Legal Transcription Services for Law Firms: What Depositions, Hearings and Recorded Interviews Cost

The public benchmark

Federal maximum transcript rates per page

The Judicial Conference of the United States sets the maximum a court reporter may charge for a transcript in a federal case. These rates took effect on 1 October 2024 for fiscal year 2025 and remain the published maximums. They matter beyond federal matters, because they are the only nationwide, publicly stated price for this work and they give you a defensible reference point in any negotiation.

Two things in that table are worth pinning to a wall. First, speed is the most expensive variable in legal transcription by a distance: the same 100-page transcript is $440 on a 30-day turnaround and $870 within two hours, a $430 premium and very nearly double the price for identical words. Second, realtime is charged per page by feed volume, from $3.70 for a single feed down to $1.80 when five or more parties take it, so the more people who want the live stream, the cheaper each one gets.

Copies are the quiet line item. At the ordinary rate, a 100-page original is $440, but two parties each taking a copy adds $220, taking the real invoice to $660 before anyone expedites anything. On a matter with four represented parties, copies exceed the cost of the original.

Worked example

One two-hour deposition, priced four ways

Two hours of testimony, about 80 transcript pages at an average pace. The same event, bought four different ways. Only the first row produces an official certified record, which is exactly why the comparison is useful: it shows what the certification is costing, so you can decide when it is worth paying for.

The third row is the one most firms underuse. It is not an alternative to the first row for a deposition, because there is no oath and no officer. It is the correct product for the large volume of audio a litigation practice generates that never needed a reporter in the first place: recorded witness interviews, client intake calls, insurance recorded statements, internal investigation sessions, expert briefings and arbitration audio.

The fourth row is genuinely useful in one specific way that is worth building into a workflow. Run every recording through automated transcription first, at effectively no cost, use it to decide which files actually matter, then buy human transcription only for those. On a matter with forty hours of audio where six hours are relevant, that single triage step saves roughly $3,000 and costs about $10.

Automation

Where automated transcription belongs in a law firm, and where it does not

The claim that a firm running ten depositions a month could spend over $5,000 with reporters or under $150 with automation is arithmetically true and practically misleading, because the two outputs are not the same document. One is a certified record admissible as the transcript of sworn testimony. The other is an unverified draft with a known error rate.

Put a number on that error rate and the boundary becomes obvious. Speech runs about 150 words a minute, so an audio hour holds roughly 9,000 words. The best engines reach about 96% on clear audio, which is 360 wrong words an hour. On a multi-speaker deposition room with crosstalk, accuracy drops 8 to 12 points, taking a single hour past 1,200 errors and frequently collapsing speaker attribution altogether. A human transcriptionist at 99% or better produces 90 or fewer, and marks the ones they were unsure of.

That last difference matters more than the count. A machine does not hedge. It produces a confident, grammatical sentence that is wrong, and there is no way to find it except by listening to the recording against the transcript, which costs more than having bought the human version. A transcriptionist writes "[inaudible 01:14:22]" and you know precisely where to look.

Keep going

Related pages worth reading next