How to Transcribe an Interview and Check Every Quote

To transcribe an interview, work from a recording you are allowed to process, create a speech-to-text draft and check it against the original. Keep the questions, identify who actually spoke and replay any passage you intend to quote. An interview transcript is a record of speech; a fluent draft or a speaker label is not proof that every word or attribution is correct. The embedded demo checks a short excerpt: at most the first five minutes, potentially less with a shorter file or lower remaining audio allowance. It does not complete a typical long interview; the full-recording route below explains how to finish the rest.

Go to the short-excerpt upload steps

Decide what kind of interview transcript you need

• Full speech record: agree which fillers, repetitions, pauses and meaningful sounds must be retained. A strict verbatim transcript needs a deliberate listening pass; this demo has no verbatim-mode switch.
• Readable working copy: use consistent conventions for any cleanup, preserving the speaker’s meaning and uncertainty. Keep this edited copy separate from wording you will quote directly.
• Summary or research findings: write these separately from the checked transcript. A paraphrase, theme or interpretation must not be presented as the interviewee’s exact words.

Fictional style comparison: verbatim — “Um, I, I think we can start in June, but only if the review is finished.” Cleaned reading copy — “I think we can start in June, but only if the review is finished.” The cleaned example removes the filler and repetition while retaining “I think” and the condition. Your project’s convention determines whether that cleanup is permitted. This is a working-copy choice, not a style switch in the demo.

Before uploading, check the participant agreement, your project’s access and storage rules, and whether an external processing service is permitted. Permission to record is separate from permission to upload or share. Review the linked privacy policy; do not infer an interview-specific retention period or a no-storage guarantee from a generic privacy label. Use an approved local or organizational workflow when the recording cannot be uploaded.

JotMe’s privacy policy

Keep the recording and its identifying notes together

• Keep the original file. Record the interview date, project or story, file name, approximate length, participants’ approved names or IDs, and the intended transcript style in your own working document.
• Prepare a short spelling list for names, organizations and specialist terms. Confirm spellings from an appropriate source; a familiar-looking name in an automatic transcript can still be wrong.
• First check that you have the intended, complete file and that both voices are audible. This initial file check is not a full review. To call a transcript fully reviewed, compare all covered speech with the audio, noting overlap or connection gaps. Then replay each intended quotation with its context. Checking a few quotes does not verify the entire transcript; do not reconstruct missing words.

Upload the interview recording and check the preview allowance

This embedded audio-to-text tool detects the spoken language automatically and returns a transcript in that language. It has no manual source-language override; the page’s interface language does not force an English transcript. Use a saved audio file you can process, with a network connection for upload and processing. A recorded video needs the separate video-to-text workflow.

Video-to-text tool

Common audio inputs include MP3, M4A and WAV, but the browser must also be able to read the duration. Keep the original if preparation fails; obtain a compatible copy through an approved recording or audio editor and check it before uploading. Renaming .m4a to .mp3 does not convert a file.

In Review before starting, compare Duration with Free time left (audio). The current demo shows at most the first five minutes per preview, and less when the file is shorter or your remaining allowance is lower. For example, a 30-minute interview with 02:00 available does not produce a complete 30-minute transcript. Check your displayed balance and the returned preview boundary.

1. Choose or drag in the interview audio. Wait for preparation and read any format or duration error.
2. Select Transcribe, then check the file, Duration and Free time left (audio) in Review before starting.
3. Select Proceed to transcribe when those details match your task. Read the resulting segments or error; a loading message is not a completed transcript.
4. Keep the original open in a separate audio player. Use the displayed segment times as references for your listening pass; speaker labels still need identification checks.
5. Where browser text selection is available, select a small passage, use your device’s Copy command and paste it into your own document. Inspect what pasted; segment text, labels and times may not all be selected together. Add the corresponding speaker label and time range separately. This embedded result has no dedicated Copy all, transcript-download or inline-edit button.
6. Save useful material before another file, Transcribe again or leaving. Label partial coverage. If browser selection is unavailable or unreliable, take source-referenced notes manually or use an accessible transcript workflow; copying across all segments on every phone or computer is not established here.

Check who spoke before checking what you want to quote

Replay a turn in the original player, including the question before it and the answer after it. A displayed time such as 00:52 is a reference within the uploaded recording; this result does not provide clickable audio playback. If an approved excerpt starts at 12:00 in the full interview, add its offset yourself: 00:52 in that excerpt is approximately 12:52 in the original, then verify by listening.

• Speaker: confirm the voice using a reliable introduction or your interview notes. Speaker 1 is not automatically the interviewer, and a short “yes” or interruption can be assigned to the wrong person. Leave identity unresolved when you cannot establish it.
• Words: check negation, qualifications and the complete sentence. “Not until June” is different from “in May”; removing an exception can change what the person meant.
• Details: listen again to names, dates, amounts, decimal points and units. Compare specialist spelling with an appropriate reference, then return to the recording rather than accepting a plausible replacement.
• Uncertainty: mark an unresolved word or overlapping passage with its time in your working copy, using a consistent notation such as [unclear 00:52]. Do not fill the gap from a guess.
• Context: retain the question and relevant surrounding answer when selecting a passage. Check the agreed use and review requirements before sharing it. An automatically polished sentence is not a verified direct quotation.

Use a small notation key for unresolved speech

• [unclear 00:52]: an unresolved word at this source time; do not guess it.
• [overlap 00:52–00:55]: voices overlap; record only wording and attribution you can establish.
• [unknown speaker 00:52]: identity is unresolved; do not substitute a person’s name.
• [reviewer note: points to chart]: an editorial addition, visibly separate from attributed speech. These are suggested project conventions, not universal transcription rules.

Check a quotation before using it

1. Confirm the speaker from the recording and reliable identification evidence.
2. Replay the whole passage with the preceding question and relevant surrounding answer.
3. Compare the exact words, including negations, uncertainty, names and numbers.
4. Preserve conditions and exceptions; do not turn a tentative answer into a promise.
5. Record the original source file and time range, including any excerpt offset. Checking what was said does not establish that the speaker’s underlying claim is factually true.

Worked example: keep the question with the answer

This author-written fictional example shows wording and speaker-attribution errors. It is not a real interview, captured JotMe output or an accuracy test.

• 00:43 — Interviewer: “Will the pilot launch in May?”
• 00:47 — Participant P03: “No, not until June, and only if the final review is complete.”
• An illustrative faulty draft says: “The pilot will launch in May.” That loses both the correction and the condition; it must not become a quotation.
• A checked working note would retain P03’s complete answer, its reference time, the preceding question and the unresolved real-world condition. It would not turn “only if” into a confirmed launch date.
• A separate fictional attribution error labels the 00:47 answer as “Interviewer.” The corrected example label is “Participant P03.” In a real recording, establish that from the voice, a reliable introduction and interview notes, including interruptions; the automatic label alone does not prove identity.

Format a working copy that someone else can check

Use a short header and one paragraph per speaking turn in your own document. Include the source file, interview date, approved participant names or IDs, transcript style, covered time range and who checked the text. Keep the unreviewed draft and checked version distinguishable. If your project requires pseudonyms, handle identifying notes through its approved access process; replacing a name alone does not make an interview anonymous. For example, keep P03_draft and P03_reviewed as separate files, with the review date and covered range in the header. Mark a five-minute excerpt as PARTIAL, not a completed interview. Use IDs in filenames and keep identifying information under the project’s access rules.

Copy this suggested working-document template

This fictional template is for your own document, not a JotMe export. Replace its placeholders, then review the actual covered audio.

• Source filename: interview_P03.m4a | Interview date: [date] | Participants: I01 (interviewer), P03 (participant).
• Transcript style: [project convention] | Covered range: 00:00–05:00 | Coverage: PARTIAL; full recording [duration].
• Reviewer: [ID] | Review date: [date] | Status: DRAFT / REVIEWED for the stated range only.
• 00:43 — I01: “Will the pilot launch in May?”
• 00:47 — P03: “No, not until June, and only if the final review is complete.”
• 00:52 — [unknown speaker]: [unclear 00:52]. Keep this unresolved marker separate from the words you can establish.

W3C’s transcript guidance emphasizes speaker identification and information needed to understand the recording. Add meaningful sounds or essential visual context when they matter, clearly labeling your additions separately from spoken words. A speech transcript of a filmed interview does not automatically describe a silent gesture, chart or on-screen name. Timestamps help when they make the source easier to find; they are references, not evidence of correctness by themselves.

W3C transcript guidance

If a checked original-language transcript already exists, use the separate text-translation tool on that corrected text, choosing the actual source and desired target languages. Retain linked original and translated versions with matching source times; identify translated wording as a translation when quoting it. If you start from audio instead, audio translation creates another recognized-source draft that needs its own listening and translation review. Translating a misheard word does not fix it.

Translate a checked transcript as text

Translate directly from an audio recording

Fix Common Interview Transcription Problems

• Noisy or incomplete recording: replay with headphones, obtain a clearer approved source if one exists, or mark the passage unresolved. An automatic transcript cannot establish words absent from the recording.
• Overlapping voices: check each turn manually and keep an unknown-speaker label when necessary. Do not assign a sensitive quotation solely from an automatic label.
• Wrong detected language: compare the actual speech and obtain a clearer permitted recording or an approved transcript. This embedded transcription tool has no manual language selector.
• Long interview or exhausted preview: save the partial text and its ending. This demo has no verified automatic continuation into an account upload. A separate account file-transcription service uses a new upload/job with its own limits and terms; signing up does not resume this demo file or prove where another job starts. To finish manually, open the original in a player, resume from the last verified time, transcribe short speaking turns into your document, and compare all remaining speech with the recording. Keep overlap notes at the join and finish with a full covered-range review. A separate full-recording service is another option only after checking its support and terms.
• Another upload or tab: check the current balance before processing. Refreshing, splitting files, changing attachments or opening another tab is not a way to reset the allowance or obtain unlimited interview transcription.

For recording a new phone memo, use the separate voice-memo guide linked below. This article starts with an existing interview file and focuses on speaker attribution, source context and checked quotations.

Voice-memo guide