Juris Compiler AI

Guide

How to Transcribe a Deposition Recording, Step by Step

By Juris Compiler AI · Updated September 27, 2026

To transcribe a deposition recording, confirm that nothing restricts sharing the file with an outside service, keep your own copy of the original, upload the audio or video, approve the price, and review the speaker-labeled transcript line by line. With Juris Compiler, a 2-hour deposition costs $90: $0.75 per minute, rounded up to the whole minute.

The result is a working transcript for review and drafting, not a certified transcript.

Steps 1 and 2: Before you upload

Step 1: Check what limits sharing, and keep your original

Confirm that nothing restricts sharing the recording with an outside vendor, such as a protective order, a confidentiality designation, a stipulation, or your client's instructions. That check is yours to make for your own matter, and if an order is unclear, talk to counsel. The Terms of Service also make you responsible for having the legal right to upload each recording.

The recording goes to encrypted storage, then to a transcription provider. Only the transcript text, not the audio, is sent to a language model for analysis. The security page lists the subprocessors and what each one receives.

Step 2: Prepare the file

The file picker lists MP3, MP4, M4A, WAV, MOV, WebM, and MKV, and it accepts other audio and video types too. For a video deposition, the audio track is transcribed. At upload, a file is turned away only if its length can't be read. Transcription is English only.

Multi-hour recordings are accepted, so a full session can usually go up as one file. One file per session helps, because each file becomes its own transcript and the contradiction finder and Ask the Record work within one transcript at a time.

Steps 3 to 5: Upload, pay, and let it process

Step 3: Upload the file

You can add up to 10 files in one upload, with one payment for all of them. Large files go up in parts, straight from your browser to storage through short-lived signed links, and a part that fails is retried. A matter name, case number, and court are optional and can be added now or later.

Step 4: See the cost and pay

The upload panel shows an estimated price for each file at $0.75 per minute, rounded up to the whole minute, with a 1-minute minimum. The billable length is then measured on the server, and the exact amount appears on Stripe's checkout page, where you approve it before you are charged. For scale, 90 minutes is $67.50 and 7 hours is $315, with no subscription and no monthly minimum. The pricing page has a calculator, and the deposition transcript cost guide compares that price with an industry estimate for a court reporter, with its assumptions stated.

Accounts created with Google or Facebook get 30 free minutes, no credit card. Email sign-ups don't. Free minutes are credited automatically, and only the rest is billed. A file fully covered by free minutes skips checkout. If you're new, you can start free.

Step 5: Let it process

Files are processed one at a time, in upload order, so a batch takes longer than a single file. No turnaround time is guaranteed. When a transcript is ready, a case summary is generated automatically and the uploaded recording is deleted.

Step 6: Review the transcript

Transcription is fully automated, and no person reads the file before you do. The review is yours.

Name the speakers

Speakers are labeled A, B, C, and so on. Rename each label to the right person, and the names carry into the PDF, the SRT file, and drafted documents. If a line is credited to the wrong person, reassign it, or split it off as a new speaker if the engine gave two people one label. Reassignments are logged.

Check the flagged lines

Every line shows the engine's confidence as a percentage: green at 90% and above, amber from 80 to 89%, and red below 80%. Lines below 65% also get an amber marker on the left edge. A high score means the engine was sure, not that it was right, so read the green lines too. Check names closely: a fixed list of legal words is boosted, but you can't add your own party names.

Correct lines against your own copy

Correct any line in place. The original machine wording is kept, and the edit history logs the line, the text before and after, who changed it, and when. To settle a flagged line, find its start time in the copy you kept in Step 1 and listen there.

Step 7: Put the transcript to work

  • Case summary. A plain-English overview, legal issues, key parties, possible actions, and a risk rating. The possible actions are suggestions for your review, not legal advice.
  • Ask the Record. Plain-English questions about this transcript, with answers grounded in the transcript. Answers aren't saved, so copy what you need.
  • Contradiction Finder. Statements in tension within this one transcript, each with speaker and timestamp, flagged as potential impeachment for your review. It does not compare one deposition against another.
  • Redaction. Masks detected identifiers on screen and in the PDF, without changing the underlying transcript. It does not apply to the SRT file, drafted documents, or Ask the Record answers. Organization names aren't masked, and detection can miss things, so read before you share.
  • Exports. A PDF with numbered lines, timestamps, and speakers, marked as AI-generated and not a certified transcript, plus an SRT caption file. There is no page:line pagination and no Word export. Both exports need a paid transcript.
  • First-draft documents. An incident report, a deposition designation sheet, and an investigative report, each labeled a draft. The designation sheet cites timestamps, not page and line. The designation sheet guide covers mapping each entry to the certified page:line.

The demo walks through a fictional deposition in the product interface: a speaker-labeled transcript with confidence scores, a case summary, the contradiction finder, and Ask the Record. No account is needed.

Step 8: Match passages to the certified transcript

Anything filed with or quoted to a court typically comes from the certified transcript, not the working one. Under Federal Rule of Civil Procedure (FRCP) 30(f)(1), the deposition officer certifies the record, and a transcript made later from the recording does not carry that certification.

By common practice, deposition testimony is cited by page and line, such as 45:3-47:12. The working transcript has no pages, so one approach is to find passages in it, then take the page and line from the certified transcript.

Rule 30(b)(3)(A) allows audio or audiovisual recording unless the court orders otherwise, so some depositions have no stenographic transcript at all. State rules differ, and courts can change the federal defaults. Check your court's rules, local rules, and any standing or pretrial order, and talk to counsel about your matter.

Everything in this guide starts after the deposition, from a recording that already exists. The recording method is set before that: the deposition notice states it under Rule 30(b)(3)(A), and under Rule 30(b)(3)(B) any party may add another method with prior notice, paying for that extra record unless the court orders otherwise. Our guide to court reporters and AI transcription compares what each one produces.

Frequently asked questions

Does the file format of my deposition recording matter?

Usually not. Juris Compiler AI accepts audio and video files, and the file picker lists MP3, MP4, M4A, WAV, MOV, WebM, and MKV. For a video file, the audio track is transcribed. A file is turned away at upload only if its length can't be read. If that happens, the app suggests trying a standard MP3, MP4, WAV, M4A, or MOV file.

What if one deposition session is split across several recording files?

You can upload the parts together, up to 10 files with one payment. Each file becomes its own transcript and is billed on its own, rounded up to the whole minute. The contradiction finder and Ask the Record work within one transcript at a time, so they won't connect statements across parts. Put the parts under the same matter to search testimony across all of them.

How do I fix a line credited to the wrong speaker?

Reassign that line to the correct speaker. If the engine gave two people the same label, split the line off as a new speaker. Reassignments are logged in the edit history. Speakers start out labeled A, B, C, and so on, so rename each label once and the names carry into the PDF, the SRT file, and drafted documents.

Does correcting a line erase what the machine originally transcribed?

No. Juris Compiler AI keeps the original machine wording and marks the corrected line as edited, with its confidence shown as 100%. That figure reflects your edit, not a new check by the engine. The edit history logs the line, the text before and after, who made the change, and when. The exported PDF shows only the current text, without confidence scores or the edit history.

Why can't I export a PDF or SRT file of my transcript?

Exports need a paid transcript. A transcript processed entirely on free minutes can still be reviewed, corrected, searched, and analyzed, but its PDF, redacted PDF, and SRT exports are locked. A file that used some free minutes and paid for the rest can be exported. The PDF is made through your browser's print dialog, so allow pop-ups for the site and choose Save as PDF.

Try it on one of your own recordings

Your first 30 minutes are free when you sign up with Google or Facebook, no credit card needed. After that it is $0.75 per minute, and Stripe's checkout page shows the exact amount before you pay.

Juris Compiler AI is a workflow and productivity tool, not a law firm, and does not provide legal advice. Using the platform creates no attorney–client relationship. Automated transcripts and AI-generated analysis may contain errors and are not a substitute for professional judgment. You are solely responsible for reviewing and verifying all output and for any decisions or actions taken based on it.