Clinical Conversation Transcription Demo

Realtime doctor/patient transcription with AI diarization and a Gemini full-audio second pass for higher accuracy.

Demo only. Do not use real patient PHI without hospital approval, signed BAA, and security review.

Audio setup

Single-microphone capture. Speaker separation is handled by AI diarization (live) and a Gemini full-audio pass after you stop.

Mono · AI diarization

Audio input device

Detected channels: —

Microphone input0%

Controls

Idle

Confidence threshold: 85%

Single microphone input. Live transcription uses Deepgram diarization; when you stop, the full recording is sent to Gemini 3.5 Flash for a high-accuracy second pass and reconciled with the live transcript.

Speaker 0 role

Speaker 1 role

Custom keyterms (comma separated — medications, names, conditions)

Boosts recognition of these phrases. Common medical terms are already included.

Live transcript

0 blocks

Start recording to see finalized transcript blocks here.

Flagged review queue

0 items

No blocks flagged. Low-confidence or clinically risky transcript blocks will appear here.

Audio setup

Controls

Live transcript

Flagged review queue

Export