Local-first speech
High-quality transcription using Whisper-class and other on-device engines. Voice activity detection trims silence so recognition stays focused on speech.
Miccy
Ambient scribe for hospitals and doctors
Dictate, structure, and paste into any field at the cursor. For the default workflow, speech and optional formatting stay on your device—not our servers.
16-language UI · 50+ dictation languages · multilingual clinics, EU & Switzerland
Documentation support only—not medical advice. Follow your institution's policies.
Demo
Demo coming soon.
Demo video
Video will appear here.
Download for macOS / Windows / Linux
Free and open source (MIT). No per-seat subscription.
View all downloadsFeatures
Ambient structuring
After transcription, optionally run note templates - cleanup, bullet summaries, SOAP-style documentation, and more. Everything can stay on your computer; if your practice already runs a local model server, you can connect it in settings (see Models & Languages).
Paste where you work
Miccy uses system-level input: your text goes where the cursor is in the active window - clinical software, browser, Word, email, or a PiS / EHR. No special integration or API needed.
Global shortcut
Choose a keyboard shortcut that fits your workflow - start and stop dictation without leaving the app you are documenting in.
Works with any EHR / PiS
Because Miccy drives the computer like a keyboard, it is not tied to a single product. Use it with any system that accepts text at the insertion point.
History & control
Review past sessions, manage retention, and keep documentation under your control on your machine.
How It Works
-
Download Miccy
Install on macOS, Windows, or Linux. Pull updates when you choose - the app can check for new releases.
-
Pick your transcription model
Select the model that matches your language, accent, and hardware. Download once; use offline after that.
-
Pick your shortcut
Set a global hotkey - push-to-talk or hold-to-record - so you can dictate without clicking around the EHR.
-
Transcribe
Place the cursor where the text should go. Start the shortcut, speak, stop. Miccy turns speech into text on your computer.
-
Structure (optional)
Enable post-processing and choose a template to polish the transcript or structure the note - on your device by default. If your clinic runs its own compatible model service, you can point Miccy at it in the app settings.
Clinical workflows
A few places teams already use dictation-to-text: you place the cursor, run Miccy, and paste into the field you are in.
Outpatient consult note
Dictate history, exam, and plan into the open encounter in your PiS or EHR—then optionally tidy bullets or SOAP-style blocks on device.
Ward round note
Capture bedside findings and jobs for the list in one pass; paste into the ward system or shared document without switching apps mid-round.
Discharge summary
Speak the course, procedures, and follow-up; structure into sections before you drop the text into the discharge editor or letter template.
Referral or handoff draft
Draft the clinical question and summary for another service or GP; polish wording locally, then paste into your referral form or email.
Models & Languages
Speech-to-text (transcription)
Download models that match your language and hardware from inside the app. The table below is illustrative; exact names appear in Miccy’s model picker.
| Family | Examples (illustrative) |
|---|---|
| Whisper | Small, Medium, Turbo, Large - strong multilingual dictation; pick Auto or a specific language for better accuracy. |
| Parakeet | Optimised for English - very fast, high quality for English consultations. |
| Moonshine | English, very fast - great when latency matters. |
Exact model names and files are shown inside Miccy’s model picker.
Which model fits my computer?
Starting points for typical practice computers - exact models are listed in the app.
Note structuring (optional, on-device or your server)
After transcription, you can run templates for cleanup, bullets, SOAP-style sections, and more. By default this runs with a built-in local setup on your machine. If you prefer, you can point Miccy at your own compatible model service on your network (for example one your IT team already hosts). Nothing needs to leave your environment for that step when you configure it to stay local.
Languages
Multilingual clinics and EU / Swiss settings often need both a familiar UI language for staff and flexible dictation languages—Miccy is built for that combination out of the box.
16-language app UI
Run the app in the language your team prefers: English, German, French, Italian, Spanish, Chinese, Japanese, Korean, Polish, Russian, Ukrainian, Portuguese, Czech, Turkish, Vietnamese, or Arabic (right-to-left)—without patching separate builds per site.
Dictation
Automatic language detection, or choose from 50+ recognition languages (Whisper-oriented list). English-only engines are labelled in the app - use Whisper for the broadest multilingual clinical dictation.
Templates & structuring
Built-in examples are tuned especially for German clinical wording; you can adapt prompts for your own language and style. Prompts stay under your control whether you use the default local setup or connect your own server in settings.
Who it is for (and not)
Miccy assists with documentation; it does not diagnose or replace clinical judgment. Local processing is the default, but your institution still sets the rules.
A strong fit if…
- You want ambient-style speech-to-text and optional structuring without a per-seat cloud scribe.
- Your priority is keeping consultation audio and note text on hardware you control for the core workflow.
- You work in hospitals, practices, or outpatient settings where the PiS/EHR accepts text at the caret, or you draft in Word, email, or browser.
- You value open source (MIT), a global UI, and the option to tune models and prompts yourself.
Not the right tool if…
- You need a vendor-certified medical device or a formal integration contract with your EHR vendor—Miccy behaves like a fast typist at the keyboard, not a certified module.
- You expect guaranteed enterprise SLAs, central admin, or off-the-shelf clinical content packs from us—this is community software you deploy and govern.
- You require every byte of traffic to stay in a specific certified cloud region while still using hosted vendor ASR; the default here is local inference (you choose any remote endpoints explicitly).
- You want fully hands-off ambient documentation with zero review—human oversight and local policy still apply.
EHR / PiS Compatibility
Miccy does not need a custom interface from your hospital vendor. It works like a fast typist at your keyboard: text appears where the cursor is. That means compatibility with any documentation system that accepts text entry - today’s PiS/EHR and future tools - without waiting on integration projects.
Privacy
Miccy is built so speech and note structuring can stay on your device for the default local workflow. Model downloads and app updates use the internet; first-time setup may require downloading speech and optional structuring models. Read the Privacy Policy for details and for your jurisdiction.
FAQ
Is Miccy free?
Miccy is free and open source under the MIT license. You may still choose to download third-party models under their own terms from within the app.
How does Miccy compare to Dragon Medical, Heidi, or other paid scribes?
Products like Dragon Medical are established paid dictation solutions, often with enterprise licensing. Ambient AI scribes such as Heidi or Abridge (and similar offerings from other vendors) usually charge subscription fees and run your audio through vendor infrastructure. Miccy targets teams and individuals who want a free, MIT-licensed path with local processing for the default workflow - you trade vendor-managed polish for control, transparency, and no per-seat bill. Product names above are trademarks of their respective owners.
Does it work offline?
After models are downloaded, core transcription and on-device structuring can run without a network connection. Updates and model downloads need the internet.
Will it work with our EHR?
Miccy types into the focused field like a keyboard. If your EHR or PiS accepts text at the caret, Miccy can paste there without vendor-specific integration.
What about patient data?
For the default local workflow, audio and transcripts are processed on your machine. Your organisation’s policies and logging tools still apply - treat workstations like any other clinical endpoint.
Which languages?
The app UI is available in 16 languages. For dictation, pick from 50+ recognition languages or automatic detection, with Whisper offering the broadest multilingual coverage. See Languages above for detail.
Why we are making clinical AI free and 100% local.
Healthcare software should not force a trade-off between powerful documentation and patient privacy. We build Miccy so clinicians can run speech and note structuring on hardware they own - no mandatory cloud handoff for the core workflow, no per-seat tax on essential tools, and a path anyone can inspect, fork, and improve.
- Anamedi Team