Oral History Transcription: How to Transcribe Long Interviews

To transcribe an oral history interview, start with an AI draft, then listen once to correct names, mark who is speaking and add timestamps at speaker changes. On TranscribTxt, while you have a free file left, the first 30 minutes are free, and a whole interview up to 3 hours is $2, paid once.

Short answer: Upload a copy of the interview; the language is detected automatically. While you have a free file left, the first 30 minutes come back free with TXT, SRT and VTT downloads, so you can check the text before you pay. The whole interview up to 3 hours is $2, paid once, and also downloads as JSON with word times and Word, with a time mark on every paragraph of the Word file. Several interviews in one week fit the $5 Week Pass. TranscribTxt doesn't label speakers, so you add interviewer and narrator initials while you review.

What an oral history transcript needs

A transcript makes a recording easier to find and quote, but according to the University of Pittsburgh Oral History Toolkit (updated Oct 16, 2026), its goal is to "enhance the audiovisual recording, and not to supplant it." The Oral History Association (OHA) adds that archives should offer, when possible, "indexes with time tags linking to the recording."

The UTRGV guide (updated Sep 15, 2026) asks for a header, speaker initials and HH:MM:SS timestamps; the Baylor style guide (revised May 2018) covers wording and unclear words.

Guide wording from the University of Pittsburgh Oral History Toolkit (updated Oct 16, 2026), the UTRGV Oral History Interviews guide (updated Sep 15, 2026), the Baylor University Institute for Oral History Style Guide (revised May 2018), and the Oral History Association's Best Practices and Statement on Ethics, all checked Nov 1, 2026. Your archive, IRB or instructor may ask for something different, and their rules come first. TranscribTxt (RESTART AI LLC) isn't affiliated with these institutions. Found an error? Write to support@transcribtxt.com and we will check and fix it.

Before you upload: consent, release forms and the original recording

OHA recommends these steps:

• Get and document the narrator's informed consent before the interview. Tell them what will happen to the recording, and if an online service will transcribe it, say so in plain words. OHA's July 2026 session on AI transcription listed privacy, informed consent and intellectual property among its ethical topics.
• Secure a signed legal release, ideally when the interview ends, following the policy of your repository.
• Copy the original recording off the device, keep redundant copies in different physical places, and upload a copy, never your only file.
• Ask the interviewer for a word list of names, places and terms; Baylor says it gives "a more accurate transcript."

TranscribTxt deletes your upload after processing (a free-preview upload is kept for up to 24 hours so you can finish the file). The audio goes to our speech-to-text providers in the United States, which don't use it to train models; our privacy policy has the details. If your project has an ethics or IRB protocol, see our research interviews page.

This is general information, not legal advice, and your repository, IRB or instructor has the final word.

How to transcribe an oral history interview, step by step

1. Upload a copy of the recording: MP3, M4A, WAV, MP4, MOV, MKV, WebM or AVI, up to 2 GB per file. Your first file needs no account.
2. The language is detected automatically. You see upload progress, then "Transcribing…", then the text. In October 2026, half of the uploaded files were transcribed in 14 seconds or less, and a 145-minute recording took 2 minutes 45 seconds. Times don't include uploading, which depends on your connection. Dialects and older recordings, such as digitized tapes, can need more correcting.
3. An interview longer than 30 minutes gets its first 30 minutes free while you have a free file left, with TXT, SRT and VTT downloads. Check it against the audio to see how much correcting the rest needs.
4. Transcribe the whole interview up to 3 hours for $2, paid once. It finishes the same upload and adds JSON with word times and Word.
5. For several interviews, the $5 Week Pass covers 7 days and up to 25 hours of audio, paid once, and doesn't renew. Eight 50-minute interviews transcribed in one week fit one pass. To price a whole collection, try our cost calculator.

How to mark the interviewer and narrator by hand

TranscribTxt does not label speakers. In the Word file, paragraphs are cut by length, not by who is talking, so a change of speaker can fall inside a paragraph.

In one listening pass:

1. Under the title of the Word file, type the header and an Acronyms line for the initials.
2. At each change of speaker, start a new paragraph with the speaker's initials and a colon.
3. If that paragraph has no time mark, take that line's start time from the SRT or VTT and type it as [HH:MM:SS].
4. Add Baylor's marks: (laughs) for non-speech sounds, (??) after an unclear word.

An example, with made-up names:

Narrator: Rosa Delgado
Interviewer: Sam Ortiz
Interview date: 03/14/2026
Interview location: Tucson, Arizona
Acronyms: RD (Rosa Delgado), SO (Sam Ortiz)

[00:41:05] SO: What do you remember about the store?
[00:41:09] RD: My father – my father built the counter himself. (laughs) We sold everything from it.

As of October 31, 2026, Microsoft's help page says Word's Transcribe needs a Microsoft 365 subscription (300 minutes of uploaded audio a month) or a Copilot license, and labels speakers. More in our post on Word's Transcribe; general tips are in our interview guide.

Verbatim or clean verbatim: style choices for oral history

The guides lean verbatim but differ in detail. Pitt says not to fix the narrator's grammar and offers a sample note calling the text a verbatim transcript. UTRGV leaves crutch words and false starts to your instructor. Baylor allows no more than two crutch words per occurrence per page, takes a middle course on false starts, always keeps repetition for emphasis, and spells out "going to" instead of "gonna."

TranscribTxt's draft often leaves out "um" and "uh," so it sits closer to clean verbatim. For a strict verbatim transcript, add back what your guide keeps while you listen. Brackets add clarification, as in UTRGV's example: "They [the school board] voted down the proposal."

Pick one guide for the whole project and name it in a note at the top of each transcript. Our post on verbatim vs clean verbatim transcription compares the two styles.

Timestamps, time-coded indexes and WebVTT

Some archives name the file type; for Pitt's, "Transcripts must be delivered as WebVTT files."

If your archive takes WebVTT, start from our VTT file (every plan, free included) and correct it in a caption editor such as CADET or Aegisub, both on Pitt's tool list.

The Word file and the VTT are separate, so a fix in one doesn't reach the other: pick your master file before you start correcting. For a rough time-coded index, our SRT to text tool with "Keep timestamps" on gives one line per subtitle with its time.

OHMS, an open-source tool from the University of Kentucky's Louie B. Nunn Center for Oral History, links a search term in a transcript or index to that moment in the recording. On November 7, 2024, the Nunn Center announced it was reworking OHMS to work with WebVTT; check the OHMS site for what the current version accepts.

More on time marks in our timestamps guide.

Review, narrator approval and archiving

• Check names, places and numbers against the audio first. Before you quote a line, play it from a second or two before its time mark to hear the lead-in.
• Send the edited transcript to the narrator for review and approval, and keep a record of their changes. OHA lists the narrator's options: deleting, restricting or redacting parts, adding clarifications, correcting mistakes, or keeping the interview closed until a set date.
• Deposit the recording, transcript and release together, in the format your repository names.
• Download every file you need when it's ready, because the upload is deleted after processing.

To cite the finished interview, see our guide on how to cite an interview in research.

Start with one interview and read the free first 30 minutes while you have a free file left. For talks or class sessions in the same project, see our lecture transcription page.