what ATLAS.ti expects
ATLAS.ti's interview-coding workflow assumes a transcript with three things: per-turn timestamps in the form #hh:mm:ss.s#, named speakers (not "Speaker 1" unless you want to relabel everything later), and clean paragraph breaks at speaker turns. the synchronized-with-audio import path that lets you click a paragraph and hear the relevant audio depends on this format being right at import time.
a .docx with bold speaker names and run-on paragraphs imports — and then you spend an hour rebuilding the timestamps and breaking the paragraphs at every speaker switch. for a study with twenty interviews, that's a workweek.
what we ship
- per-turn timestamp prefix in ATLAS.ti's expected format:
#hh:mm:ss.s#on every speaker turn. matches the synchronization import pattern documented in ATLAS.ti's import docs. - named speakers from your codebook. relabel "P03" or "InterviewerA" once in the editor, propagate to every row for that speaker. saves the find-and-replace pass after import.
- per-turn paragraph breaks so ATLAS.ti's segmentation step has clean edges to work with. no more breaking apart paragraphs that the transcription tool merged.
- .docx export with the correct paragraph styling so the auto-coding pre-pass parses it. plain text export for researchers who code in a separate document.
workflow
- drop the interview audio. transcription runs and produces a first-pass transcript with diarization.
- review in the editor. relabel speakers from "Speaker 1" to study-codebook IDs (P03, P04, etc.) — bulk fix, one click, propagates everywhere.
- export to ATLAS.ti format. open ATLAS.ti, create a new hermeneutic unit, add the transcript document, import. the timestamps sync to the audio file automatically.
- code as normal. clicking a coded segment plays the audio from that timestamp.
where this fits
- fits: semi-structured interviews, focus groups, ethnographic field recordings. the same shapes that work with NVivo and MAXQDA work here.
- fits: longitudinal projects. custom vocabulary persists across files, so participant names and study-specific terms learned on file one transfer to file twenty.
- doesn't fit (yet): network analysis on transcribed text. ATLAS.ti's network views work on coded segments, not on the transcript directly. our export is a clean coding substrate; the network analysis happens after coding, in ATLAS.ti.
privacy
for IRB-restricted recordings — and especially for data residency requirements that prohibit cloud processing — run the file in private mode. the ATLAS.ti export works identically; the audio just stays on your laptop.
citation
when you publish, the methodology footnote we recommend reads: "transcribed using audiohighlight v[X], ATLAS.ti export profile, on [date]. transcripts produced in private mode and stored locally per IRB protocol [Y]." version numbers are stable across our releases for replication.