Real-time transcription
Speech turned into text live, with speaker labeling (diarization), optimized for Brazilian Portuguese.
Medpro.scriba listens to the encounter, transcribes in real time and extracts clinical data as the conversation happens — attention points, gaps and a SOAP draft for you to review. No dictation, no forms, no losing eye contact with the patient.
Physician
How long have you had this chest pain?
Patient
About three days, worse when I climb stairs…
flagAttention point
Exertional chest pain — assess cardiac risk
SOAP draft
S: Chest pain for 3 days, on exertion.
Real time
Audio → SOAP
Notes written after the encounter are slow, error-prone and pull attention away from the patient. Scriba moves documentation into the visit itself.
Much of the day goes to writing notes after the patient has left — slow work that piles up.
Filling screens and forms during the visit steals eye contact and attention from the patient.
Documenting from scratch, from memory, lets important details slip between one visit and the next.
Interactions and red-flags only surface at review — when it's already too late to act during the visit.
How we solve it
Scriba captures the audio, transcribes it live and extracts structured clinical data as the conversation happens. At the end it hands back a SOAP draft the physician reviews and consolidates into the record.
On every encounter, Scriba follows the conversation in the background and turns speech into clinical data — without interrupting physician or patient.
The microphone stays in the browser; audio goes up through the Medpro proxy. AI credentials never reach the client.
Transcription and extractions appear during the visit by streaming — not after it ends.
HTTPS and WSS, mandatory authentication and data in a dedicated database. Each physician owns their own records.
On completion, the encounter is locked for review and a diagnosis-validation job (RAG) runs against references.
Speech turned into text live, with speaker labeling (diarization), optimized for Brazilian Portuguese.
Clinical alerts with severity (high, medium, low) — interactions, red-flags, contradictions — to confirm, discard or defer.
Tracks the essential documentation items and shows, as a checklist, what is covered and what is still missing.
Diagnoses, medications, symptoms, vital signs and allergies captured from the conversation, structured and deduplicated.
The AI suggests specialty, risk level and primary diagnosis based on the content of the visit.
A note structured in four sections (Subjective, Objective, Assessment, Plan) for the physician to review, edit and consolidate.
In three steps
Open the encounter and start recording. The browser requests the microphone and the transcript appears live on screen.
While you see the patient, Scriba surfaces attention points, gaps, extractions and the clinical assessment in real time.
On close-out, review the SOAP draft, adjust what you need and consolidate the documentation into the encounter record.
AI credentials stay on the server, communication is encrypted and each physician owns their own records.
The browser never receives transcription or AI keys — every integration is server-to-server.
Audio and events travel over HTTPS and WSS, with mandatory authentication on every route.
Session data lives in a dedicated database, and each physician owns the encounters they create.
On completion, diagnosis validation (RAG) checks the hypotheses against cited clinical references.
Clinicians on the Medpro platform can already use Scriba at scriba.medproapp.com.br. Open it now or talk to a specialist.