Super Voice Mode processes audio, dictation, meeting transcripts, and meeting notes locally on your Mac by default. We do not require an account, and the app sends no audio, transcripts, feature usage, or behavioral telemetry to us. Standard download, model-catalog, and update requests do reach our delivery server; the limited request analytics we derive from them are described below. Content is sent to an external provider when you choose that provider for the voice assistant, AI Correction, or AI Writing.
~/.config/voicemode/. The app does not upload your history library; individual text requests are sent to Claude or ChatGPT only when you select one of their models.Vocabulary is on by default. While a dictation is running, it reads a bounded amount of text around the insertion point and visible text in the focused window through macOS Accessibility, so recognition can match names and technical terms that are already on screen. The captured text is used in memory for the current dictation and then discarded; only a term that is actually recognized in your speech becomes part of the dictated output. The captured text itself is never sent anywhere, never written to a log, and never saved to disk. The app logs which app the text came from and how many terms it found, not the text. It also keeps a small local diagnostics file of counts, such as how many terms were considered and applied in each dictation and how long each step took; it contains no transcript text, no on-screen text, and no term names. Secure text fields, elements macOS marks as protected content, password managers and authenticator apps, and windows whose titles indicate secrets are skipped entirely. You can turn this off with the Vocabulary switch in the app.
Meetings are saved as ordinary local files in your selected vault. A meeting folder can include structured meeting metadata, generated notes, a transcript, meeting-local speaker labels, user-entered names, and any meeting-window snapshots you manually capture. These files are not uploaded by Super Voice Mode.
Release builds do not create voiceprints, biometric templates, or cross-meeting speaker identity records. Speaker labels are local review aids for a meeting file, not a system for recognizing people elsewhere.
Depending on where you and your meeting participants are located, recording, transcription, diarization, labeling, or other meeting-processing features may be regulated. You are responsible for deciding whether you can record, transcribe, summarize, or label a meeting and for obtaining any required consent.
If you enable the voice assistant feature with an external backend (Claude, Codex, or Ollama), your transcribed text is sent to that service according to its own privacy policy. This is opt-in and clearly indicated in settings. The local LLM assistant backend sends nothing externally.
Text-to-speech processing runs entirely on your device. Generated audio is played locally and is not stored, transmitted, or logged. Custom voice data and character files stored at ~/.config/voicemode/characters/ remain on your machine and are never uploaded. See our Terms of Service for usage restrictions on generated audio.
Super Voice Mode uses Sparkle for update checks. The app also reads a signed model catalog from the same delivery server. Like any web request, these requests expose standard connection metadata to the server, including IP address, coarse country, user agent, and app version. We do not store the raw IP address. We store a salted, truncated hash for deduplication for up to 30 days and retain aggregate counts for up to 120 days. This lets us estimate active app versions and countries without adding an app telemetry ping or device identifier. You can disable automatic Sparkle update checks in the app; model-catalog requests occur when the app checks or downloads supported models.
Payments are processed by our payment provider. We do not store credit card numbers or payment details. The payment provider's privacy policy governs that transaction. We receive your email address and license key for the purpose of license activation and support.
When you activate a license, the app sends your license key and a hashed machine identifier to validate the license. The machine identifier is a one-way hash; it does not compute to anything that could identify your hardware. No other data is sent.
For paying customers, we store:
We also store the short-lived, pseudonymous delivery-request records described above. We don't store audio, transcriptions, meeting content, dictated text, or in-app feature usage.
We do not sell your personal data. We may share limited data with service providers who help us operate and improve the app, such as:
These providers are contractually required to use your data only for the services we engage them to provide.
You can request deletion of your data (email and license records) at any time by contacting us. Since the app stores everything locally, uninstalling it and deleting ~/.config/voicemode/ removes all local data.
Super Voice Mode is not directed at children under 13. We do not knowingly collect personal information from children.
We may update this policy from time to time. We will post the updated policy on our website with a revised date. For material changes that expand data collection, we will provide notice through the app or on our website before the change takes effect. Your continued use of the app after such notice constitutes acceptance of the updated policy.
Questions about this policy: privacy@voicemode.io