Interview transcription, ready for review

The review desk

Review questions, answers, and speakers

A useful interview transcript keeps each answer connected to its question. Speaker detection groups voices. It does not know anyone's name or verify their identity. Listen first, then replace a label with a name or a neutral role such as "Interviewer."

See an example: a two-person interview

This clip is synthetic owned practice input. It is not a transcription-provider output and not a customer recording. Download it and upload it above if you want to try the real tool. Playing or downloading the sample does not start a transcription request.

The authored turns below are the reference script, with the start time of each beat in the recording. The last beat mixes both voices on purpose so you can hear overlap. Compare returned words, speaker turns, and timestamps with this recording. Confirm the proper name North Quay.

1. 00:00Interviewer
   What changed during the North Quay pilot?
2. 00:02Guest
   We moved the Tuesday interviews to the library. Attendance stayed the same.
3. 00:06Interviewer
   So the location changed, not the number of participants?
4. 00:10Guest
   Exactly. We have not compared the feedback yet.
5. 00:14Interviewer / GuestOverlap
   InterviewerSorry, go on.

   GuestWe still need the library notes.

How interview transcription works

1. Upload your recording

Choose a supported file or record in the browser. Speaker detection starts enabled with a two-speaker hint. Open settings to change the count or turn detection off. Select the spoken language and add important names before submitting. If login interrupts a draft, follow the file-restoration prompt. A filename alone cannot restore an audio file.

2. Review the result

Wait for the task on this page. Open the completed transcript, search for a phrase, and replay the relevant passage. Rename speaker labels in the speaker panel and wait for the saved confirmation. If a task fails, read its error before starting a new attempt. Refreshing a processing task restores its status.

3. Export your text

Save your corrections, then choose DOCX for further editing or timestamped text for a reference copy. DOCX keeps your reviewed prose and, when speaker segments are available, adds a clearly labeled original timestamped transcript with your saved speaker labels. An intentionally blank reviewed transcript stays blank. Your task stays in My Recordings alongside work created in the general workspace. Reopening it or exporting the saved transcript does not start a second transcription.

Check the quote, not just the spelling

Search for Attendance in the practice interview, replay the answer and the following question, and check the word same. Replacing it with increased would reverse the meaning. Confirm North Quay against your own interview notes. Names and unfamiliar terms can be misheard.

Replay the whole exchange before extracting a sentence. Check speaker attribution, qualifications, and any interruption. Keep unclear words marked for review instead of guessing. The National Park Service's interview-transcribing guidance explains why written transcripts cannot preserve every nuance of the original recording.

Use this workflow for reporting interviews, customer conversations, and oral-history drafts. It supports transcript review. It is not a certified legal transcription service or a research coding system.

Supported files and plan limits

This page accepts MP3, WAV, M4A, MP4, MOV, FLAC, OGG, WEBM, AAC files, up to 1 GiB and 60 minutes each. Browser recording requires microphone permission and browser support. Actual audio must be readable. Changing an extension does not convert a file.

Transcription uses account credits: a clip of up to 30 seconds costs 0.5 credits; longer recordings use minutes rounded up. The server validates the file, duration, and balance. Check the current balance above and plans and usage limits before submitting. This is not unlimited free or fully local processing.

Cross-talk, distant voices, noise, and similar voices can produce missing words or incorrect speaker groups. Speaker count is a hint, not a guarantee. Separate microphones and clear turn-taking make a recording easier to review. Optional summaries and translations follow the existing plan and credit rules and reuse the saved transcript.

Frequently asked questions

Can I correct speaker labels?

Yes. Open the completed result's speaker panel, edit a detected label, and move focus out of the field to save. Wait for the confirmation before leaving. Reopen the same task from your recordings to check the saved label. Labels identify groups within this recording. They are not verified identities.

Is the transcript ready to quote without review?

No. Listen to each intended quote in context and check words, numbers, names, and speaker attribution. Automatic text is a draft, including when the recording sounds clear.

Can I recover an interview after refreshing?

Yes, once a task has been created. Keep the page's task link or open it from My Recordings. Sign in with the owning account to restore progress and saved results. Unsaved local edits and files that were never submitted are different. Follow any prompt to reselect the original file.

Who can access my recording?

Task and download requests require the owning account. Processing happens in the cloud. Only create a sharing link intentionally, and use the existing revoke control when access should end. Keep participant consent and your retention policy in mind.