How to Transcribe an Interview: A Practical Step-by-Step Guide

Whether it is for an article, a research project or a hiring process, a written transcript makes an interview searchable, quotable and easy to share. Typing it by hand takes roughly four to six times the length of the recording. With automatic speech recognition, most of that work disappears — you mainly review and tidy.

1. Record so the transcript comes out clean

Transcription quality is decided mostly before you press record:

• Get the microphone close. A phone on the table between two people works; a lapel mic is better.
• Pick a quiet room. Background music and cafés are the biggest cause of errors.
• Avoid talking over each other. Overlapping speech is hard for any recogniser.
• Remote interviews: record each side locally if your tool allows it, or use the call app’s recording.
• Ask for consent before recording, and tell the interviewee how the recording will be used.

2. Transcribe the recording

1. Open TidyTranscribe on a computer (Chrome or Edge is fastest for long files).
2. Set the spoken language and choose the Accurate model for interviews — names and natural speech benefit from the larger model.
3. Drop in the recording: MP3, M4A, WAV and video files all work.
4. Watch the progress bar. Long files are processed in segments, so you can start reading early.

Because everything runs in your browser, the recording is never uploaded. That matters for interviews, which often contain personal details or unpublished information.

3. Download both TXT and SRT

The TXT file is your working transcript. The SRT file contains the same text with timestamps, which is useful when you need to find the exact moment a quote was said and listen to it again.

4. Add speaker names

The tool does not label speakers automatically, so add them during your review. A simple format that works well:

INTERVIEWER: How did the project start?
ANA: It began as a side project in 2021…

Tip: do a quick pass first where you only insert a line break and a name every time the speaker changes. Then do a second pass for wording.

5. Review and clean up

• Names, places and jargon are the most common errors. Use find-and-replace for repeated ones.
• Decide on verbatim vs. clean. Verbatim keeps “um”, false starts and repetitions; clean read removes them. Journalists usually quote clean but keep meaning exact.
• Check every quote you publish against the audio, using the SRT timestamps to jump to it.
• Mark unclear parts as [inaudible 12:34] instead of guessing.

How long does it take?

It depends heavily on your device: a recent computer with GPU acceleration is much faster than a phone or a CPU-only machine, and the Accurate model is slower than Balanced. Try a few minutes of audio first to get a feel for the speed before starting a long recording. Reviewing the transcript usually takes about as long as the interview itself — still far quicker than typing from scratch.