Enhance & note drafting
Available onWebmacOSWindowsiOSAndroid
Enhance turns a raw transcript into a structured note (a SOAP note, a letter, a case-conference summary) using a large language model guided by the prompt you pick. The transcript itself is produced on your device; Enhance is the step where a model (cloud or on-device, depending on platform) rewrites it.
How it works
Section titled “How it works”- Transcribe (or import) a consultation.
- Pick a prompt. Each prompt defines the note style and can pin a specific model. See Prompt library.
- Enhance. The note is saved as a new version alongside the original, so you can always get back to the raw transcript.
Enhanced notes are written in structured formatting (headings, bold labels, lists) and render that way in the app.
Drafting routes
Section titled “Drafting routes”Where the language model runs differs by platform. This is the key privacy decision:
- On-device (Apple platforms): with Apple Intelligence, drafting runs on the Mac/iPhone itself. Nothing leaves the device.
- On-device (Windows, Vulkan-capable GPU): a local Qwen3 model runs on your graphics card. Nothing leaves the PC. Scribe0 tests the card first and only offers this if it can actually run the model.
- Sovereign cloud (the default cloud route): an Amazon Bedrock endpoint in AWS Sydney/Melbourne; transcript text is processed in Australia only. This is the default whenever drafting isn’t running on-device.
- OpenRouter (alternative): a general cloud route with a wider model selection, available for those who prefer it.
Available onmacOS · Apple Intelligence or local Qwen3Windows · Vulkan-capable GPUiOS · iOS 26 + Apple Intelligence
Platform specifics
Section titled “Platform specifics”- Press Enhance (Ctrl+E). The note streams into the pane live as it generates, then renders with formatting.
- Three routes: on-device (a local Qwen3 model on your graphics card), Bedrock (sovereign), or OpenRouter. Configure them in Settings → LLM (keys are stored in Windows Credential Manager).
- On-device needs a Vulkan-capable graphics card. Pick it in Settings and Scribe0 downloads the model (about 2.4 GB, once) and runs a short test on your card. If your card cannot run it, Scribe0 says so and you stay on a cloud route rather than getting something slow or unstable.
- The sovereign Bedrock endpoint is the default cloud route.
- A prompt with a pinned model overrides the global model choice.
- Three routes: Apple Intelligence (on-device), Bedrock (sovereign), or OpenRouter. The app prefers on-device when available, and the sovereign Bedrock route is the default cloud option; OpenRouter is an alternative.
- Optional PII redaction masks identifying details before any cloud call. macOS-only.
- Two routes: Apple Intelligence (on-device, needs iOS 26) or Scribe0 Cloud, the sovereign Bedrock endpoint in Australia. There is no OpenRouter route and no model picker on iPhone.
- The cloud route asks your permission before the first transcript leaves the phone, and it can be swapped for on-device on any prompt.
- Cloud enhances run on a free daily allowance with no sign-up; on-device drafting is unlimited.
- Settings ▸ Data and AI lists exactly what is sent, to whom, and what never leaves the phone.
- Cloud only: there is no on-device model on Android, so Enhance always sends the transcript text to the sovereign Bedrock endpoint (Australia). The app asks for consent once, before the first cloud call.
- Works with no sign-up on a free daily allowance; the app shows how many enhances are left. A Scribe0 Cloud key (Settings) lifts the cap.
- Each enhance adds a tab, so you can re-run a transcript with a different prompt and compare drafts.
- Cloud drafting with a choice of models; the most extensive prompt tooling of the apps (snippets, text replacements).