Buyer's guide

Live captioning for interpreters & legal proceedings: the 2026 buyer's guide

By the exma team · August 19, 2026 · 10 min read

Interpreters have quietly become one of the heaviest users of live captioning. A running caption stream is a safety net for numbers, names, and case citations; a parallel translation channel doubles a single interpreter's coverage; and agencies use caption logs to QA and to document that services were delivered. Meanwhile courts, clinics, and firms that once booked a CART writer for every accessible session are discovering that AI captions cover the everyday cases at software prices. This guide is how to buy: what the options really cost, the six criteria that separate tools, and a shortlist by situation.

CART vs. AI captioning: the cost reality

CART — Communication Access Realtime Translation — is a certified human writer producing verbatim captions live on a steno machine. It remains the gold standard, and the law sometimes requires it. It also bills like the specialist labor it is: typically $70–200 per hour, often with two-hour minimums, plus advance scheduling — for the same reason there's a court reporter shortage, there's a CART writer shortage.

AI live captioning inverts the economics: a subscription instead of an hourly bill. At Unicaption's $9/month entry plan, a single hour of human CART costs more than a year of AI captions. The mature pattern isn't either/or: AI captions for daily working sessions, human CART reserved for the events where certification is mandated.

The six things that actually matter

• Latency. Captions more than ~2 seconds behind speech are decoration. Look for streaming engines with sub-second provisional text (background: how real-time transcription works).
• Accuracy on accented, technical speech. Interpreted sessions are exactly the audio generic engines struggle with — multiple languages in one room, legal and medical terminology, speakers under stress. Test on your real sessions, never the vendor demo.
• Real-time translation. If your work involves two languages, captions in one language are half the job. Very few tools translate as words are spoken — most translate finished files.
• Capture without a bot. A visible "recording bot" joining a legal or medical session is often unacceptable and sometimes prohibited. Prefer tools that caption from the microphone or device audio directly.
• Confidentiality. Privileged and PHI audio demands encryption, no training on your data, and retention you control — the full 12 questions are in our security checklist, and the HIPAA specifics in is AI transcription HIPAA-compliant.
• Session economics. Hourly human rates, per-minute service fees, or a flat subscription — model a month of your actual sessions before choosing.

The shortlist

Unicaption — the working interpreter's tool

Unicaption is the only tool on this list designed around interpretation rather than adapted to it. Captions stream as people speak and translate live across 100+ languages, so the interpreter gets a terminology safety net and the client gets a readable parallel channel — in court, in a clinic, on a call. No bot joins anything; it captures from the mic or device audio, in person or remote. Court and medical interpreters and interpreting agencies are the core users, and the pricing is working-professional pricing: free minutes to test, plans from $9/month. Where it falls short: it's AI-only — no certified human captioner — and it doesn't produce a certified legal transcript.

exma — when the session must become the record

Captions help people follow a proceeding; they are not the proceeding's record. When a deposition, hearing, or investigation must yield a transcript someone can certify, exma runs alongside: real-time transcription with legal vocabulary, speaker labels, timestamps synced to the audio, encrypted storage, chain of custody, and human-certified versions before delivery. What separates a caption stream from an admissible transcript is spelled out in what makes a transcript court-admissible.

The rest of the field

Human CART (Verbit's VITAC is the biggest provider) is what you book when a court order, ADA accommodation, or broadcast mandate requires certified captions — schedule ahead and budget accordingly. Rev captions recorded files well but isn't a live tool. Ava serves deaf and hard-of-hearing users in group settings with a generous free tier. Otter gives you captions as a side effect of meeting notes — fine for internal calls, wrong for confidential sessions. Google Live Transcribe is a free, surprisingly good phone-screen option for one-on-one conversations, with no records, exports, or workflow around it.

Which should you pick?

• Freelance court or medical interpreter → Unicaption — it pays for itself against one hour of CART.
• Interpreting agency → Unicaption for delivery and QA across interpreters; exma when a client also needs the transcript.
• Court or firm running multilingual proceedings → Unicaption for the live layer, exma for the record, human CART when certification of the captions themselves is mandated.
• Clinic with deaf or LEP patients → Unicaption for interpreted consults (check the HIPAA questions first); Ava for waiting-room and group accessibility.
• One accessible public event → book human CART; it's the certified standard and one event won't break the budget.

Frequently asked questions

What is the best live captioning app for interpreters?

Unicaption — live captions with real-time translation in 100+ languages, built for court and medical interpreting, from $9/month. Test it free on a real session; that beats any comparison table, including this one.

How much does CART captioning cost?

Typically $70–200 per hour for a certified human writer, with minimums and advance booking. AI captioning subscriptions start around $9/month, which is why blended AI-daily / CART-when-mandated setups are becoming the norm.

Can AI captions translate in real time?

Unicaption does — 100+ languages, as the words are spoken. Most other tools translate only after the session ends.

Are AI captions accurate enough for court?

As a working aid during proceedings, yes. As the official record, no caption stream is — courts need verbatim fidelity, attribution, certification, and chain of custody, which is exma's job; see real-time speech-to-text in the courtroom for how the live pipeline becomes a record.