov suggests breaking at a period or comma where possible.

In practice, a break like “the Lord is my / shepherd” is harder to read than “The Lord is my shepherd,” even though both hold the same words. In a live feed you cannot hand-edit every break. You can, however, help the speaker finish thoughts in clear sentences, which gives the captions natural places to break.

Speed: how fast is too fast

Speed has two parts: how fast the speaker talks, and how long each caption stays on screen.

Speaking pace

Section508.gov notes that speech above 180 words per minute may be too fast for captions. A common glossary figure for a comfortable reading pace is roughly 150 to 160 words per minute. A typical conversational pace is below the 180 line, but excited speakers, auctioneer-style emcees and anyone reading a prepared script aloud can cross it.

The fix is cheap: brief the speaker. Ask them to:

  1. Pause at the end of each sentence.
  2. Slow down for names, numbers and scripture or policy references.
  3. Avoid talking over applause, music or another speaker.

This is the single most effective readability improvement at a live event, and it costs nothing.

How long each caption stays up

The published minimums and maximums are close to each other:

  • WSU advises at least one second on screen, ideally about 0.3 seconds per word, and no lingering through long silences (WSU).
  • DCMP recommends a minimum of 40 frames (one second and ten frames) and a maximum of six seconds, with an exception for background-music notation (DCMP).
  • UCOP says segments generally should not be shorter than about 1.33 seconds or hold more than six seconds of audio (UCOP).

The pattern: a caption that flashes by in half a second cannot be read, and a caption that sits for ten seconds feels stale. Think “one to six seconds.” A speaker who pauses between sentences naturally produces captions in that range.

Type, contrast and placement

Legibility is the part most people notice first.

  • Font: sans-serif, such as Helvetica or Arial, with proportional spacing. Section508.gov recommends 18-point sans-serif text.
  • Size: WSU recommends at least 22-point. When in doubt, go bigger. Captions read from a few feet away need more size than captions read on a handheld screen.
  • Case: mixed case. All caps slow reading and take more width.
  • Contrast: Section508.gov suggests white text on a translucent black background. WSU specifies a 3:1 text-to-background contrast ratio as the floor.
  • Position: centered in the lower third of the frame unless it covers important text. Keep other on-screen text in the upper two-thirds. Avoid scrolling or flashing captions (Section508.gov).
  • Do not cover people: DCMP says captions should avoid covering faces, mouths, names or essential graphics.

If you also run captions or text on a big screen, distance changes everything. Text that works on a phone fails across a hall. The principles in this piece on font size at viewing distance carry over: size depends on how far the farthest viewer sits.

One advantage of a QR-code setup is that each person reads on their own screen, held at their own distance, instead of squinting at a projector. It also keeps captions off your slides, so you do not have to fight with the lower third of the stage.

Accuracy is part of readability

A caption that is perfectly formatted but wrong is not readable in any useful sense. Section508.gov’s list begins with synchronization and accuracy for a reason.

Generic speech-to-text is usually weakest on names and specialized words. For live events, that means pastors’ names, scripture book names, medical terms, sponsors, product names and the names of speakers on a panel. These are the words the audience most needs to get right.

Live Caption AI addresses that with:

  • A Worship & Ministry model for church vocabulary.
  • A Medical model for health-related talks and education sessions.
  • A Finance & Earnings model for investor calls and financial presentations.
  • Custom Terms: up to 100 names, places and terms the speaker types in before going live.

A ten-minute prep routine goes a long way. Before the session, list every name on the program, every unusual place or product, and any acronym the speaker will say out loud. Add them as Custom Terms. Pick the model that matches the room. Then you are not hoping the software guesses your guest speaker’s surname.

Live Caption AI is an AI captioning tool. It is not a certified or human-reviewed service. It is built for churches, classes, meetings and events, not for courtroom or clinical use.

Speaker labels and non-speech sounds

When more than one person speaks, readers need to know who said what. Section508.gov says to identify speakers when it is not obvious, and to caption meaningful sounds such as music and sound effects. The University of Michigan lists speaker identification and sounds such as laughter, applause and music as part of complete captions. It also treats stable labels and consistent presentation as a usability matter: less confusion for people relying on captions.

For the FCC’s rules on covered video programming, 47 CFR §79.1(j)(2) calls for conveying nonverbal information such as speaker identity, music, sound effects and audience reaction to the greatest extent possible.

A few habits make labels work in a live setting:

  • Introduce speakers by name at the start of a panel, so the captions carry the name.
  • Use one style all day: the same name, the same format.
  • Skip labels for a single, obvious speaker at a podium. They add clutter.

We cover this topic in more depth in our guide to speaker labels in captions, and you can browse more in the speaker labels tag.

Live versus recorded: what changes

Recorded video can be edited. Live captions cannot, at least not in the moment. The W3C Web Accessibility Initiative notes that when live captions are later posted as a recording, minor editing may be needed to improve accuracy. That is a useful workflow: caption live for the room, then clean up the text before you publish it.

Live Caption AI saves every session’s transcript automatically to the speaker’s account. Transcripts are readable only by the speaker, and speakers can delete their transcript from History. You can export a transcript as TXT, SRT, JSON or PDF, and AI summaries are included. Session audio is never stored. For a church, that means a sermon transcript you can review and tidy before posting. For an event team, it means a written record of the session without keeping an audio file.

A readability checklist for event day

Use this before doors open.

  1. Brief the speaker. Slow pace, clear sentence endings, pauses for names.
  2. Load Custom Terms. Names, places, acronyms, up to 100.
  3. Pick the right model. Worship & Ministry, Medical or Finance & Earnings, whichever matches.
  4. Check the microphone. Clean audio in, readable text out.
  5. Post the QR code where everyone can see it: on screens, programs, signs and seat cards.
  6. Tell people how it works. Scan the code. No app to install, no account to create.
  7. Test from the back of the room. Read the captions on a phone from the farthest seat.
  8. Plan for translation. If part of the audience needs another language, pick the source and target language before you go live.

If you serve a multilingual audience

Live Caption AI translates into one of 34 languages per session. The speaker picks one source language and one target language, and viewers can switch between the original and the translation. For an AV company running a multilingual conference, that means one target language per session. For a church with a large Spanish-speaking group, it means a second language on every phone with no extra equipment.

Readability rules apply to translated text as well. Translated lines can run longer than the original, which is another reason short sentences from the speaker help.

What this means for your budget

Many organizations hear “captioning” and think of booking a stenographer or renting hardware. A QR-based setup removes both. There is no hardware to rent and no stenographer to book. Attendees use their own phones.

Where the guidance asks for legal compliance, be careful. Live Caption AI can be used as part of an organization’s ADA effective-communication plan, but it does not make an organization ADA-compliant on its own. It is not an ADA certification. Treat it as one part of a plan that also covers your own policies and requests for accommodation.

Our perspective

We hear the same story from churches and event teams. They turn captions on, and then nobody reads them. The text is too fast, the names are wrong, or the screen is too far away.

Our view is that readability is a setup problem you can solve before the event starts. Slow the speaker down. Put the names in first. Let each person read on the phone in their hand.

That is why Live Caption AI is built around a QR code, vocabulary models and up to 100 Custom Terms. No hardware. No stenographer. No app for attendees to install.

The guidance in this article is not complicated. Two lines. Short lines. Steady pace. Clear labels. We think any organization can do it, and do it affordably.

Ryan

Put it into practice

If you want captions on every phone in the room, start with the Live Caption AI pricing page. The Professional plan is $19.99 a month with unlimited captioning and summarization. A Free plan and a Business plan (from $199 per month) are also listed there. Translation credit packs are available as one-off purchases: 30 credits for $6.99, 60 credits for $11.99 and 120 credits for $19.99.

Live Caption AI

Set up Custom Terms, choose your model, share the QR code and go live. Explore Live Caption AI.

FAQ

How many lines should a caption have?

Two lines at most. Section508.gov recommends no more than two lines and no more than 45 characters per line, and fewer characters are better. Washington State University recommends fewer than three lines and fewer than 32 characters per line, which suits narrower screens.

What font size is best for captions?

Section508.gov recommends 18-point sans-serif text, and Washington State University recommends at least 22-point. Use mixed case and strong contrast, such as white text on a translucent black background. Bigger is safer when viewers sit far from the screen.

How fast can captions be before people cannot keep up?

Section508.gov notes that speech above 180 words per minute may be too fast for captions. Each caption should stay up at least about one second, and published guidance generally caps a single caption at about six seconds. The most reliable fix is a speaker who pauses between sentences.

Do live captions have to meet an FCC reading speed?

The FCC’s Part 79 rules for covered video programming set qualitative standards. Captions must be accurate, synchronous, complete and placed so they do not block important visuals, and they must display at a speed viewers can read. They do not set a fixed number of characters or words per minute. Those rules cover video programming such as TV, so check which rules apply to your situation.

Should captions show who is speaking?

Yes, when it is not obvious. Section508.gov and the University of Michigan both advise identifying speakers and keeping the style consistent. On a panel, have the moderator introduce each guest by name so the captions carry it.

Sources

caption readability

Try it

Put captions in the room.

Free on-device captions forever. Broadcast, translation, and AI summaries from $19.99/month.

No credit card required. 7-day trial, cancel anytime.