Deposition transcription at $0.05/min (for internal review, not court record)

May 5, 2026 · 5 min read

First: this is not a replacement for a certified court reporter. If you need a transcript admissible in court, signed by a stenographer or certified transcriber, that's regulated work and Whisper is not it.

This article is about everything else a small law firm does with deposition audio: internal review, witness prep, discovery first-pass, exhibit cross-reference, paralegal summary work. The $4-8/page court reporter charges add up fast — and most of that paper sits in a drawer for cases that never go to trial. Here's how to drop that cost for the non-record use cases.

What we mean by "internal review"

• Pre-trial witness prep: partner reads the transcript before prepping the witness, marks favorable/unfavorable lines
• Discovery review first-pass: paralegal scans transcript for keywords, flags 5-10% of pages for attorney review
• Exhibit cross-reference: match deposition lines to specific exhibits
• Internal case file: searchable text in your case management system
• Settlement analysis: partner reviews opposing witness deposition before settlement conference

None of these are read by a judge. None require a stenographer's seal. They're first-pass review work.

Cost comparison for a 6-hour deposition

$18 vs $800-2,400 for the same data. The court reporter transcript is still needed when you go to court — you order it then. For everything before, why pay 50x?

The 5-step workflow for legal teams

1. Get the audio

Record on a Zoom deposition (most depositions are remote in 2026). Local recording → audio_only.m4a. Or take the audio file from the videographer if it's an in-person deposition.

2. (Optional) compress for upload

ffmpeg -i depo_2026-05-04.m4a -b:a 64k -ac 1 depo.mp3

6 hours of audio at 64 kbps mono = ~165 MB. Within LessRec's 1 GB limit.

3. Upload to LessRec

Use a firm account ($25 pack = 500 minutes = ~$0.05/min). The 10 free minutes is enough to test accuracy on a sample first.

4. Run keyword search on the .docx

Open the .docx output, hit Cmd-F. Search for the issue terms ("breach", "warranty", "notification", "signed"). Mark the pages where they appear. This is your first-pass discovery review.

Whisper's accuracy on clean deposition audio (single mic per speaker, professional recording) is around 95-97%. For aggressive cross-examination with overlapping speech, drop to 88-92%. Either way, the search-and-flag workflow finds 90%+ of the moments you care about.

5. Use the transcript for paralegal summary work

Junior paralegal reads the .docx, summarizes per topic, flags 30-60 pages of attorney-review priority. With a clean machine-generated transcript, this is 4-6 hours of work instead of 2-3 days of listening to audio.

What about diarization (witness vs counsel)?

LessRec doesn't auto-label speakers (yet — Q3 2026). For a deposition, the workflow is: skim the first 10 minutes, identify each voice, then add speaker headers manually as you find them. About 15 minutes of cleanup work for a 6-hour deposition.

Or: AssemblyAI (more expensive, ~$0.10/min) does diarization automatically if labels matter to your workflow. Trade-off.

Privacy and confidentiality

Deposition audio is privileged. Where it lives matters.

• LessRec stores audio + transcripts on Hetzner Helsinki (EU GDPR jurisdiction). 7-day auto-delete by default.
• We don't train on your data, ever. Don't sell, don't share, don't index.
• For tighter compliance (HIPAA-adjacent matters, attorney-client work product) — email hello@lessrec.com for a private-tier with 24-hour retention and encrypted-at-rest.
• Self-host option: faster-whisper INT8 runs on a $40/mo Hetzner box. Take the open-source model, run it on hardware you control, no third party touches the audio. We can help you set it up.

When the certified court reporter is non-negotiable

• Trial transcript needed for the record
• Cross-jurisdiction case requiring stenographer certification
• Insurance defense work requiring claims-quality records
• Anything entered into evidence

For all of these, you order the certified transcript when you need it. Whisper is for internal use beforehand and around it.

For court reporting firms specifically

Several court reporting firms use Whisper as a backbone for their own internal review tier. Workflow: stenographer captures live record (the certified product), AI generates the rough draft for the team to review and clean. Cuts the proofreading and clean-up time per page significantly.

If you run a court reporting firm and want to integrate Whisper transcription as your internal review tier, we have a white-label option starting at $2k/mo flat for unlimited firm use. Email hello@lessrec.com.

FAQ

Can I use this transcript in court?

No. Whisper output is not certified. Use it for internal prep only. When you need a record, order the certified transcript from a court reporter.

Will Whisper get the legal terms right?

Common terms (motion, objection, hearsay) — yes. Industry-specific or unusual terms — sometimes wrong. Search the transcript for known problem words ("voire dire" sometimes comes out "voir dyer") and fix in batch.

What if the audio has whispers or muffled witness moments?

Whisper's accuracy drops on muffled audio. The transcript will mark it as "[unintelligible]" or just transcribe a wrong word. Listen to the audio at those points yourself.

Are you HIPAA compliant?

Default LessRec tier is not HIPAA. The private-tier option (mentioned above) signs a BAA. Email for details.