Choose the interview recording
Upload an MP3, WAV, M4A, or MP4 up to 200 MB or 20 minutes. Keep the original recording available for the final quote check.
Interview transcription is not only about producing a block of words. A useful draft keeps each answer close to its source so a writer, researcher, or editor can verify a quote, locate a detail, and distinguish a question from an answer during the editing process.
Create a private, timestamped interview draft for quote checking and editorial review.
This browser workflow accepts common audio and MP4 files, runs the model locally, and presents the result as timestamped segments. Add the interviewer name, guest name, organization names, places, and specialist vocabulary before processing so the words that matter are easier to find during review.
Speaker labels are assigned manually in the current version. That keeps the limitation visible: automatic diarization is not promised, and conversations with interruptions, cross-talk, or distant microphones require careful listening. Export the checked result as TXT, JSON, SRT, or VTT.
Upload an MP3, WAV, M4A, or MP4 up to 200 MB or 20 minutes. Keep the original recording available for the final quote check.
Add the interviewer, guest, organizations, locations, and specialist terms that should be checked in the transcript.
Replay timestamps for quotations, edit the answer text, assign manual labels, and export the format needed by your writing or editing workflow.
Segment timing helps an editor locate the original wording instead of relying on an unverified paraphrase or a copied block of text.
Names, figures, dates, job titles, and direct quotes deserve a second listen even when the general meaning of the transcript looks correct.
The transcript is designed to speed up research and editing. It is not a publication-ready record without human review.
Example output only. The actual transcript depends on the recording, language, microphone, and background noise.
[00:07:26.110] Speaker 2: The biggest change was moving customer support into the product team.
Yes. Assign Speaker 1 through Speaker 4 to each segment during review. The current local build does not automatically identify speakers.
Click the segment timestamp or a word to jump to its estimated audio position, then compare the text with the recording before publishing.
Yes. Add the guest name, organization, products, places, and specialist words to the names and terms field.
TXT is convenient for reading, JSON preserves structure, and SRT or VTT are useful when the interview will be paired with video captions.