Discovery & usability
UX & product research
You quote participants back to the team, and a paraphrase is not something you can defend.
Interview & Research
What they said. Word for word.
A verbatim transcript with the speakers kept apart and the timestamps intact — and no AI summary written over the top of it.
Start for FreeFree 15 minutes · no credit card
plus.subanana.com
Transcript
Interviewer00:04:12
Walk me through the last time you gave up on it.
P0400:04:19
I kept the tab open for two days, then closed it.
Interviewer00:04:31
What would have made you come back?
P0400:04:36
Honestly? Somebody telling me how long it would take.
If the exact words are the evidence, this is for you.
Four kinds of interview, one requirement — the record has to survive being checked.
Fieldwork & oral history
Academic & field research
You need timestamps and speaker turns you can code yourself, in the tool you analyse in.
Sources & panels
Journalism
The quote has to be exact, and you need to find the moment it was said in under a minute.
Hearings & case notes
Compliance & intake
Who said which line matters as much as what was said, and someone else has to read it.
How does it work?
Onboarding study · P04
48 min · 2 speakers
Upload the recording, or record in the browsertell it how many speakers to expect
Verbatim transcript
Interviewer00:04:12
Walk me through the last time you gave up.
P0400:04:19
I kept the tab open for two days, then closed it.
No summary
Nothing is written over what was said
Transcript with speakers
DOCX · TXT · Markdown · PDF
Spreadsheet for coding
XLSX · pick your own columns
Timed text
SRT · VTT · TXT with timestamps
A translated copy
the original always stays alongside
Ask the transcript
answers quote the cue and its time
Nine reasons researchers keep the transcript, not the summary.
Not a feature list — the things that decide whether you can quote it in front of the people who were there.
Verbatim, and nothing on top
- No summary, no chaptersNothing is invented on your behalf. The transcript is the deliverable.
- Quotable as spokenWhat you put in a findings deck is what the participant said, not a paraphrase.
- Hesitations stay inFalse starts and long pauses are kept — sometimes those are the finding.
- Change your mind laterSwitch the project type and reprocess for a summary. Once, at no extra cost.
Ask it, and it shows its receipt
- Ask across 90 minutesInstead of scrubbing for the moment you half-remember hearing.
- Speaker and timestampEvery answer carries them, so you can check it against the audio.
- Anchor to a cueQuote a specific line into the question when the answer has to sit on it.
- Threads are savedA re-read months later starts where you left off, per project.
It leaves in your analysis shape
- XLSX, columns you pickIndex, timestamp, speaker, text, translation — in whatever order you code in.
- Readable formatsDOCX, TXT, Markdown or PDF when it needs to be read rather than coded.
- Timestamps your wayIn or out, with a blank line between turns if that is your format.
- SRT and VTTFor when the transcript has to line up against the recording itself.
It reads your guide first
- Attach a background packThe interview guide, a domain primer, or a URL — handed over before it starts.
- Glossary per studyOn top of the workspace list that applies to everything you run.
- Pull terms from a PDFLet the AI find candidates and approve them yourself before transcription.
- It matters most when hardAccents, jargon and crosstalk are exactly where a guide earns its keep.
Cleanup you approve, line by line
- A score out of 100Suggestions sorted by kind, in ten to thirty seconds.
- Structure without rewordingApply sentence breaks and paragraphing while the wording stays as spoken.
- Keep the hesitationsWhen the false starts are the finding, leave that group unapplied.
- Or skip it entirelyNothing is applied unless you press the button.
It stays as private as you set it
- Sharing is opt-inInvite named members, or turn on a public link only when you mean to.
- Guests are read-onlyAnyone opening a shared link gets a read-only view and read-only chat.
- Not used for trainingYour content does not go into model training, on any plan.
- Delete means deletedRemove a project and its transcript goes with it.
You say how many voices
- Set the count up frontSpeaker number is something you tell it, not something it guesses afterwards.
- Rename onceChange “Speaker 2” to “P04” and every one of their turns follows, everywhere.
- Every turn timestampedSo a quote can always be found again in the audio.
- See who did the talkingUseful when you suspect the interviewer did too much of it.
Choose how hard it thinks
- Three tiers, per projectStandard, Advanced or Premium — picked per recording, not set for the account.
- Speed or qualityWeigh it either way depending on how soon you need the transcript.
- Spend it where it countsSave the expensive pass for the recordings that are genuinely hard to hear.
- It only changes hearingThe mode never changes what is done with the words, only how well they land.
Interview in one, analyse in another
- Transcribe as spokenIn the participant's language, then translate the whole transcript into yours.
- Turn for turnThe original sits beside the translation, so any quote can be checked.
- Both columns, one fileExport as XLSX, or a ZIP with the original and the translation draft.
- Appendix and bodyQuote the original in one and the translation in the other, from one source.
And the rest of it.
None of these change the method, but each one removes an afternoon of tidying up.
Upload a recordingRecord in the browserScreen recordingEditable transcriptFind and replaceFull-text searchTimestamps on or offBlank line between turnsCustom spreadsheet columnsDOCX exportXLSX exportTXT exportMarkdown exportPDF exportSRT exportVTT exportFoldersBulk exportShare linkWorkspace rolesSwitch project type
Fifteen minutes are free, and nothing you upload is used to train a model.
Transcribe an interviewQuestions people askbefore they trust a transcript
Yes. Tell it how many speakers to expect and the turns come back separated, with timestamps. You rename the labels once and the whole transcript follows.
Not unless you ask. This project type is deliberately verbatim — no chapters, no auto summary, nothing written over what was said. If you want a summary, switch the project type and reprocess.
Export to XLSX and choose the columns — index, timestamp, speaker, text — so it drops straight into whatever you code in. DOCX, TXT with or without timestamps, Markdown, PDF, SRT and VTT are all there too.
95+ languages, including Cantonese and code-mixed speech. Add one translated copy alongside — the original transcript stays the record, the translation sits next to it.
No. Your content is not used to train models, and you can delete a project and its transcript whenever you want.
It will not code your themes for you, and it should not be the only thing standing between you and a published quote — check the passage against the audio, which is why every line keeps its timestamp.
Questions people askbefore they trust a transcript
Yes. Tell it how many speakers to expect and the turns come back separated, with timestamps. You rename the labels once and the whole transcript follows.
Not unless you ask. This project type is deliberately verbatim — no chapters, no auto summary, nothing written over what was said. If you want a summary, switch the project type and reprocess.
Export to XLSX and choose the columns — index, timestamp, speaker, text — so it drops straight into whatever you code in. DOCX, TXT with or without timestamps, Markdown, PDF, SRT and VTT are all there too.
95+ languages, including Cantonese and code-mixed speech. Add one translated copy alongside — the original transcript stays the record, the translation sits next to it.
No. Your content is not used to train models, and you can delete a project and its transcript whenever you want.
It will not code your themes for you, and it should not be the only thing standing between you and a published quote — check the passage against the audio, which is why every line keeps its timestamp.