Omi Health

Safe voice AI
for healthcare.

The most accurate medical speech-to-text we’ve measured — with zero invented drug names. #1 of 28 systems on our open benchmark; run it in our cloud, yours, or fully on-device.

Consultation audio

“Start amoxicillin 500 mg for the chest infection, and repeat HbA1c at review.”

Every drug name transcribed correctly
Flagship · cloud API

omi-medical-1.

The flagship model — #1 medical accuracy of the 28 systems we measured, with zero fabricated drug names. Live as an EU-cloud API, or self-hosted in your own tenant.

Medical-term error rate 28 systems · lower is better
omi-medical-1 (ours)0.94%
ElevenLabs Scribe v20.97%
Google Chirp 31.11%
Gemini 3.1 Pro1.34%
Zero fabricated drugs, every test set
Open source · on-device

omi-medical-edge-1.

The best open on-device medical speech-to-text we measured. Open weights (CC-BY-4.0), runs fully offline on Mac, CUDA or CPU — no audio ever leaves the device.

on-device · offline
$ pip install omi-med-stt
$ omi-med-stt consult.wav
loading omi-medical-edge-1 (0.6B) … ready
transcribed locally — no audio left this machine
Research & Open Source

We publish what we'd want to evaluate ourselves.

Evaluation frameworks, benchmarks, and open-licensed models. Reproducible, not marketing.

Build on it.

Get an API key and test the engine on your own audio — free during preview.