Technology
One engine. Real-time speech and clinical language.
Ozana AI runs on a single low-latency speech engine that powers both the Interpreter and the Scribe. This page summarises how it works, what it stores (nothing at rest), and how it integrates with clinical systems.
Real-time speech pipeline
Speech is captured, transcribed, translated and re-synthesised in a streaming pipeline that keeps end-to-end latency under 200 milliseconds. The same stream feeds the Scribe, so the SOAP note is drafted while the consultation is still in progress.
Zero-data-storage architecture
Audio and transcripts are ephemeral. They live only in memory for as long as the pipeline needs them, then they are discarded. No recordings are written to disk. No transcripts persist. Only the clinician-verified note leaves the session, sent to the EHR or downloaded by the clinician.
Multilingual clinical models
The models are tuned for medical vocabulary across 80+ languages, including symptoms, medications, procedures, safeguarding and consent. The output language is set per session - the clinician can read the note in English while the consultation happened in Arabic or Polish.
Security & integration
End-to-end encryption in transit and in memory. Runs in the browser, no patient app required. SOAP-structured, EHR-ready output. See Compliance for the full posture.