Murmurtale The short version
Murmurtale is a macOS dictation app for writers. Our design goal is that your words stay yours.
- Free (on-device) tier: nothing you dictate leaves your Mac. Audio and text are captured, transcribed, and cleaned up entirely on your device. We can’t see them because they never reach us. You can dictate up to 10,000 words with no account; a free account then gives you 10,000 local words every month, at no cost — we count your words, we never see them.
- Pro (cloud) tier: our servers store nothing. When you choose a cloud engine, your audio and text are transmitted over an encrypted connection, processed in memory, and not stored on our servers — and we keep no content logs. Our cloud subprocessors process your content transiently; a subprocessor may retain it briefly — up to 30 days — for trust-and-safety monitoring before automatic deletion, and never uses it to train models.
- Accounts sign in with Apple, Google, or email. We collect the minimum account data needed for identity, billing, and your word count (an email address or sign-in identifier, a monthly word count, and, for Pro, billing status). We do not sell your data or use your dictation content to train any model.
- If crash/error diagnostics are enabled, Murmurtale sends only operational metadata such as app version, OS/device class, pipeline stage, and sanitized error codes — never dictated text, audio, prompts, dictionary terms, manuscripts, or account secrets.
- Murmurtale is AI-powered. Speech-to-text and cleanup are performed by AI models — on your device by default, and in Pro cloud mode by the subprocessors listed in section 2. AI here only transcribes and tidies your words: we make no automated decisions about you with legal or similarly significant effects, and your content never trains any model (ours or a vendor’s).
1. What Murmurtale does with your voice
Free / on-device mode (the default)
Everything happens locally on your Mac:
- The microphone is accessed only while you hold the dictation key (fn or right ⌥) or while continuous mode is armed.
- Audio is written to a local buffer, transcribed by an on-device speech model, and cleaned up by on-device rules and an optional on-device language model.
- No audio and no text are transmitted to us or to any third party. Your dictation itself is fully usable offline on any plan; the first 10,000 words need no account or connection at all. Beyond that, creating a free account and syncing your monthly word count needs a brief connection, but the words you dictate never leave your Mac.
- Learning “your world” happens on your Mac. Murmurtale can suggest dictionary terms it notices you dictating often, and it learns from your corrections — all entirely on-device over your local data. It is not keystroke logging, and nothing is sent anywhere to produce these suggestions. You can set this to Suggest / Auto-add / Off in Preferences.
Pro / cloud mode (optional, opt-in)
If you turn on a cloud engine (cloud transcription and/or cloud cleanup):
- Your audio and/or transcript are sent over an encrypted connection (TLS 1.3) to the Murmurtale backend, which forwards them to our processing subprocessors.
- On the Murmurtale backend, content is processed in memory only and is never logged or persisted; we store no dictation content on our servers and do not use it for analytics or model training. Our subprocessors process your content transiently; a subprocessor may retain it briefly — up to 30 days — for trust-and-safety monitoring before automatic deletion, and never uses it to train models.
- Our backend keeps only minimal non-content operational metadata (timestamps, request/byte counts, error codes, account ID and tier) to run the service, meter usage for billing, and debug outages.
Crash and error diagnostics
Murmurtale can send privacy-safe diagnostics for crashes and handled dictation-pipeline failures: app/build version, OS/device class, account tier, pipeline stage, sanitized error codes, and small stack fingerprints. Diagnostics never include dictated text, audio, prompts, style excerpts, dictionary terms, manuscript content, email addresses, or payment/account secrets. You can turn diagnostics off in Preferences.
2. Third-party subprocessors (cloud mode only)
Our requirements are the same for any subprocessor, and they are non-negotiable: a subprocessor may process your content only transiently to return the requested result, may retain it only briefly (up to 30 days) for trust-and-safety monitoring, never longer, and may never use it to train models. The table below is the current roster; if it ever changes, the replacement is held to the same requirements and we update this page.
When you use Pro cloud features, we route requests to:
| Subprocessor | Purpose | Data handling |
|---|---|---|
| Anthropic (Claude) | Cloud cleanup / editing pass | Content is processed transiently and never used to train models. Anthropic may retain content briefly — up to 30 days — for trust-and-safety monitoring before automatic deletion. |
| AssemblyAI | Cloud speech-to-text | Murmurtale actively deletes each transcript from AssemblyAI the moment your result is returned. Uploaded audio is processed transiently; AssemblyAI may retain content briefly — up to 30 days — for trust-and-safety monitoring before automatic deletion, and never uses it to train models. |
These vendors are engaged strictly to process a single request and return a result. We update this list if subprocessors change.
3. What is stored on your device (and your iCloud)
The following live on your Mac, under your control, and are never uploaded to Murmurtale’s servers: dictionaries, History of your dictations (auto-pruned, default ~30 days), recent audio buffers (for retry/playback), and settings.
If you use Apple’s iCloud Drive / Desktop & Documents sync, these files may be synced by macOS within your own iCloud account under Apple’s terms — that is your Apple account, not Murmurtale’s servers. You can disable iCloud sync in macOS System Settings.
4. Microphone use
Murmurtale requests macOS Microphone permission to capture dictation; the mic is active only during a hold-to-talk press or armed continuous session. It also requests Accessibility permission to insert text into other apps and detect the global dictation hotkey; that permission is used only for those functions.
5. Account & billing data
You can dictate up to 10,000 words with no account. Beyond that, an account is required to continue — sign in with Apple, Google, or email. For every account we process your email address or sign-in identifier, a monthly word count (a tally, never your dictated words, used to enforce the Free tier's 10,000-words-a-month allowance and reset it monthly), and your subscription status / entitlement. If you subscribe to Pro, we additionally process payment via Stripe and your cloud word usage against your plan's allowance, tracked the same way as a count, not content. We do not receive or store your full payment card details. We use this data only to provide and bill the service, and we do not sell it.
6. What we do NOT do
- We do not sell or rent your personal data.
- We do not use your dictation content to train AI models (ours or vendors’).
- We do not retain your dictation audio or text on our servers.
- We do not run advertising or content-based profiling.
7. Data retention summary
| Data | Where | Retention |
|---|---|---|
| Dictation audio/text (free) | Your Mac only | Local; auto-delete Off, 1, 7, or 30 days (default); delete or clear anytime |
| Dictation audio/text (Pro cloud) | In-memory at backend/vendors during the request | Not stored by us; a subprocessor may retain it up to 30 days for trust-and-safety monitoring before automatic deletion, never for training |
| Operational metadata (non-content) | Backend | Minimal, retained only as long as needed for ops/billing |
| Crash/error diagnostics (non-content) | Backend | Minimal, retained only as long as needed to debug reliability |
| Account email/identifier + tier | Backend / auth provider | Duration of your account |
| Monthly word count (local and, on Pro, cloud) | Backend | A running count only, never dictation content; resets monthly |
| Payment data | Payment processor | Per processor’s policy |
8. Your rights & choices
- Dictate fully offline — on any plan, your words never leave your device; only a monthly word count is synced to your account, and the first 10,000 words need no account or connection at all.
- Delete local data — delete individual History items or clear all History in-app; each action also removes associated local audio. You can choose automatic deletion after 1, 7, or 30 days, or turn it Off.
- Access / deletion of account data — because we hold no dictation content, account-data requests concern only your email/subscription record. Use the routes on the Contact page.
- Depending on your location, you may have rights under GDPR/UK GDPR, CCPA/CPRA, or similar laws.
9. Children
Murmurtale is not directed to children under 13. We do not knowingly collect data from them.
10. Security
We use TLS 1.3 for data in transit and follow least-data principles (we can’t leak content we never store). No method of transmission or storage is 100% secure; we work to protect the minimal data we do hold. See the Security page for current controls and responsible disclosure.
11. Changes
We may update this policy; we’ll revise the version and date and, for material changes, notify Pro users by email.
12. Contact
Brad Pierce Enterprises LLC, 1399 Kimblewick Rd, Potomac, MD 20854, USA · legal@officialbradpierce.com — privacy contact routes are also on the Contact page.