VeriVox / Blog
Forensic Audio
Intelligence
Guides, research, and insights on court-defensible transcription, AI admissibility, chain of custody, and voice verification.
2026-07-02 · 7 min read
Is an AI Transcript Admissible in Court? What Judges Actually Require
Authentication, human verification, and provenance — what makes an AI transcript survive cross-examination, and what doesn't.
2026-07-02 · 9 min read
Body Cam & Dashcam Audio: Usable Transcripts from the Worst Recordings
Wind, sirens, radio chatter, engine noise, overlapping commands. Multi-pass scatter on chaotic audio is signal, not failure.
2026-07-02 · 7 min read
Chain of Custody for Audio Evidence: A Practical Guide
Hash it, log it, preserve it. A chain-of-custody checklist for audio evidence that survives scrutiny.
2026-07-02 · 7 min read
Enhancing a Faint Recording Without Getting It Thrown Out
Named recipes. Reproducible commands. Documentation that survives cross-examination. Undocumented cleanup is a gift to opposing counsel.
2026-07-02 · 7 min read
The [Inaudible] Problem: What Transcripts Hide
When five AI passes disagree, that scatter is data. The ear-queue, agreement scores, and why disagreement belongs in the record.
2026-07-02 · 7 min read
How to Transcribe a Jail Call for a Criminal Case
Jail calls are privileged. Uploading them to a cloud service breaks that privilege. Local-first transcription preserves it.
2026-07-02 · 9 min read
Transcribing Your Own Recording for a Legal Matter — Free, Local, Private
A parent with a voicemail. A tenant with a threat. Personal use is free forever — you shouldn't have to pay to prove the truth.
2026-07-02 · 7 min read
Preparing Audio Transcripts for Canadian Courts
Certification, authentication, and admissibility in Canadian proceedings. Who can prepare a transcript, and what the court requires.
2026-07-02 · 8 min read
Voice Identification for Lawyers: What a Similarity Score Can (and Can't) Prove
Enroll a voice, score the segments, flag conflicts. Speaker similarity is corroboration — not machine identification.
2026-07-02 · 7 min read
Whisper Hallucinations: Why One AI Pass Should Never Become Evidence
Cornell found 1.4% hallucination rates. Michigan found errors in 8/10 samples. On faint audio, a single pass invents. Multi-pass voting catches it.