October 28, 2026 · 14 min read

Stenographer vs. Speech Recognition: Which Captions Fit

Stenographer or speech recognition for live captions? Compare accuracy, cost, ADA rules and when each fits a church, event or meeting.

A stenographer (a human CART captioner) gives you the highest accuracy and is the safer choice when captions are a formal accommodation or the stakes are high. Speech recognition gives you instant, low-cost captions for everyday sessions, with no booking and no per-hour fee. Most churches, event teams and meeting hosts end up using both, each where it fits.

This guide shows how the two work, how they differ on accuracy, cost and speed, what U.S. accessibility rules say, and how to decide for your next session.

Live Caption AI

Add Live Captions to Routine Sessions

Live Caption AI provides instant captions for churches, meetings, classes, and events, with attendees reading on their phones via QR code.

Explore live captioning

TL;DR

Stenographers and respeakers produce edited, context-aware captions. AI produces near-verbatim text automatically.
Human captioning is often cited at 98% to 99%+ accuracy, against the low-to-mid 90s for speech recognition. Live human CART runs about $120 to $290 an hour in the same source.
Use a human when captions are a formal accommodation, the event is high-stakes, or the content is heavy on jargon. Use AI for routine services, classes and meetings.
Live Caption AI is an AI tool, not a certified or human-reviewed service. It costs $19.99 a month on the Professional plan.

How the two methods actually work
Accuracy, speed and noisy rooms
What it costs a church or event team
What U.S. accessibility rules say
When to book a human captioner
When AI captions are the right call
Where AI gets better with preparation
Captions in more than one language
Our perspective
A practical way to start
A simple decision checklist
FAQ
Sources

How the two methods actually work

A stenographer uses a steno machine, a chorded keyboard, to write sounds in shorthand. Software turns that shorthand into English in real time. In live captioning this is called CART (Communication Access Realtime Translation). Some human captioners instead use respeaking: they listen and repeat the speech, with punctuation, into software trained on their voice.

Speech recognition (ASR, or automatic speech recognition) uses machine learning to turn audio into text with no person in the loop. A speaker’s phone or laptop picks up the audio. The software returns captions within moments.

The key difference is judgment. Research comparing the approaches finds that human captioners, whether using stenography or respeaking, edit for clarity and readability, while AI generates near-verbatim text. A human can tell “their” from “there” from context, know who is speaking, and drop a false start. AI, according to a Forbes Communications Council piece, struggles more with homonyms and speaker identification.

That edited output matters for readers. Academic work on deaf and hard-of-hearing users suggests that false starts and repetitions in near-verbatim ASR can be less readable than the cleaned-up text a professional produces.

The U.S. Department of Justice describes real-time captioning as an auxiliary aid: a transcriber types what is said at a meeting or event so it can appear on a screen. The guidance does not name a specific technology.

How the two methods actually work — overview diagram

Accuracy, speed and noisy rooms

Here is how the two compare in practice.

Factor	Stenographer (CART)	Speech recognition
Typical accuracy	Often cited at 98% to 99%+	Low-to-mid 90s
Output style	Edited, readable, speaker-aware	Near-verbatim, including false starts
Setup	Book ahead, share the agenda and names	Start a session and share a QR code
Specialized terms	Captioner prepares from your materials	Depends on the model and the terms you load
Lead time	Plan around 10 business days at one university	None
Cost	About $120 to $290 an hour	Flat monthly subscription

The accuracy figures come from the Forbes piece. Treat them as typical ranges, not promises. Real results shift with audio quality.

What moves AI accuracy most is the room, not the software:

Microphone distance. A phone on the pulpit or a lapel mic beats a phone across the room.
Crosstalk. Q&A with several people talking at once is hard for AI and for humans.
Accents and soft speakers. Both methods do better with a clear, steady speaker.
Names and jargon. This is where generic AI garbles the most, and where preparation helps (more on that below).

A human captioner can also ask for a repeat or look up a term. AI can’t.

What it costs a church or event team

Professional live human captioning in the U.S. runs about $120 to $290 per hour, depending on remote or onsite delivery and attendance. That figure is per session. A church with two Sunday services and a midweek study can pass that in a month. A conference with several rooms multiplies it quickly. For a deeper breakdown, see our CART captioning cost guide.

AI captioning is priced as a subscription. Live Caption AI’s Professional plan is $19.99 a month with unlimited captioning. The Business plan starts at $199 a month. There is also a Free plan with basic on-device captions. Prices are listed on our pricing page.

Human and AI captioning price structures

The trade is simple. Human captioning costs more per hour and buys certified accuracy. AI costs a small fixed amount and buys frequency: you can caption every service, class and meeting instead of only the few you can afford to book.

There is also a middle path. The same Forbes source notes that hybrid models, which pair ASR with human correction, can cut captioning costs by up to 55% compared with fully human work, while serving readers better than ASR alone. That is a separate type of service from what Live Caption AI provides.

What U.S. accessibility rules say

The ADA requires state and local governments (Title II) and businesses open to the public (Title III) to communicate effectively with people who are deaf or hard of hearing. That includes providing auxiliary aids and services such as real-time captioning or qualified interpreters when needed. Title II entities must give primary consideration to the aid the person with a disability requests. The ADA Tool Kit checklist says the same for state and local governments.

Two points matter here:

The law names the outcome, not the tool. The DOJ says the right aid depends on the nature, length, complexity and context of the communication and on the person’s usual communication methods.
The obligation doesn’t wait for a crowd. The National Association of the Deaf says the duty does not depend on how many attendees ask for the service.

The NAD is also direct about automatic captions: it says computer-generated captions are generally inaccurate and unreliable, so they do not by themselves ensure effective communication. Its best-practice standard calls for qualified captionists and captions that are accurate, synchronous, complete, legible and well positioned.

So be careful with claims. No AI tool, ours included, makes an organization ADA-compliant on its own. Live Caption AI is not a certified or human-reviewed service. What it can do is be one part of a written effective-communication plan. Our guide to ADA effective communication walks through that plan.

When to book a human captioner

There is no federal threshold that says “AI is fine under 200 people.” Universities set their own rules, and they are a useful guide because they have worked through these cases. They are institutional policies, not ADA law.

A formal accommodation. The University of Iowa says human CART is necessary when captioning is a formal accommodation. It suggests considering it when a participant has a hearing impairment, the event is public, or an online event has more than 50 participants.
Large or critical events. Ohio State allows approved automated captioning except for events expected to exceed 200 participants, events conveying critical legal, financial or medical information, events needing immediate action, and events where participants are known to need captions.
Reliance on captions. Harvard says automated captions can supplement but are not yet a replacement for professional live captioning when someone relies on them to communicate.
Jargon and high stakes. The University of Georgia recommends professional human captioning for formal accommodations, commencements, institutional keynotes and heavy technical vocabulary.

Lead time matters. Texas A&M University-Corpus Christi, which describes CART as professional stenographic transcription, suggests planning about 10 business days ahead to get a captioner. If a deaf member of your congregation or a registered attendee asks for CART, ask the person what works for them, and book early.

When AI captions are the right call

AI works well when the goal is to make everyday speech easier to follow for everyone in the room:

Weekly worship services, Bible studies and announcements
Classes, trainings and staff meetings
Community events, panels and town halls with no accommodation request
Rooms where some attendees are hard of hearing, non-native English speakers, or simply far from the speaker
Sessions you could not otherwise caption at all

In each case, the real alternative to AI captions is often no captions. A free, instant option that reaches every phone in the room beats an unbooked stenographer.

AI also gives you something humans do not: a transcript by default. If you save one, review it before you publish it. Georgia’s guidance says recorded automated captions should be reviewed and corrected before sharing.

Where AI gets better with preparation

Generic speech-to-text stumbles on names, places and field terms. That is the gap Live Caption AI is built to narrow:

Specialized models. Choose a Medical, Worship & Ministry, or Finance & Earnings model before you go live.
Custom Terms. Type in up to 100 names, places and terms: your pastor’s name, the town, the product, the speaker list.
No hardware. The speaker runs it on a phone or laptop. Attendees scan a QR code and read on their own phones. No app, no account.

We do not offer a legal or technical vocabulary model, so for deposition-style legal work or dense engineering content, a human is the right pick.

Captions in more than one language

Multilingual events are where AI changes the math the most. Booking a human captioner is hard enough. Booking one for each language is rarely realistic for a small team.

With Live Caption AI, the speaker picks one source language and one target language per session, from 29 available languages. Viewers can switch between the original and the translation on their own phones. Only one target language runs per session, so if your room needs three languages at once, you need a different plan, such as separate sessions or interpreters.

Keep expectations honest:

Machine translation sits on top of speech recognition. Errors in the first step carry into the second.
Idioms, humor, scripture phrasing and fast speech translate less reliably than plain, steady sentences.
Translation is text only. We do not offer translated audio.
For legal, medical or other high-stakes content, use a qualified human interpreter or translator.

Speaking a little slower, using short sentences and loading names as Custom Terms help the translation as well.

Our perspective

We built Live Caption AI because most churches and small event teams were choosing between a $120-plus hourly booking and nothing. Nothing is not a good option for a person who cannot hear the sermon.

We do not think AI replaces stenographers. A skilled CART writer is the right answer for a formal accommodation, a complex proceeding or a room where one person depends on every word. We say so plainly, and we say that Live Caption AI is not certified or human-reviewed.

What we do believe: captions should be normal, not a special request. When a pastor, teacher or organizer can start captions in a minute, with no hardware and no stenographer to schedule, more people can follow along. The best plan is layered. Use AI for the regular schedule. Keep a human captioner on call for the sessions where the law, the stakes or a specific person’s request call for one. And never store audio you do not need: session audio in Live Caption AI is never stored.

— Ryan

A practical way to start

If you are replacing costly human captioning for routine sessions, or adding captions where there were none, start small. Run Live Caption AI on your next service or meeting, load your Custom Terms, and ask a few attendees how readable the captions are. Save the transcript as TXT, SRT, JSON or PDF if you want a record, and use the AI summary to share notes afterward.

The Professional plan is $19.99 a month with unlimited captioning. The Free plan lets you try basic on-device captions first. See Live Caption AI plans and pricing for the full list, including in-app translation credit packs.

Live Caption AI

Explore Live Caption AI

If you want to see how the QR code setup works at an event, read our guide to ADA-friendly event captions with cloud live transcription.

A simple decision checklist

Before each event, ask:

Has anyone requested captions as an accommodation? If yes, book a human captioner and ask what format they prefer.
Is the event public, large, or tied to legal, medical or financial decisions? Lean human.
Is the vocabulary heavily technical? Lean human, or prepare an AI model and Custom Terms and test it first.
Is this a routine session with no request? AI captions are a practical choice.
Do you need a second language? Use AI translation for general content and a human interpreter for high-stakes content.
Is your audio clean? Put a good microphone close to the speaker whichever method you choose.

Write the answers into your effective-communication plan so you can show how you decide.

FAQ

Is a stenographer more accurate than speech recognition?

Yes, in most live settings. Human captioning is often cited at 98% to 99%+ accuracy, while speech recognition typically lands in the low-to-mid 90s. A human also edits for readability and identifies speakers. AI closes the gap on clear audio with prepared vocabulary, but it does not match a certified professional on complex content.

Can AI captions satisfy the ADA?

The ADA requires effective communication, not a specific technology. The National Association of the Deaf says automatic captions generally do not ensure it on their own. No AI tool makes an organization ADA-compliant alone. If a person with a hearing disability requests captioning, consider their preference and talk to a qualified accessibility professional or attorney.

How much does a human captioner cost compared with AI?

Live human CART in the U.S. typically runs about $120 to $290 an hour. Live Caption AI’s Professional plan is $19.99 a month with unlimited captioning. The two serve different needs, so many organizations budget for both.

Do attendees need to install anything to read AI captions?

Not with Live Caption AI. The speaker runs it on a phone or laptop, and attendees scan a QR code to read captions in a browser. There is no app and no account.

Can Live Caption AI translate into several languages at once?

No. Each session has one source language and one target language, chosen from 29. Viewers can switch between the original and the translation. For several target languages at once, plan separate sessions or human interpreters.