PRIVACY POLICY

This is the plain-English version of how Voquence handles your data. It says exactly what leaves your Mac, what stays on it, and who can see it. If anything is unclear, email support@voquence.com and Jason will answer directly.

• Local Whisper mode: nothing leaves your Mac. Your voice is transcribed on-device. Voquence the company never sees your audio or your transcript.
• Cloud mode (Bring Your Own Keys):your audio goes from your Mac to OpenAI's Whisper API using your OpenAI key. Your transcript goes to Anthropic's Claude using your Anthropic key. Voquence the company never sees either.
• Managed Cloud (subscription):transcription still happens however you've set it (Local Whisper stays fully on-device). For the polish modes, your transcript is sent to Voquence's proxy, which forwards it to Anthropic's Claude using Voquence's key and returns the result. The proxy does not store your transcripts or outputs; it records only a usage count (how many polishings, not what they said).
• API keys: currently stored on your Mac in a local JSON config file. Migration to the macOS Keychain is upcoming (v0.3.x). Either way, your keys live only on your machine.
• App usage data:the Voquence app collects nothing. No analytics, no telemetry, no “phone home” pings.
• Website (voquence.com): uses Google Analytics 4 for page views and download counts. No personal data is collected by us beyond what GA4 standardly sees (anonymous browser/region info).
• Email signups: if you enter your email in our wishlist form, we store it in Resend to send you launch announcements. You can ask us to delete it anytime.

Voquence is built and maintained by Jason Regino, a solo indie developer in California. There is no corporate parent. There is no data team. When you email support@voquence.com a human reads it.

Voquence runs in two modes. Choose either, switch any time, in the Settings panel.

Local Whisper (recommended, fully offline)

When you turn on Local Whisper in Settings, Voquence downloads OpenAI's open-source Whisper model to your Mac (about 141 MB, one-time). After that, every voice recording is transcribed on-device using that model. The audio is captured by your microphone, held briefly in memory, passed to whisper.cpp, and discarded. None of it leaves your Mac. Voquence the company has no servers in this path and no way to see what you said.

Cloud (Bring Your Own Keys)

When Local Whisper is off, Voquence sends your recorded audio directly from your Mac to OpenAI's Whisper API, using the OpenAI API key you pasted into Settings. The resulting transcript is then sent directly from your Mac to Anthropic's Claude API (using your Anthropic key) so it can be processed by the mode you picked: Clean Text, Tweet Thread, Book Description, and so on. The polished result comes back and is pasted at your cursor.

In BYOK cloud mode, OpenAI and Anthropic see your audio and your transcript under the terms of their own privacy policies and data retention rules (linked below). In this mode Voquence the company is not on the wire. Your keys give you a direct line to those providers; we are not a proxy.

Managed Cloud (subscription)

If you subscribe to Managed Cloud, you skip the keys entirely. Your transcription still runs wherever you've set it (Local Whisper stays fully on-device). When a polish mode runs, your transcript is sent from your Mac to Voquence's proxy at voquence.com, which forwards it to Anthropic's Claude using Voquence's own key and returns the polished result. The proxy is a pass-through: it does not store, log, or read your transcripts or outputs. The only thing it records is a usage counter (how many polishings you've run this month, never their content) so we can enforce the plan's monthly allowance.

As of v0.4 (current), your OpenAI and Anthropic API keys are stored on your Mac in a local JSON file at ~/Library/Application Support/voquence/config.json. The file is readable only by your macOS user account by default. On a FileVault-encrypted Mac (the macOS default since Big Sur), it's encrypted at rest.

We are migrating API key storage to the macOS Keychain in an upcoming release. The Keychain is the correct long-term home for credentials on Mac. When that ships, the migration will run once on update: your keys move into Keychain, and the JSON file is cleared.

You can avoid writing keys to disk entirely by setting OPENAI_API_KEY and ANTHROPIC_API_KEY as environment variables. Voquence reads them on launch if the config file is empty.

Voquence asks for three macOS permissions during onboarding. Each does exactly one thing.

• Microphone: captures your voice during recording. Voquence only listens between the moment you press the hotkey and the moment you press it again. The mic is off at all other times.
• Accessibility:lets Voquence paste its output at your cursor in any app. Apple's Accessibility API is the only sanctioned way to do this on Mac. Voquence uses it to senda paste keystroke. It does not read other apps' text, watch your typing, or take screenshots.
• Apple Events: required for the Cmd+V paste simulation to work in some apps. Voquence triggers a single keystroke event when a transcription completes. Nothing else.

Voquence does not take screenshots of your screen, read the contents of other apps' windows, log your keystrokes, or watch your clipboard. (For comparison: another Mac dictation app was found taking periodic screenshots of the active window for AI “context.” Voquence does nothing of the kind, by design.)

The Voquence app collects nothing. The website (voquence.com) uses Google Analytics 4 to count page views, download clicks, and referrers. This helps Jason know whether his marketing is working. GA4 sees what it standardly sees: anonymous browser and region information, no personal identifiers.

If you enter your email address into the wishlist form on the homepage, we store it in our Resend account to send you launch announcements. We do not share or sell email addresses. Email support@voquence.com and we'll delete yours within a day.

voquence.com is hosted on Vercel. Vercel logs HTTP requests at the edge as part of normal site operation. Those logs are retained per Vercel's standard policy and we do not access them except for debugging.

• OpenAI: only if you turn on cloud mode and add your OpenAI API key. Audio is sent directly from your Mac to OpenAI using your key. See OpenAI's privacy policy.
• Anthropic: only for the content modes that call Claude, and only with your Anthropic key. See Anthropic's privacy policy.
• Google Analytics 4: on voquence.com only. Page views and download counts.
• Resend: only if you enter your email in the wishlist form. Used for launch announcements.
• Stripe: processes Founding License and Managed Cloud payments. Stripe sees your payment details (card number, billing address) under Stripe's privacy policy. Voquence never sees or stores your card number.
• Vercel: hosts the voquence.com website and serves the auto-updater endpoint.
• Hugging Face: hosts the Whisper model file. When Local Whisper mode downloads the model the first time, it pulls the 141 MB file from huggingface.co. No personal data is sent.
• Supabase:stores your Managed Cloud account login (email and encrypted password or OAuth identity) and your subscription and usage status. Supabase sees your email and account state. They do not see your dictation, transcripts, or polished outputs, which flow through Voquence's own proxy with our Anthropic key only while you are actively using Managed Cloud. Free tier users and BYOK users do not create a Supabase account. See Supabase's privacy policy.

Voquence is not directed at children under 13 and we do not knowingly collect personal information from them.

You can ask us to delete any data we hold about you (mainly: wishlist email address) at any time by emailing support@voquence.com. We'll confirm deletion within a few business days.

If you're in the EU/UK, you have rights under GDPR/UK GDPR including the right to access, correct, delete, and port your data. Email support@voquence.com to exercise any of these.

We'll update this page if how Voquence handles your data changes. The “last updated” date at the top will move, and we'll note significant changes in the changelog and in release notes inside the app.

Email support@voquence.com. A real human (Jason) reads every message. Replies come back as “Jason from Voquence” from the same address.