Choose one of three approaches. The steps for that option appear below.
Choose an option above. On a phone, swipe the tabs left or right to see all options →
Otter.ai transcribes and labels speakers. You check every quote against the audio and get approvals.
Get ready
Transcribe and label
Replay the quotes
Get approvals
Use: You
The interview recording and the agreed quotation scope. Record consent before any upload.
Upload only after consent is recorded.
Use: Otter.ai
Import the recording, label the speakers and list quotable passages with transcript timecodes and an empty approval column.
What to paste into the tool
Task: Otter transcript draft; Otter generated notes packet. Use the Import control rather than a chat command; save the draft transcript and any summaries, action items, outlines, or template notes as the named packets. Second message, after your check: Task: Speaker labels and quote candidates. Label speakers across this transcript. List quotable passages with their transcript timecodes. Add an approval-status column and leave it blank.
Use: You
Replay every candidate quote word for word against the untouched audio; map timecodes to the original recording; repair discrepancies.
Release the import result only after processing has finished; check five random speaker labels by listening before trusting the rest; move to interviewee review only after each selected quote has a checked original timecode.
Use: You
Share only the selected quotes with the interviewee through their chosen channel and record each response; unanswered quotes stay unapproved.
Release the package only after every quote has an explicit approved or unapproved status.
Descript transcribes and detects speakers. You correct the whole draft by ear; a reviewer replays the quotes.
Check terms and prep
Transcribe and detect
Correct while listening
Replay the quotes
Get approvals
Use: You
Have ready: the interview recording. Check permission and data terms before adding a file to Descript.
Start the project only after permission and data terms are clear.
Use: Descript
Import the file, run transcription and speaker detection, and keep the text as an unchecked draft.
What to paste into the tool
Task: Descript detected-speaker draft. Use the project import and Identify speakers controls; keep uncertain labels as aliases until the editor checks samples
Use: You
Listen through the entire original while reading the draft; correct wrong words and add original paragraph timecodes.
Advance from detection only when the imported file has finished transcribing; test five random speaker locations before trusting the rest; continue to quotation checking only after the full listening pass is recorded.
Use: You
Replay each selected passage separately, compare every word and timecode, fix mismatches and add a review status field.
Submit quotations only after each selected passage is checked.
Use: You
Ask the interviewee to approve the exact quotation wording, record approved or unapproved per entry, and combine the checked list with the transcript.
Finish only when each quotation carries its approval status.
You transcribe and choose quotes first. ChatGPT re-transcribes excerpts for comparison; you reconcile and get approvals.
Transcribe locally
Cross-read the excerpts
Replay and repair
Get approvals
Use: You
Listen to the source and type the speaker-labelled transcript locally; select quotations with timecodes.
Share the excerpts only after consent is recorded.
Use: ChatGPT
Request a second transcription of the cleared excerpts, then a list of every wording or speaker difference against your local text.
What to paste into the tool
Task: Independent excerpt transcript. Transcribe the excerpts without filling inaudible words; keep ambiguous speakers as numbered voices and identify uncertainty explicitly. Task: Difference list. These two texts cover the same audio excerpts. List every passage where wording or speaker attribution differs, with the excerpt and position in each text. Do not resolve the differences.
Use: You
Replay every listed difference and each selected quotation against the original recording, record original timecodes and repair the transcript and quote log.
Accept the response for comparison only when the attached excerpts have been processed; check the difference list covers both texts; replay every difference and chosen quotation against the original before passing the transcript on.
Use: You
Present the checked quote log for speaker approval through a chosen route; mark absent approval as unapproved and keep the local transcript.
Close the log only once each entry records approved or unapproved.
Best when: A transcript draft should come before quote selection
What it does: Otter.ai transcribes; you check quotes and get approvals
Strong at: Import is a documented transcription entry point
Limit: Speaker sampling leaves unsampled passages untested
Free plan: Yes, limited (300 monthly minutes, 3 lifetime imports)
Best when: You intend to check the complete script by listening
What it does: Descript drafts; you correct, replay and collect approvals
Strong at: Detection precedes correction while the recording stays unchanged
Limit: Every paragraph needs human attention in the listening pass
Free plan: Yes, limited (60 media minutes a month)
Best when: Quote choices are made before any machine reading
What it does: ChatGPT cross-reads excerpts; you reconcile and finish
Strong at: The second reading is restricted to selected excerpts
Limit: The full transcript stays a local typing task
Free plan: Yes
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Never paste credentials or full payment details.
Quote wording is checked by playback against the original recording.
Record a paragraph timecode and test speaker names at five random timecodes.
Each selected quotation has an approved or unapproved status before publication.
Check the result against your original inputs. Approve it before sharing or acting.