Transcribe oral history and life-story interviews, word for word
Upload an oral history interview, a grandparent telling their life story or an archive tape, and get a verbatim transcript with a speaker and timestamp on every paragraph, plus an optional translation under each one. Dialect is kept as spoken. A recording of up to 15 minutes is free in full; a longer one costs $5 for the whole interview, up to 6 hours.
- Verbatim: dialect, repetitions and mixed languages kept as spoken
- A timestamp on every paragraph, ready for a time-coded index
- A CSV of speaker turns for a catalog record or a spreadsheet
- Optional translation under each paragraph, so a heritage language and English sit together
- Up to 6 hours per interview for $5
Long interviews in the language people grew up with
Life-story interviews are long because memory takes its time. A session with an elder easily runs two or three hours, with pauses, tea, a phone ringing and stories that come back around. They are often told in the language of childhood: Punjabi, Greek, Polish, Vietnamese, Arabic, a village dialect of Kurdish. The grandchildren who will read the transcript may only read English.
Choose a translation language before you upload and every paragraph shows the original words first and the translation underneath. The original stays the record. The translation is an AI reading that helps family members and researchers follow along.
Keeping the voice of the narrator
Oral historians argue about how much to tidy a transcript. This tool takes the verbatim side: it writes what was said, in the words and dialect used, and does not smooth speech into textbook grammar. A narrator who switches from Spanish to English in the middle of a sentence gets both languages, each in its own words. Words that cannot be made out are marked [inaudible] instead of being guessed.
If your archive prefers a lightly edited transcript, the verbatim draft is a good place to start. Edit on screen before downloading, or keep the verbatim file as the master and make an edited copy for publication.
Indexes and catalog records
- Timestamps like [1:12:40] counted from the start of the file give you the times for an index of places, people and events
- Topics in order with a one-sentence neutral summary each work as a first draft of a session abstract
- The speaker-turn CSV (File, Time, Speaker, Text, Translation) imports into Excel, NVivo, ATLAS.ti or MAXQDA
- Notable quotes come out word for word with speaker and time, useful for an exhibit or a family book
- Rename Speaker 1 and Speaker 2 to Interviewer and the narrator's name with Find and replace
Before you start
Old cassettes digitized to MP3 or WAV work fine, but hiss, a quiet narrator and a loud interviewer cause mistakes. Proper names, place names and dates are the things to check by ear. Speaker numbers are given per ten-minute part, so on a long session read the labels again before you rename them. Languages outside the supported list may still come out, but check the free first part first.
How the tool works, what it costs and how your files are handled is on the Interview Transcription page; prices for every tool are on the pricing page.
Last updated
Questions and answers
Can ChatGPT transcribe a 3-hour oral history interview?
A chat assistant can be useful on a short excerpt. For a full session the file often exceeds what it accepts, and a long transcript gets shortened or stops partway because it does not fit in one reply; timestamps for a long file are not reliable either.
Is the transcript accurate enough for an archive?
It is a draft made by an AI speech model. It is good on clear speech and needs a human check on names, dates and unclear passages before it goes into a catalog.
What does a long interview cost?
The whole interview is $5, paid once, up to 6 hours. If you are working through a collection, Unlimited at $19 a week covers every interview up to 12 hours each and can be canceled any time.
Does the AI interpret or code the interview?
No. It writes the words, labels speakers and gives a neutral summary of topics. Coding and interpretation are left to you.
Is the recording private?
Yes. Only you can see it. It is sent to an AI model provider only to be processed and is never used for training, and you can delete the job at any time.