Speak anywhere
Dictate into notes, chat, docs, forms, or code prompts without changing windows.
macOS 14+ desktop dictation
ScribeFoundry
Offline-first dictation for desktop.
Press a hotkey, speak naturally, and turn speech into clean text anywhere.
RAW
need a crisp project update with blockers and next steps
CLEAN
Project update: blockers are clear, next steps are ready, and the follow-up can go out today.
Project update
Blockers are clear. Next steps are ready.
Follow-up can go out today.
Workflow preview
See the intended dictation loop in 18 seconds.
Hold the hotkey, speak once, let local cleanup shape the text, then insert it into the active desktop app.
Video not loading? Open the workflow preview.
Benefits
Compact, local-first, and built around the cursor.
Local by default
Local dictation runs offline after the selected model is downloaded and cached.
Clean output
Post-processing can tidy punctuation, formatting, custom terms, and voice commands.
Built for workflows
Mode-specific output helps short replies, structured notes, reports, and prompts.
Optional API enhancement
Use OpenAI, Anthropic, or compatible endpoints when you want remote refinement.
How it works
Four small moves, one continuous workflow.
01
Press
Default hotkey: Option+Space on macOS.
02
Speak
A compact capsule shows capture state and waveform feedback.
03
Transform
Local rules and model post-processing prepare readable text.
04
Insert
Text is typed or pasted through the desktop insertion pipeline.
Workflow modes
Useful shapes for spoken work.
Notes
Capture quick thoughts before they disappear.
Chat reply
Turn rough speech into a concise response.
Email draft
Shape spoken intent into a useful first draft.
Coding prompt
Dictate agent requests with cleaner phrasing.
Issue report
Capture steps, expected behavior, and context.
Documentation
Draft explanations while staying in flow.
Translation
Optionally translate output into supported languages.
Meeting notes
Summarize spoken follow-ups after a call.
Privacy model
Local first, explicit when network is involved.
- Local dictation runs offline after the selected model is downloaded and cached.
- First-time model downloads and optional API post-processing require network access.
- Dictation audio is processed locally by default; optional API cleanup only runs when you enable a provider.
- API keys are stored in the OS Keychain and are never revealed in the UI.
- No remote sync or telemetry is part of the documented desktop architecture.
Gallery
Real app surfaces from a clean local profile.
Main dashboard
Recent dictation activity, model readiness, and workflow controls.
Recording capsule
A small desktop overlay for push-to-talk capture.
Settings
Models, input language, output style, API provider, and system controls.
FAQ
Practical answers about local dictation.
Which platforms are supported?
ScribeFoundry targets macOS 14+. Apple Silicon has the best experience; Intel support is partial.
Does dictation work offline?
Yes, for local dictation after the selected local model has been downloaded and cached.
What happens after I release the hotkey?
ScribeFoundry stops capture, transcribes locally with the selected model, applies cleanup or translation settings, then inserts the final text at the active cursor.
Which local models are available?
The app presents simple choices: Fast for smaller downloads and quicker startup, Quality for the default accuracy-focused path, and Robust as an experimental Apple Silicon option for noisy or far-field audio.
Can it translate output?
Yes. You can keep the dictated language or translate the final inserted text into supported output languages including English, Vietnamese, Spanish, French, German, Japanese, Korean, and Simplified Chinese.
Can it use API providers?
Yes. API post-processing is optional and uses the provider settings you enable.