Articles
How to transcribe an interview
How to transcribe an interview recording into a timestamped, searchable, quotable transcript.
By GistCite team, Product and engineering · For journalists, podcasters, researchers and students
Published · Reviewed
Key takeaways
- An interview recording needs a transcript with timestamps so a quote can be checked against the audio.
- Whisper upload handles recordings up to 500 MB with word timestamps and optional translation to English.
- A vocabulary hint helps with names and jargon a general model would otherwise misspell.
- Results are cached by content hash, so re-running the same file costs one unit the second time.
Recording to text
An interview recording is only useful once it can be searched and quoted. Uploading it to Speech to Text sends it to Whisper, which returns a transcript with word-level timestamps, for files up to 500 MB.
This matters for an interview meant for publication, where a direct quote has to match what was actually said, not a paraphrase from memory. Whisper upload handles the recording regardless of whether it was captured on a phone, a dedicated recorder or a video call.
Names and jargon
A vocabulary hint can be given before transcription to help with names, places and jargon a general model would otherwise get wrong, and the recording can be translated to English at the same time if needed.
Getting a name right the first time saves an editing pass later, particularly for an interview with several speakers or unfamiliar terminology.
Editing and exporting
The result downloads as plain text for a write-up or as SRT subtitles if the interview will run as video. Results are cached by content hash, so transcribing the same file again costs one unit instead of the full rate.
Because the transcript keeps word-level timestamps, correcting one misheard name does not require re-listening to the whole recording, just the few seconds around it.
From transcript to a draft
Once the interview is transcribed, AI Create From Source can draft a blog post, newsletter or brief from what was actually said, and Ask This Source can pull a specific quote with its timestamp for a fact check.
Standard and Deep quality are both available for the draft, so a quick version and a more thorough one can be produced from the same interview without transcribing it twice.