# Transcribing Your Own Recording for a Legal Matter — Free, Local, Private

You recorded something that matters. A voicemail. A threat. An admission. A moment of clarity in a chaotic situation. You know what you heard — but the court needs a transcript.

You are not a law firm. You are not billing a client. You are a person with evidence and a need to make it legible.

Personal use is free forever. That is not marketing. That is the mission.

Why VeriVox exists

VeriVox was built by someone in exactly your position: a parent, alone with a recording that mattered, facing a courtroom with no budget for a forensic expert and no intention of handing a machine's hallucination to a judge.

The tools exist — speech recognition, voice identification, enhancement models. They were not built for this. They were built for enterprise transcription, for podcasters, for meetings.

We built the tool we needed. Then we made it free for everyone who needs it the same way.

What you are facing

You have a recording. It is not clean. It is a voicemail from a moving car. A jail call with compression artifacts. A threat captured through a floor. Wind. Background noise. Overlapping voices.

You tried a free transcription service. It produced:


[00:14] Thank you for calling.
[00:21] [inaudible]
[00:28] Have a great day.

The words that matter — the admission, the threat, the evidence — are either wrong or bracketed.

You tried uploading it to a paid service. They want $150 for a 10-minute recording. Or they ask you to create an account, accept terms that let them use your audio to train their models, and hope their single-pass AI does not hallucinate the critical line.

You should not have to choose between affording the truth and protecting it.

What you need

You need a transcript that:

  • Is accurate — verified by a human ear, not just a machine
  • Is defensible — documented method, preserved uncertainty
  • Respects privilege — the recording never leaves your control
  • Costs nothing — because justice should not have a paywall

That is what VeriVox does.

Understanding what courts require for AI transcript admissibility is the foundation — but having the right tool to meet those requirements is what makes it possible for a non-expert.

How it works

VeriVox runs entirely on your computer. You do not upload the recording. You do not create an account. You do not hand your evidence to a third party.

Five stages, one export:

1. Ingest — The recording is hashed (SHA-256). A custody manifest is written. Before anything touches the audio, there is proof of what it was.

2. Enhance — Multiple noise-reduction recipes are applied. Each is scored. The audio is not "cleaned up" — it is processed in documented, reproducible ways. Nothing is added.

3. Transcribe — Five transcription passes run in parallel. Different models, different approaches. One opinion is a guess. Five opinions vote.

4. Consensus — The passes are aligned word-by-word. Where they agree, confidence is high. Where they scatter, the disagreement is flagged — those lines need your ear.

5. Verify — You listen to the flagged sections. You correct hallucinations. You mark truly inaudible lines. The export labels every line: verified by you, corrected by you, or machine-only.

The result is a transcript you can defend.

If you need to establish chain of custody for your recording, VeriVox handles that from the moment you load the file.

What the export looks like

Not a wall of text. A ledger.


TRANSCRIPT — evidence.m4a
SHA-256: 7f3a91cc482b3f7...8d4be081
Operator: [Your name]
Date: 2025-11-04

[00:14.2–00:17.8] ✓ CALLER: You were never supposed to be there.
  agree 5/5 · verified by ear

[00:22.1–00:25.3] ✎ CALLER: I already told you what happened.
  agree 3/5 · corrected
  (ASR original: "I already to you what happen")

[01:03.4–01:05.1] ⚑ CALLER: [grave excerpt — verified]
  agree 0/5 — machine could not resolve

[02:48.0–02:49.5] · BACKGROUND: [traffic noise, exclude from excerpts]
  machine-only

Every line declares how it earned its place. When you hand this to a court, opposing counsel can question the method — and you can answer every question.

Why local-first matters

Your recording is evidence. It may also be privileged. Uploading it to a cloud service — even a reputable one — means:

  • A third party has access to it
  • It may be retained for "quality assurance" or "model training"
  • It is stored on servers you do not control
  • Subpoenas can reach it

If your recording involves:

  • A family matter (custody, abuse, threats)
  • A criminal defense (jail calls, witness statements)
  • An employment dispute (harassment, retaliation)
  • Anything you would not want stored on someone else's server

Then it should not leave your machine.

VeriVox processes locally. The models run on your laptop. The audio never uploads. It can run fully offline, air-gapped if you need it.

Zero retention, because there is nothing to retain.

Who this is for

VeriVox was built for:

  • Parents — custody cases, evidence of abuse, threats
  • Tenants — landlord harassment, illegal evictions, documented violations
  • Employees — workplace harassment, retaliation, wrongful termination
  • Victims — threats, stalking, evidence of violence
  • Witnesses — recordings that corroborate testimony
  • Pro se litigants — self-represented people who cannot afford expert services

If you are handling your own case and have evidence audio — this is for you.

What it costs

Personal use is free. Forever.

No trial period. No feature gates. No upsell. The full pipeline — enhancement, multi-pass transcription, consensus voting, voice verification, human-certified exports — is yours.

We do not monetize personal users. We monetize the firms that bill with it.

Law firms, court reporters, forensic experts, and agencies pay a commercial license. That revenue funds the free tier. You get the same tool they do.

What you do not get for free

Support. We do not staff a help desk for free users. The tool is documented. If it breaks, file an issue on GitHub.

Hosting. You run it on your machine. We do not host your audio, run your processing, or store your files. That is by design.

Legal advice. We cannot tell you how to present the transcript in your case. We can tell you the transcript is defensible — what you do with it is between you and the court.

What you do get

The same pipeline the firms use:

  • Multi-pass consensus transcription
  • Chain-of-custody manifest
  • Reproducible enhancement recipes
  • Enrolled voice verification
  • Human-verified exports with full provenance

And the knowledge that when you hand that transcript to a courtroom, you can say:

  • Here is the original recording (hash-verified, unaltered)
  • Here is the method (documented, reproducible)
  • Here is what the machines could not agree on (preserved in the record)
  • Here is what I verified by ear (labeled, certified)

That is what makes it defensible.

When to pay for help

If your case involves:

  • High stakes (criminal charges, large financial exposure)
  • Expert cross-examination expected
  • Complex audio (many speakers, long duration, severe degradation)
  • Court-ordered certification requirements

You may need a forensic audio expert or certified court reporter. VeriVox produces a verified transcript — but it cannot testify.

For lower-stakes matters, self-represented cases, and internal preparation — the free tier is enough.

How to get it

VeriVox is in early access. Request access at verivoxai.com.

You will receive:

  • The desktop application (macOS, Windows, Linux)
  • The local transcription models (Whisper, Wav2Vec2, others)
  • The enhancement pipeline
  • The auditor workbench (browser-based verification interface)

Everything runs locally. Nothing phones home.

Why we made it free

Because someone in the middle of the hardest moment of their life should not have to choose between proving the truth and paying rent.

Because access to justice should not depend on access to capital.

Because the tools exist, the science is public, and the only thing standing between you and a defensible transcript is someone willing to build it and give it away.

We built it. We are giving it away.

The mission

Free for the people it was built for.

That is the line. That is the promise. Personal use is free forever — not as a trial, not as a loss leader, but as the reason VeriVox exists.

The firms fund it. You use it. The evidence speaks.

Request early access →

For more on how VeriVox's open-source approach ensures you're not locked into a proprietary system, read about our licensing and mission →

FAQ

Is there a catch?

No. Personal use is free. If you are not billing a client, you do not pay.

What counts as "personal use"?

You are preparing a transcript for your own legal matter. You are not a law firm, court reporter, investigator, or expert billing for the work. You are the person with the recording — or you are helping someone pro bono.

What if I need help using the tool?

The tool is documented. Community support is available (GitHub, forums). We do not staff a help desk for free users, but the user base helps each other.

Can I use this for my small business?

If you are transcribing for your own matter (employment dispute, contract case), yes — that is personal use. If you are transcribing for clients or billing for it, that is commercial use and requires a license.

How do you make money?

Law firms, court reporters, forensic experts, and agencies pay for commercial licenses. That revenue funds the free tier, the open-source core, and continued development.

Will you ever charge for personal use?

No. That would break the mission. The promise is "free forever" — we mean it.

Can I donate to support the project?

Not yet. When we open donations, we will announce it. For now, the commercial licenses fund everything.

What if the audio quality is really bad?

Multi-pass consensus is designed for difficult audio. Where one model fails, others may succeed. Where they all disagree, that disagreement tells you which lines need your ear — and the export labels them honestly.

What if my case involves something sensitive?

The audio never leaves your machine. We do not see it, store it, or transmit it. The models run locally. It is as private as your own files.

Can I share the transcript with my lawyer?

Yes. The transcript is a text file with embedded metadata. Share it however you share documents. The original audio stays on your machine unless you choose to transfer it.

What if the court requires certification?

Some courts require certified transcripts for official proceedings (trials, appeals). VeriVox produces a verified transcript with full provenance — but it is not "certified" in the credentialing sense. Check your court's rules. For many matters (motions, tribunals, self-represented cases), verified is enough.