EU Managed Cloud Beta — Free until 30 June 2026 — No credit card required
Privacy-first clinical AI

Clinical notes that stay under your control.

AI-powered documentation and workflow automation. EU managed cloud beta open now — go fully offline on Mac, or self-host in your own infrastructure. We never train on your data.

No credit card required. Works in any browser.

Scribe UI on MacBook — transcript on left, SOAP note on right, evidence chips visible
On-device or private cloud
Zero training on your data
Evidence-anchored notes
Open evaluation frameworks
How it works

From conversation to verified note.

Four steps. Each one reviewable. Nothing reaches the medical record without clinician approval.

1

Listen

Record the consultation. Real-time audio capture with automatic speaker detection.

2

Transcribe

Speaker-identified transcript with timestamps. Doctor, patient, and other participants labeled automatically.

3

Extract & structure

Clinical data pulled from the transcript — chief complaint, history, assessment, plan — with evidence anchoring to the source.

4

Review & finalize

Verify every claim against the original transcript. Edit, approve, and export. You stay in control.

Pipeline flow visual
Deployment

One platform. You choose the boundary.

Same pipeline, three environments. Start anywhere — upgrade without starting over.

For Mac users who want fully offline

Omi Scribe for Mac

Everything runs on your laptop. Zero internet, zero cloud.

  • On-device speech-to-text & LLM
  • Encrypted local storage
  • No internet required
  • macOS only (Apple Silicon)
For health systems

Enterprise Self-Hosted

Deploy inside your own infrastructure. Full governance control.

  • Your VNET, your GPUs, your rules
  • SSO, RBAC, audit logs, retention
  • Connect to Foundry, Bedrock, or Vertex
  • EHR & workflow integrations

Not sure where to start? Most teams begin on the managed cloud and move to self-hosted when governance or scale requires it.

Beyond the note

The note is the starting point,
not the finish line.

Documentation is only half the admin burden. What comes after — referrals, prescriptions, EHR entry, billing — takes just as long. Omi Scribe automates post-note workflows safely inside your environment.

Referral letters

Drafted from the consult note — ready for review and send.

Prescription prep

Pre-filled for clinician approval before signing.

EHR pre-fill

Structured data pushed to your records via approved integrations.

Billing support

Suggested codes based on the documentation — ready for coder review.

Automation panel with next actions beside a completed note
Security & trust

We never see your data unless you choose to send it.

Privacy isn't a feature we bolt on. It's how the architecture is designed.

Data minimization

Processing happens on-device or inside your private environment. We only see what you explicitly choose to send.

Zero training policy

We do not use customer audio, transcripts, or notes to train our models. No exceptions.

Review-first workflow

Nothing enters the permanent medical record until a clinician reviews and approves it.

Enterprise governance controls

Self-hosted deployments include full administrative control.

SSO / OIDC RBAC Retention policies Audit logs Admin dashboard

Omi Scribe for Mac and Managed Cloud are built to meet HIPAA and GDPR requirements. For self-hosted enterprise deployments, we provide the technical controls — your organization applies them within its own compliance program.

Open & transparent

Trust built on verifiability, not marketing.

We publish what we can. Evaluate our work. Run it yourself.

What's open today

  • Medical STT Benchmark — evaluation framework for speech-to-text models on medical conversations. 26 models benchmarked on 55 real-format dialogues.
  • Clinical SOAP Note Evaluation Benchmark — lightweight benchmark for evaluating clinical note generation. Measures safety, factual grounding, coverage, and quality across scribe systems.
  • Sum Small — open-source SOAP model — a fine-tuned clinical language model that generates SOAP notes from medical dialogues. Outperforms GPT-4 on our benchmarks at a fraction of the size.

Coming soon — Community Edition

A free, open-source AI scribe. Run it locally via Docker. Connect to cloud APIs or local models with Ollama. Fully hackable.

  • Self-contained Docker deployment
  • Modify prompts, templates, endpoints
  • Free and open-source forever
Get notified

Frequently asked questions

Is patient data sent to your servers?

Managed Cloud: EU beta available now — data is processed in the EU under strict access controls. Australia and US hosting are opening later via waitlist.
Mac: No — all processing is on-device. Data stays encrypted on your laptop.
Self-Hosted: No — data stays inside your own infrastructure.

Do you train on my audio, transcripts, or notes?

No. We do not use customer content to train our models. This applies to all deployment options.

Is Omi Scribe HIPAA / GDPR compliant?

Omi Scribe for Mac and our Managed Cloud are designed to meet HIPAA and GDPR requirements — including encryption, access controls, data residency, and a zero-training policy. For self-hosted enterprise deployments, we provide the full set of governance controls (SSO, RBAC, audit logs, retention); your organization applies them within its own internal compliance program. Enterprise deployments can be covered by a BAA.

I don't have a Mac — can I still use it?

Yes. The managed cloud version works in any modern browser — no install needed. The Mac app is an additional option for clinicians who want fully offline processing on macOS.

Can I connect it to our EHR and internal systems?

Yes. The self-hosted enterprise deployment integrates with your local stack — EHR systems, internal APIs, and approved integration patterns — for post-note workflows like referrals, prescriptions, and billing.

What note types do you support?

SOAP notes today, with additional templates planned (H&P, progress notes, discharge summaries). Enterprise customers can configure custom templates through the admin dashboard.

What's the upgrade path from cloud to self-hosted?

Start a pilot on managed cloud, then move to self-hosted when your governance or scale requires it. The pipeline is the same — no restart needed.

What models do you use?

We develop in-house speech-to-text and clinical language models with published evaluations. On self-hosted deployments, you can run our models on your GPUs or connect to your own model framework (Azure Foundry, AWS Bedrock, Google Vertex).

Start in minutes. Scale with control.

Start the EU managed cloud beta for free, join the Mac beta, or talk to us about self-hosted deployment.