How to transcribe an interview quickly and accurately

By the BF-Blessy team · Updated November 7, 2026 · 2 min read

An hour of interview can take four hours to type by hand. With a transcription tool the machine writes the first draft in minutes, but the result is only as good as your method. This guide explains a routine used by journalists, researchers and students that gives you accurate text without doing the whole job twice.

Before you start: get the file ready

Unuk Transcription accepts common audio and video files such as MP3, WAV, M4A, OGG, WEBM and MP4. A single file can be up to about 30 minutes, so cut a long interview into parts. Name the parts clearly (interview-1, interview-2) so you can put the text back together in order.

Choose the spoken language first

Tell the tool which language is spoken. A correct choice matters more than anything else for accuracy, because the tool listens for the sounds of that language. The tool supports many world languages and several African languages, including Swahili, Kinyarwanda, Lingala, Hausa, Amharic, Yoruba, Zulu, Somali and Shona. If you chose the wrong language, change it and transcribe again.

Let the tool cut the audio into short segments

The audio is divided into short segments of a few seconds. This is useful because you can listen to one small piece, read its text and correct it, instead of scrubbing through an hour of audio. If the cuts fall in awkward places, you can re-cut the segments at natural pauses before transcribing.

Transcribe, then review segment by segment

Press Transcribe and wait for the draft. Then review it in order. For each segment, play the audio and read the text. Correct names, numbers and technical words, which are the most common errors.

• Names of people and places: confirm the spelling with the speaker or a document.
• Numbers and dates: listen twice.
• Words from another language: check that they are written correctly.

Handle overlaps and unclear parts

When two people talk at the same time, or the sound is poor, mark the passage with a clear note such as [unclear] and the time. Do not invent words. You can split a segment that holds two speakers, or merge short pieces that belong together.

Add speaker labels

For interviews, note who is speaking at each change: Interviewer and Guest, or the real names if the speaker agreed. Consistent labels make the transcript much easier to read and quote.

Export in the right format

Choose the format that fits your next step: a Word document for editing and quoting, plain text for notes, subtitles files (SRT or VTT) if you will place the text on a video, CSV or JSON if you will process the data. See our guide on export formats for a clear comparison.

Respect the people you recorded

Ask permission before recording, explain how the interview will be used, and keep the audio and transcripts safe. Delete files you no longer need.