01
Define the codebook
Choose which API labels are relevant and document what they do and do not mean in the study.
Qualitative research
Scale interview review while keeping interpretation with the researcher
Use transcripts, time-local emotion labels, and speaking-style scores as candidate annotations for recorded interviews. Researchers can search and compare larger collections while preserving context, uncertainty, and human interpretation.
One file / one result
Candidate acoustic codes aligned to the transcript
I was relieved when the process ended, but I am still uncertain about what comes next.
relieved 0.79worried 0.58sincere 0.67
Unified analysis request
curl https://speech-api.oruk.ai/v1/audio/analysis \
-H "Authorization: Bearer $ORUK_API_KEY" \
-H "X-Request-ID: $(uuidgen)" \
-F "model=oruk-resonance" \
-F "file=@recording.wav"Workflow
02
Batch prerecorded files
Submit interviews under stable model and pricing versions and retain request-level provenance.
03
Review and adjudicate
Researchers inspect segments, override candidate labels, and report model-assisted coding transparently.
Useful applications
- Find candidate passages for close reading
- Compare aggregate expression patterns across interview sets
- Attach reproducible acoustic features to coded excerpts
- Prioritize human review without replacing it
Deployment safeguards
- Obtain research consent for automated speech analysis
- Do not treat model labels as participants’ self-reported feelings
- Evaluate cultural, linguistic, and recording-condition differences
- Document model version, thresholds, overrides, and missing data