Comparison
Both dictate offline into any application. The difference is whose models do it, and where the text goes afterwards.
As of 19.09.2026
Superwhisper recognises with Whisper and tidies the text with a language model of your choosing, up to GPT-5 through your own API key. It costs $8.49 a month, runs on Mac, Windows, iOS and Android, and works offline, though the vendor itself says the local models only run really well on Apple Silicon. MediSync Dictate uses its own models, trained on German clinical language, can run both steps entirely on the machine, and brings a processor agreement under Art. 28 GDPR on every plan.
The most important difference is who gets to see the text. Superwhisper lets you choose the language model that tidies it and connect it with your own API key. For many jobs that is a strength. For a practice it means history text going to a vendor your organisation has no processor agreement with, and the liability is yours, not the tool's.
The second is vocabulary. Whisper is a general model across more than 100 languages. Our recognition and Smart Dictation are trained on German clinical language: doses, lab values, blood pressure, units, abbreviations. The difference shows up on exactly the words the documentation hangs on.
The third is what sits on a clinical desk, and what is allowed to go wrong there. A foot pedal and a dictation microphone are supported and set up on Enterprise, and the content guard rejects any result that adds a word or loses a number. A language model that changes a dose while smoothing the prose is not a comfort problem.
Recognition and formatting run on our own models, locally or in Germany. A processor agreement under Art. 28 GDPR is part of every plan. There is no third vendor that gets to see your history text.
Smart Dictation knows doses, lab values, blood pressure, units and the abbreviations that appear in a discharge letter. A general model across 100+ languages is a different thing.
MediSync Local runs on Windows 10 from 1809, on Windows 11 and on macOS with Apple Silicon. A German hospital desk is mostly Windows.
The content guard rejects any result that adds a word or loses a number, and the raw text stands. A freely chosen language model comes with no such check.
The devices already on the clinical desk are supported, and on Enterprise we set them up.
Campus licences, licence keys for machines with no internet, a private tenant or fully local operation, plus the processor agreement, the technical measures and the operating concept.
MediSync Scribe records the conversation and makes a note, MediSync Connect hangs it off the PVS or KIS, and if you build your own, the same models are an API.
| Feature | MediSync | Superwhisper |
|---|---|---|
| Whose models run | Our own, trained on German clinical languageThe legal frame hangs off that as much as the accuracy does. | Whisper for recognition, plus third-party language models such as GPT-5 |
| Price per user | €10 net a month, €8.33 net yearly | $8.49 a month, with yearly and lifetime options |
| Free plan | Permanent, unlimited MediSync Local, 100 cloud minutes | Permanent, unlimited Whisper models; a 3,000-word Pro trial |
| Works offline | Yes, on Windows and on Apple Silicon | Yes; the vendor says local models only run really well on Apple Silicon |
| Where the text goes to be tidied | The machine, or our cloud in Germany | The chosen language model, optionally through your own API key |
| Legal frame for a practice | Processor agreement under Art. 28 GDPR on every plan, § 203 accounted for | Not published |
| Certifications | Runs on ISO/IEC 27001-certified infrastructure | SOC 2 Type II, listed on the Enterprise plan |
| German clinical vocabulary | Smart Dictation is trained for it | More than 100 languages, generally |
| Foot pedal and dictation microphone | Yes, set up on Enterprise | Not published |
| Invented words | The content guard throws the result away | Not published |
| Operating systems | Windows and macOS on Apple Silicon | macOS, Windows, iOS and Android |
Every price is in the currency the vendor publishes it in. We do not convert dollars into euro: the rate moves, the table does not.
Patient data is involved and your organisation needs a processor agreement; German clinical terms have to come out spelled right; local operation has to work on Windows as well; a foot pedal or dictation microphone is on the desk; or tidying the text must never lose a number.
In the model and in the contract. Superwhisper recognises with Whisper, a general model, and tidies the text with a language model of your choosing. With us both steps are our own models, trained on German clinical language, and cloud operation comes with a processor agreement under Art. 28 GDPR. Our local mode also runs on Windows.
Technically you can with Superwhisper. Under data-protection law, every recipient of patient data needs a legal basis and a processor agreement. A key a doctor enters themselves brings neither. With us there is no third recipient.
The prices are in different currencies and we do not convert them: €10 net against $8.49 per user a month, both with a free plan underneath. At this level price is not what decides it.
Yes, Whisper covers more than 100 languages, German among them. What is missing is the clinical part: doses, lab values and the abbreviations a discharge letter is made of. That is exactly what Smart Dictation is trained on.
With us yes, on Windows 10 from 1809 and Windows 11, and equally on macOS with Apple Silicon. Superwhisper itself notes that its local models only run really well on Apple Silicon and that Intel Macs work best with cloud models.
Every figure about the other vendor comes from their own page or from a reseller's. Next to each one is the date we read it.