Medical dictation software
A product by Softix
Dictate findings.
Stay offline.
Every word counts.

Runs speech recognition directly on your workstation — no cloud subscription, no patient data leaving your network. Press a hotkey, speak, and the transcript injects straight into your reporting system.

Activate anywhere with Ctrl Space
70+
Radiology terms boosted in decoder
0
Bytes sent to any external server
16k
Hz audio sampling — clinical grade
Startup

Loads once.
Ready in seconds.

On first launch, MedScribe loads the speech model from the local ./model directory. Subsequent launches use a cached version — startup takes under 3 seconds. No internet, no login, no waiting on a server.

MedScribe initialization screen
MedScribe loading sequence — model cache hit, engine ready
Live dictation

Works inside your
reporting system.

MedScribe sits alongside Word, your RIS, PACS, or any reporting tool. Dictate your findings, then click "Dictate into App" — the transcript types itself directly into the active document. No copy-paste.

MedScribe running alongside Microsoft Word
MedScribe running alongside Microsoft Word — live transcription into an abdominal ultrasound report
Features

Built for the imaging workstation.

Everything your reporting workflow needs — nothing that puts patient data at risk.

Fully offline inference
The speech model runs entirely on your CPU — no cloud API, no internet required. Works in air-gapped hospital networks.
Direct text injection
Switch to your RIS, PACS, or Word — click Inject and MedScribe types the transcript directly into the active field.
Medical vocabulary
4.5× boost applied to 70+ radiology terms — atelectasis, cardiomediastinal, hydronephrosis resolve correctly every time.
pleural effusionatelectasispneumothoraxconsolidationcardiomediastinalhydronephrosisinfarctstenosis+60 more
Live preview
MedScribe re-runs inference every 800 ms while you speak and shows a live transcript — catch errors before you stop recording.
Privacy

No patient data leaves
the workstation.

This is not a policy claim — it is an architectural one. MedScribe has no network code.

  • Air-gap compatible. Runs on machines with no network access.
  • Audio never written to disk. Lives in RAM only, released after inference.
  • Transcripts stay local. Never synced, never uploaded.
  • Offline license verification. One-time hardware handshake — no server call ever again.
Network activity log
Cloud endpoints (AWS / Azure / GCP)Blocked
Any external HTTP requestBlocked
Audio capture → RAM onlyLocal
Model weights (disk cache)Local
Transcript draftsLocal
A product by Softix
MedScribe is built by Softix — a software studio based in Lebanon building offline-first tools for professionals. One-time license, runs locally, ships when ready.
Visit Softix

Ready to bring offline dictation
to your department?

Per-workstation licensing · no subscriptions · works day one