Guides

How to transcribe an interview without wasting hours

A practical interview transcription workflow: record clean audio, use AI for the first pass, verify quotes and export structured notes.

Start with audio quality, not the tool

No transcription model can recover words that were never captured. Put the microphone closer to the speaker, avoid a shared table, and record room tone for a few seconds. For phone interviews, ask the participant to use headphones so their audio does not loop back into the call.

Use AI for the first pass

Automatic transcription is excellent at producing a rough draft quickly. It is not a publishing tool. Use it to get the structure, the timestamps and the searchable text, then verify only the quotes and facts you plan to use.

Work in short blocks

Long files fail in more ways than short ones: upload limits, memory, network interruptions and cost. Recording or splitting in short blocks gives you progress feedback and lets you recover from a failed segment without starting over. The synchronous AI limit is currently two minutes or five megabytes per request.

Ask for the output you actually need

A generic summary is rarely useful for an interview. Ask for key quotes, a timeline, themes or action items. Keep the transcript as the source of truth and treat the summary as an index into it.

Verify before you publish

Read the quote against the timestamp in the original audio. Check names, numbers, titles and anything that could change the meaning. Automated systems can be confidently wrong, especially with accents, crosstalk and unusual names.

Store less, export more

Keep the original recording in your own storage, export the transcript you need, and delete what you no longer need. If you work with vulnerable sources, check your organization's data policy before using any cloud transcription service.