# Audio Transcription: Method, Styles and Interview Examples > Audio transcription explained: choose a transcript style, record clean audio, transcribe audio to text in five steps, and read interview examples. _Source: https://professionalstoolkit.com/articles/audio-transcription — The Professional's Toolkit · updated November 2026_ --- Audio transcription is turning recorded speech into written text: an interview, a lecture, a sales call or a team meeting becomes a document you can search, quote and share. The method is short. Record clean audio, pick a transcript style before you start, let a person or software produce the first draft, then check that draft once against the recording. That last review pass is what separates a usable transcript from a page of misheard words. **4-7 hours** — to transcribe one hour of audio by hand, per the University of Bath research data guide - **8 hours** — of work per interview hour for a clean, final transcript, review included (Oregon State University Libraries) - **3 styles** — summary, clean verbatim, full verbatim - **5 minutes** — the timestamp interval Oregon State suggests ## Audio transcription styles: pick one before you start Every transcript answers one question: what will you do with it? A researcher coding interviews needs different text than a manager who only wants the decisions from Tuesday's call. The University of Bath research data guide separates a summary transcript, which keeps key points and a few exact quotes, from an exact transcript that keeps fillers, cut-off words, laughter and timed pauses [1]. Most people land in between, on what transcribers call clean verbatim. | Style | What it keeps | Use it for | |---|---|---| | Summary | Key points, topics, a few exact quotes | Meetings, quick reference, deciding what deserves a full transcript | | Clean verbatim | Every sentence, minus ums, false starts and repeats | Interviews, articles, reports, most research | | Full verbatim | Everything, including fillers, pauses, laughter and noises | Conversation analysis, linguistic or evidential work | Choose once and keep it for the whole project. Mixing styles makes ten interviews hard to compare, and switching halfway means redoing the early ones. If you are writing a thesis, your methods guide may define these terms differently; follow its wording. One boundary is worth stating. Certified transcription for courts or medical records is its own trade, with its own rules. This guide is for people who need usable text from their own recordings. ## Start with the recording, not the transcript Bad audio is the most common reason a transcript fails. No software repairs it afterward. Put the microphone close to whoever speaks, or in the middle of the table when several people talk. Turn off fans, close the door and silence phones. Ask people not to talk over each other. Then make a short test recording and listen back before the real session begins, while you can still move the device. For a remote call, record from the meeting platform itself rather than holding a phone next to the laptop speaker. Each person's own microphone matters too: a basic headset close to the mouth beats a laptop microphone across the room. Ask permission before you press record, and say out loud that the session is being recorded. It takes ten seconds. It also gives you a clean marker at the start of the file, which makes the first minute easy to find later. > 💡 **Say names in the first minute:** Ask each person to say their name at the start of the recording. Matching voices to speaker labels later becomes a two-second job instead of guesswork. ## How to transcribe audio to text in five steps The steps are the same whether you type every word yourself or let speech recognition write the draft. Only step 2 changes. 1. **Label the file.** Name it with the date, the place and who spoke, and note the total length. A folder of files called Recording 14 is where projects stall. 2. **Produce a first draft.** By hand, slow the playback and keep your fingers on the keyboard. With software, upload the file or record straight into a transcription app, then wait for the draft. 3. **Mark the speakers.** Oregon State University Libraries suggests a header naming the narrator, interviewer, date, location and recording length, a new paragraph with initials each time someone speaks, and a timestamp every five minutes [2]. 4. **Review against the audio.** Re-read the whole draft while the recording plays and correct what you hear. Flag anything you cannot make out with a tag such as [inaudible 12:40] instead of guessing. 5. **Clean up and store.** Remove or replace names if the transcript will travel outside your team, then save it in a format you can search later. The time difference in step 2 is large. Bath warns that typing it yourself can take 4 to 7 hours for one hour of audio [1]. Software turns that into a wait of minutes plus a review, and the review never goes away. Accuracy depends on the audio, the accents and the vocabulary. Vendor accuracy figures are the vendors' own claims, not independent tests. If you want the free route in detail, see [how to transcribe audio to text with free tools](/articles/how-to-transcribe-audio-to-text). > 💡 **Slow the playback, not your typing:** When you transcribe by hand, slow the audio to a pace you can type at and control playback from the keyboard. Stopping to reach for the mouse costs more time than typing does. ## Interview transcription examples: one answer, three ways Here is a labeled illustration, invented for this guide. The interviewer is I and the respondent is R. The same 15 seconds of an interview about a product launch, written in the three styles from the table above. **Full verbatim** > I: So how did the, um, the launch go? > R: It was... honestly it was, like, rough at first. [laughs] We lost, we lost two days on the, the payment page. **Clean verbatim** > I: How did the launch go? > R: Honestly, it was rough at first. We lost two days on the payment page. **Summary** > Launch: rough start, two days lost on the payment page. The full version tells an analyst that the respondent hesitated and laughed. The clean version is what you quote in a report. The summary is what goes into the notes for a team that only needs the fact. A good interview transcript also opens with a header, for example: Interview 04, 12 March, video call, I = interviewer, R = respondent, 47 minutes. That one line tells a reader months later what they are looking at. Speaker tags work best when they are short and consistent. Initials or one-letter roles are easier to scan and easier to search than full names repeated on every line. Keep the same tag for the same person across every file in a project. ## Common mistakes that ruin a transcript Most bad transcripts fail for the same few reasons, and none of them is the software. - **Trusting the first draft.** Automatic drafts read smoothly even where they are wrong. A misheard number looks exactly like a correct one. - **No speaker names.** A transcript where every line belongs to Speaker 1 or Speaker 2 is hard to quote and impossible to audit. - **No timestamps.** Without them, checking one sentence means replaying the whole hour. - **Waiting weeks.** Transcribe while you still remember the context, the names and the jargon. Memory fills gaps that audio cannot. - **Sharing raw names.** Transcripts travel. Decide before sending one outside your team whether people's names, clients or figures should stay in it. > ⚠ **Names and numbers fail first:** Speech recognition mishears proper names, figures and jargon while the sentence around them reads fine. Check every name and number against the audio before you quote it. One more habit saves hours. Keep the recording next to the transcript, under the same name. When someone questions a quote six months from now, the audio is the only proof of what was said, and a transcript without its source is just a claim. ## When doing it by hand stops working The manual method holds for a few interviews. It breaks with volume. Take a labeled illustration with round numbers: a researcher with 10 one-hour interviews faces 40 to 70 hours of typing at Bath's estimate, before any analysis starts. A team that records every call hits the same wall faster, because recordings pile up quicker than anyone can process them. **Once the backlog grows faster than you clear it**, the answer is software that drafts the text while you keep the review. If your file is a video rather than audio, the workflow changes a little, and our guide on [how to transcribe a video](/articles/how-to-transcribe-a-video) covers it. For recurring calls, the better move is a tool that listens to the meeting and writes the transcript as it happens; [meeting transcription](/articles/meeting-transcription) explains the ways to set that up. ### Tools that do it for you For files you already have, three tools in our catalog fit the job best. Otter.ai [imports audio and video files and returns a transcript](/articles/otter-ai-transcription) with a summary and an outline, M4A files from Apple Voice Memos included. Fireflies.ai [transcribes uploaded MP3, M4A, WAV, MP4 and WEBM files](/articles/fireflies-ai-transcription) on its Free plan as well as on paid ones. Krisp imports one audio or video file at a time from its Core plan. ### Otter.ai — solves this Import an audio or video file, including M4A from Apple Voice Memos, and Otter.ai returns a transcript with speaker labels, a summary, action items and an outline. It also transcribes live recordings in its web, desktop and mobile apps. - Where it stops: On Basic only the first 30 minutes of each transcript can be viewed, and file imports are a small allowance for the account's lifetime. - Plan: Imports on Basic (free); pdf, docx and srt exports from Pro - How Otter.ai does it: [How Otter.ai transcribes files](https://professionalstoolkit.com/articles/otter-ai-transcription) - More: [Otter.ai](https://professionalstoolkit.com/tools/otter-ai) ### Fireflies.ai — solves this Upload MP3, M4A, WAV, MP4 or WEBM files, or pull them from Dropbox, Box, Google Drive, OneDrive or Zoom cloud recordings. Uploads get generic speaker labels you can rename, and find-and-replace fixes repeated mistakes. - Where it stops: Free has no downloads of transcripts or audio, and recaps are produced only for files longer than 3 minutes. - Plan: Uploads on Free; downloads from Pro - How Fireflies.ai does it: [Fireflies.ai uploads and downloads](https://professionalstoolkit.com/articles/fireflies-ai-transcription) - More: [Fireflies.ai](https://professionalstoolkit.com/tools/fireflies-ai) ### Krisp — solves this Import an audio or video file on desktop, web or mobile and Krisp returns a transcript with speaker labels and timestamps, exported as .txt. Live meetings are transcribed by the desktop app without a bot. - Where it stops: Uploads go one file at a time, with no batch; no free plan, only a 7-day trial. - Plan: Core - How Krisp does it: [Krisp note taker and uploads](https://professionalstoolkit.com/articles/krisp-ai-note-taker) - More: [Krisp](https://professionalstoolkit.com/tools/krisp) ## FAQ **Can I transcribe audio to text for free?** Yes. Some note-taking apps accept uploaded files on a free plan, Fireflies.ai Free among them, and office suites include free dictation tools. The trade is your time: free routes usually leave more cleanup and fewer export options. **How accurate is automatic audio transcription?** It depends on the recording: microphone distance, background noise, overlapping voices, accents and vocabulary. Accuracy percentages published by vendors are their own claims, not independent tests, so review every transcript you plan to quote. **What is the difference between transcription and dictation?** Dictation turns your speech into text live, as you talk into a document. Transcription converts a recording that already exists, often with several speakers, into a written record. ## The bottom line Audio transcription is a short method done carefully. Decide the style before the first file, record close to the speaker in a quiet room, get a draft by hand or by software, and always review it against the audio with speakers and timestamps marked. Typing it yourself works for a handful of interviews; past that, the hours add up faster than the recordings get used. Let software write the draft, keep the review for yourself, and store each transcript next to the recording it came from. ## Sources [1] University of Bath Library — https://library.bath.ac.uk/research-data/working-with-data/transcription (Nov 2026) [2] Oregon State University Libraries — https://guides.library.oregonstate.edu/oralhistory/describe_transcribe (Nov 2026)