SUGGESTED WORKFLOW
Work with interviews
Move from a local transcript to human-led coding, with an explicit boundary around cloud services.
An editorial workflow to adapt, not a benchmark reproduced by this site.
Prepare
Audio you are permitted to process, consent conditions, a working codebook, and a short excerpt for a pilot.
Check permission and setup
Use an approved device and storage location. Start with a non-sensitive sample; confirm whether transcription runs locally and whether any optional API is enabled.
Transcribe and correct
Consider noScribe for a graphical workflow or Whisper for a code-based setup. Manually correct a short segment before processing the rest; log model and language settings.
Pilot the codebook
Import corrected text into your coding environment, such as QualCoder. Define the unit of analysis, code inclusion rules, and examples; code independently before discussing differences.
Keep interpretation traceable
Link each interpretation to passages and memos. Record AI suggestions separately from accepted codes. Preserve ambiguity and negative cases.
Expected output
A corrected transcript, coding memos, disagreements, a revised codebook, and a record of any AI assistance.
Check your result
Check names, numbers, speaker changes, and omitted passages against the audio. Independently code a pilot excerpt and explain disagreements rather than optimizing a single agreement score.
Limits
A local application may still download models or call optional cloud services. Verify the actual configuration before using restricted material. This workflow has not been run by this site.