Legal transcription: services vs software

"Legal transcription" names two things that are not much alike. One is a service: you send audio to people, and people listen to it and write it down. The other is software: a machine turns a recording into text, on demand, priced by the minute. Both hand you a document with the same words on it, and they are not the same purchase.

What follows is about the difference in kind. It is deliberately not an argument about which produces a better transcript — that is a measurement question, we publish no number for it, and a page that answered it briskly on the way to a sale would be worth distrusting.

What a human service is for

You are buying a person's attention and their judgement. A human transcriptionist hears a mumbled name in the context of the sentence around it, notices when two people talk over each other, and decides what to do about a stretch that is genuinely unclear. All of that happens while they work, and what comes back is finished prose with those decisions already made.

The distinct case, and it is a different job again, is a court reporter: a person producing the record of a live proceeding as it happens. That is a role, not a file format. If someone is producing the record, that is the work, and a tool that transcribes an existing recording does not touch it.

What software is for

You are buying throughput and access. A tool will take more hours of audio than a person can, will return them faster, and is priced per audio minute — so the cost tracks how much material exists rather than how long anyone spends on it. That matters most when the honest answer to "how much of this do you need transcribed" is "all of it, and I do not yet know which parts matter".

What separates one tool from another is not the words on the page. It is which properties of the output you can check without listening to the audio again:

• Whether the stretches it could not transcribe are marked on the page, with times, rather than closed over.
• Whether the completeness of the transcript is measured against the source file independently, rather than by asking the transcriber whether it got everything.
• Whether a speaker label says where it came from — the recording's own structure, a person's identification, or a machine's guess — instead of all three looking alike.
• Whether there is a record of what was actually processed, so the document can be tied back to the file it came from.

These are checkable on a recording that does not matter, before you trust a tool with one that does.

The difference that actually decides it

Both approaches face the same audio and the same ambiguities in it. They differ in who resolves them and whether you can see where.

A person resolves ambiguity as they go, and the result reads as continuous prose. A good transcriptionist marks the parts they could not make out, but the decision about when a passage is unclear enough to mark is theirs, made once, and not visible to you afterwards. You are trusting the judgement, which is the thing you paid for.

A tool that discloses resolves nothing for you. It marks where it had trouble and leaves the trouble on the page, which is less comfortable to read and much easier to audit: you can see the gaps, go to the audio at those timestamps, and settle them yourself. The failure modes are opposite. A service can hand you a clean document that quietly smoothed over something. A tool can hand you a document littered with markers that is entirely honest and still needs your afternoon.

Choosing, in practice

The question that separates the two cases is not how important the recording is. It is what you need to be able to do with the result.

If a person is producing the record of a proceeding, that is a service, and this is not a decision. If you have hours of recorded material and the task is to find what is in it, cite it, and be able to say where each line came from, that is the shape software is for. Many matters have both, at different stages, and treating it as a loyalty test is how firms end up paying for the wrong one twice.

One more thing worth saying plainly, because it decides more of these than anyone admits: if what you need is the lowest possible cost per hour of bulk transcription, this is not that, and it is not trying to be. Pincite Audio is priced per audio minute — the current rate is on the pricing page — and it is built for the recordings where it matters that the transcript is complete, attributable, and citable.