Transcribe an interview

---
name: transcribe-an-interview
category: write
description: Transcribe an interview with accurate words, speakers, timecodes, uncertainty, and source provenance. Use when recorded speech must support editing, research, captions, quotation, or review.
---

# transcribe-an-interview

Create a searchable representation that never hides what the recording cannot establish.

## Procedure

1. Preserve the source and record file identifier, date, duration, language, channels, and checksum.
2. Choose verbatim or edited-verbatim scope and define notation.
3. Identify speakers only from reliable production evidence.
4. Transcribe words, meaningful pauses, crosstalk, non-speech sound, and language changes.
5. Insert regular timecodes and exact markers for quotes, uncertainty, or source switches.
6. Mark `[inaudible]`, `[unclear]`, and phonetic guesses without inventing text.
7. Verify names, terminology, numbers, and quotations against audio and authorized references.
8. Proofread while listening at normal speed and spot-check a second time.
9. Save versioned transcript and correction history with access controls.

## Guardrails

- Never fabricate inaudible speech from context.
- Do not identify an unknown speaker by guesswork.
- Protect sensitive interview content and machine-transcription uploads.

## Done

- A timecoded transcript document records source, speakers, words, sounds, and uncertainty
- Names, numbers, quotations, and sampled time ranges are verified
- Corrections and source provenance are recorded