AI-Transcriber
GPU-accelerated clinical transcription that runs on your own hardware. Record the visit or dictate the note; get back speaker-separated transcripts and structured clinical documentation — without a single second of audio leaving your network.
Interactive demo coming soon. Request a walkthrough →
Documentation that keeps up with the visit
Most physicians finish their charting after their patients have gone home. AI-Transcriber exists to give those hours back: record the encounter or dictate a paragraph mid-note, and the transcription happens on GPUs in the clinic's own infrastructure — with the finished text delivered straight into the chart.
The pipeline is built from best-in-class open models: Whisper-family speech recognition tuned for long clinical audio, pyannote speaker diarization to separate physician from patient, and clinical entity extraction. Because the whole stack is self-hosted, the privacy posture is structural — patient audio is processed and stored inside your network, full stop.
Inside PocketEMR it disappears into the workflow: a live recorder in the chart, a dictate button inside the note editor, transcripts viewable on screen, and notes that populate themselves the moment transcription completes. Long visits are welcome — uploads up to 700 MB are supported for extended recordings.
Features
What's included
GPU Speech Recognition
Whisper-class models running on dedicated GPUs transcribe full visits — not just short dictations — with clinical vocabulary accuracy.
Speaker Diarization
Tells the physician and the patient apart automatically, so the transcript reads as a conversation, not a wall of words.
Structured Clinical Notes
The visit becomes a structured note — SOAP or free-form — ready to review, edit, and file to the chart.
Self-Hosted by Design
Audio and transcripts are processed on your own GPU hardware. Nothing leaves your network — data sovereignty by architecture, not by policy.
Record Anywhere in the Chart
Live visit recording with wake-lock and take-by-take review, or one-button dictation straight into a clinical note — the text lands where you are working.
Built into PocketEMR
First-class inside PocketEMR — transcription jobs, playback, transcripts on screen, and notes that auto-populate when the transcription finishes.
Ready to try AI-Transcriber?
We're onboarding a select group of Ontario specialists. Get a personalized demo and see how AI-Transcriber fits your workflow.