Transcribe an HR or workplace investigation interview
Investigation interviews transcribed on your own machine, with the recording never leaving it. No vendor account, no upload, no copy of an allegation sitting on someone else’s server.
A timestamped, editable transcript you can quote from in a report — produced without widening the circle of people with access.
Language and translation need the Multilingual model.
Drop an audio or video file
or , paste with Ctrl + V, or
- MP3
- MP4
- WAV
- M4A
- MOV
- WEBM
- OGG
- FLAC
- AAC
The file is read by this page and never uploaded — no account, no size limit but your own RAM.
Preset: balanced English model, timestamped output.
Download
Click a line to jump the audio there. Click the text to fix a word — your edits go into every download.
How it works
- Open this page — the tool is already set up for “HR investigations”.
- Drop the file in, browse for it, or paste it with Ctrl+V.
- The model runs on your own device; the transcript appears as it goes.
- Correct anything misheard, then download text, subtitles or notes.
Confidentiality is the whole procedure
A workplace investigation depends on people believing that what they say will be handled carefully. The recording typically contains a named complainant, a named respondent, named witnesses and specific allegations — material that can end up in a tribunal bundle, and material whose uncontrolled spread can itself become a grievance. Uploading it to a transcription service puts a copy outside the organisation entirely, usually without the participants knowing.
Transcribing locally keeps the file where your policy assumes it is. It is read by the page, processed by your own processor and discarded when the tab closes. There is no upload endpoint on this site, which you can confirm by watching the network tab or by transcribing with the network switched off.
Quoting accurately in the report
Investigation reports live or die on accurate quotation, and an automatic transcript will get some words wrong. Every quotation you intend to use should be checked against the audio — click the line and it plays from exactly there, which makes verifying twenty quotes a ten-minute job rather than an afternoon of scrubbing.
Label the speakers as you go. In an interview with an investigator, a note-taker and an interviewee, unlabelled turns become ambiguous quickly, and ambiguity in an investigation record is exactly what gets challenged later.
Retention and disclosure
A transcript is much easier to copy, forward and disclose than an audio file — which is useful for the process and a risk for everything else. Decide before you start who will hold the transcript, where, and for how long, and apply the same retention rule you apply to the recording.
This tool does not store anything for you and cannot delete anything for you; what it stores in your browser is the speech model, which contains no case material. Everything about handling the file and the transcript remains an organisational matter.
Things that save a re-run
- Label investigator, note-taker and interviewee before exporting — unattributed quotes get challenged.
- Verify every quotation you plan to use by clicking the line and listening.
- Apply the same retention rule to the transcript as to the recording; it is easier to spread.
Read the transcript against the audio before you rely on it. Every speech model — this one and the paid cloud ones — mishears names, numbers and crosstalk, and it does so confidently. The player and editor here exist so that check takes minutes rather than an afternoon.
Frequently asked questions
Does the recording go outside the organisation?
No. It is read and processed inside your browser, and the site has no upload endpoint. Nothing is transmitted, so no copy is created outside your device.
Can it tell the interviewer and the interviewee apart?
Not automatically — Whisper does not identify speakers, and a wrong automatic label in an investigation record would be actively harmful. You assign labels manually; one click applies a name to that line and every line after it until the next label.
Is the transcript admissible or evidential?
Treat it as a working document. The audio is the record; a machine transcript is an aid to reading it. Verify any passage you rely on, and expect the transcript to be scrutinised if the matter escalates.
What about interviews conducted over Teams or Zoom?
Export the meeting recording and drop the file in — MP4 and M4A both work. Note that if the meeting was recorded to the platform’s cloud, a copy already exists there; transcribing locally does not undo that.