Quoth — Privacy Policy — Bit 0100 Bit Software

What we collect

Nothing. Quoth has no account system, no backend service, no telemetry, no analytics, no crash reporting, and no third-party SDKs that collect user data. Your voice and what you dictate stay on your Mac.

What is processed on your Mac

Audio is captured to a 60-second in-memory ring buffer while you're dictating, reset on every new dictation, and never written to disk.
Speech-to-text runs on-device — Apple SpeechAnalyzer by default, or WhisperKit as an optional fallback.
AI polish runs on-device via Apple FoundationModels (Apple Intelligence) when available, with a deterministic rule-based fallback otherwise. No prompt or output leaves your Mac.
Polished text is inserted into the active app via the system pasteboard, or the macOS Accessibility / Apple Events APIs.
Imported audio or video files are transcribed entirely on this Mac using the same on-device engines — the file is read locally and never uploaded.

Stored on your Mac only

Your personal dictionary, snippet substitutions, and per-app shape overrides.
App preferences (dictation shortcut, microphone, HUD opacity, speech engine, and similar) in the app's UserDefaults.
The menu bar shows the last five outputs in memory only — never written to disk, gone when the app quits.
Optional saved history (off by default). If you turn on “Save dictation history,” each dictation's transcript and polished output are stored in a local database encrypted at rest with AES-GCM, using a key kept in your macOS Keychain (device-only, never synced). It auto-deletes on a schedule you choose (7 / 30 / 90 days, or keep forever), and you can search, export, or clear it any time. Nothing is saved unless you enable it.
Aggregate usage counts for the Stats tab (number of dictations, word totals, time) — plain numbers with no transcript content.
All on-disk data lives in ~/Library/Application Support/Quoth/ and is removed when you uninstall the app or delete that folder.

When Quoth uses the network

One-time speech-model download: macOS may fetch the Apple SpeechAnalyzer model the first time you use the Apple engine; choosing WhisperKit downloads the selected model once from Hugging Face (huggingface.co) and caches it locally.
A one-time Gatekeeper notarization check with Apple on first launch — no account, receipt, or usage data is exchanged.
bit 0100 bit Software LLC does not see, receive, or process any of these requests.

Permissions Quoth requests

Microphone — capture your voice while you're dictating.
Speech Recognition — run Apple SpeechAnalyzer for transcription.
Accessibility — insert polished text into the app you're typing in.
Contacts (optional) — only if you choose to import names and companies into your personal dictionary. Contacts are read on this Mac and never uploaded.

Review or revoke these any time in System Settings → Privacy & Security.

Contact

Questions about this policy? Email support@bit0100bitsoftware.com.