Is Vocateca private? Here's exactly what leaves your Mac.
Vocateca transcribes and diarizes entirely on your Mac — your audio and transcripts never touch a Vocateca server, in Free or Pro. The only network requests it makes are the ones required to fetch the content you subscribed to, check for app updates, and — only if you sign in or buy Pro — talk to the account backend. Free needs no account at all.
Why audio privacy matters
A voice recording is biometric data — it identifies you the way a fingerprint does — and transcripts of interviews, therapy sessions, or closed-door meetings often carry things nobody involved agreed to hand to a third party. Vocateca is built so that sensitivity never has to leave the room: transcription happens on the same machine the recording was made on.
What runs locally
Everything that touches the actual audio and its content stays on-device.
Transcription
Parakeet-TDT (FluidAudio, MLX) is the default engine, with WhisperKit (Whisper, MLX) available as a fallback — both run as on-device Core ML / MLX models on Apple Silicon, no server round-trip.
Speaker diarization
Who-said-what segmentation runs on-device via FluidAudio's diarization models.
Storage
Transcripts, episode metadata, and settings live in a local SQLite database at ~/Library/Application Support/Vocateca/state.sqlite; transcript files themselves are plain Markdown, SRT, TXT, HTML, or OKF on disk.
Search
Full-text search runs against that local database. Nothing is indexed or queried remotely.
Exactly what leaves your Mac
Every network request Vocateca makes falls into one of these five categories.
1. The content you subscribed to
Whenever a show refreshes, or you add or search for one.Podcast RSS feeds are fetched directly. YouTube audio/metadata is fetched via the bundled yt-dlp; Instagram via the bundled gallery-dl. Podcast search queries Apple's iTunes Search API. A Spotify link is resolved by reading Spotify's public embed page, then the underlying audio is fetched from its actual source. Every one of these requests goes to the platform the content actually lives on — never to a Vocateca server. This is inherent to fetching what you chose to follow.
2. App update checks
Automatically in the background (Sparkle's standard behavior), and whenever you click "Check for Updates…".A request to vocateca.com/appcast.xml asks whether a newer, signed build is available. No personal data is sent — only the current version.
3. Tool & model provisioning
Once, the first time a feature needs a component that isn't already bundled.yt-dlp and ffmpeg are downloaded from GitHub, gallery-dl from Codeberg — each download's SHA-256 is checked against a pinned hash before it's trusted. Transcription and diarization model weights are downloaded once from Hugging Face on first use, then run fully offline from then on.
4. Pro account & entitlement
Only when you sign in, complete checkout, refresh your entitlement, cancel, or redeem a voucher — never for Free use.Magic-link sign-in and entitlement checks talk to Vocateca's own backend (the auth., api., and hook. subdomains of vocateca.com); purchases are processed by Mollie through that same backend. Free users never create an account and never contact this backend.
5. Integrations you opt into
Only if you configure a webhook or turn on Notion push — both are off by default.Webhooks send the transcript payload to the exact URL you enter, when a transcription completes. Notion push sends transcript content to api.notion.com using the token you connect. The Obsidian integration writes Markdown files straight into your local vault — no network request at all.
What Vocateca does not do
- No telemetry, analytics, or crash reporting of any kind — verified against the source: none of those SDKs exist in the codebase.
- No audio or transcript is ever uploaded to a Vocateca server, in Free or Pro.
- No account, sign-in, or personal data is required to use Vocateca for free.
Your controls
Privacy dashboard
Settings → Your data shows exactly how much audio and transcript data is stored on disk.
Retention
Set how many days audio and transcripts are kept, independently — 0 means keep forever.
Export & delete
Export your settings and subscriptions, or delete everything Vocateca has stored in one step.
Factory reset
Wipes all on-disk data and the Keychain entries Vocateca created — Instagram session, integration secrets, webhook secrets.
Keychain
Account tokens, webhook secrets, and Instagram session data are stored in the macOS Keychain, never in a plain file.
Log redaction
Copying or exporting the diagnostic log automatically redacts URLs, file paths, and sensitive field values.
This page describes how Vocateca actually behaves, in plain language. For the formal legal privacy policy — data controller, legal basis, your rights under GDPR — see the Datenschutzerklärung.
Frequently asked
Does Vocateca upload my audio or transcripts?
No. Transcription and diarization run entirely on your Mac; audio and transcript content is never sent to a Vocateca server, in Free or Pro.
Do I need an account to use Vocateca?
No. Free requires no sign-in and no account. An account is only created if you choose to sign in or purchase Pro.
What actually leaves my Mac?
Only network requests required to fetch the podcasts or videos you subscribed to (from their own platforms), check for app updates from vocateca.com, download tools/models once on first use, and — only for Pro — talk to Vocateca's account backend. Optional integrations you explicitly configure, like webhooks or Notion push, also send data, only when triggered.
Does Vocateca use analytics or tracking?
No. There is no telemetry, analytics, or crash-reporting SDK anywhere in the app — verified directly against the source code.