How to transcribe an interview without losing context

1. Prepare the recording and agree on its use

Keep the original file and a short note about who is speaking and what may be shared. Permission to record a conversation does not automatically settle how its words should be published. Confirm the intended use with the people involved, and avoid uploading confidential material without an appropriate agreement.

For ClipToDraft, use a supported audio or video file, such as MP3, M4A, WAV, or MP4, up to 50 MB. File size, the maximum length of one recording, and monthly minutes are separate limits. A Free account allows 30 minutes per month but only five minutes per recording; check your plan before uploading a full interview.

2. Create a transcript, then review each important passage

Open the audio-to-text tool, choose your file, and select Create transcript. When the result is ready, open it in the workspace. Use a timestamp to return to the relevant audio, then edit the segment text. Wait for the saved confirmation before leaving.

1. Check names, dates, amounts, negations, and any phrase you intend to quote.
2. Identify voices from the recording. Add a role or name as a text prefix only when you are sure; otherwise mark the speaker as uncertain.
3. If people talk over each other, replay the passage. Mark unclear words rather than filling a gap with a plausible guess.
4. Preserve the question, a self-correction, and any later answer that changes the meaning.

3. Keep the correction and the uncertainty

The original script below is an illustrative interview about a community workshop, not a customer conversation. A 19-second English recording made with synthetic voices was used to check the upload, text-editing, playback, and TXT/SRT export workflow on October 13, 2026. It is a workflow check, not an accuracy benchmark for real interviews.

Friday replaces Thursday, but the estimate remains an estimate. In a transcript, keep the correction. In a separate summary, you can write that the workshop opens Friday and that about twenty people are expected, with registration still open. Do not turn that paraphrase into a direct quotation.

Interviewer: When will the community workshop open?
Guest: Thursday, sorry, Friday. We expect about twenty people, but that is not a confirmed total.
Interviewer: Can I quote twenty as the final number?
Guest: No. Say we expect about twenty people. Registration is still open.

4. Export the text and keep a route back to the audio

Export TXT for reading and writing, and SRT or VTT when you need segment timing. TXT does not retain those timestamps. Keep the original recording alongside a timed export, and note the source time of each quotation separately. If you transcribed an excerpt, add its start offset when referring to the full interview.

Editing segment words in ClipToDraft preserves their timestamps; it does not retime, split, or merge subtitle cues. Check a timed export with the recording before using it as captions. For a written interview, retain speaker labels and relevant non-speech information, such as a meaningful pause or laughter, where needed to understand the exchange.

5. Separate the transcript from your interpretation

The Oral History Association recommends reviewing first-draft transcripts and checking names and dates against the recording. Keep an editing note if you remove repetitions or other spoken material. A cleaned reading copy, a direct quotation, and your summary serve different purposes; label them clearly.

Before sharing, read each quotation with the question and the following answer. Check that your headline or summary does not remove a condition the speaker relied on. If you create an AI draft later, use the reviewed transcript and perform this check again.

Before you finish

• The recording fits the upload and duration limits, and its use is agreed.
• Speaker labels, corrections, names, and uncertain words have been checked.
• Each quotation can be traced to the original audio and retains its context.

Common questions