Privacy · By construction
Privacy Policy
VisionSpeech transcribes your audio on your Mac, iPhone, or Apple Vision Pro. Audio is not uploaded for transcription. Transcript history stays local by default and, on iPhone and Vision Pro, syncs through your iCloud account only if you explicitly enable iCloud Sync. We do not run servers that see your data, have user accounts, or collect analytics.
The short version
What the app does with your speech
When you record, audio is captured by the microphone and transcription runs on the device itself, using the model you choose: NVIDIA Parakeet v3 (via FluidAudio) or OpenAI Whisper Large v3 (via WhisperKit). The audio is not uploaded by VisionSpeech. It is discarded after transcription unless you explicitly enable recording retention on Mac. Retained recordings remain in VisionSpeech's local Application Support directory and can be removed from the app.
Delivery depends on the platform: Vision Pro copies the result for pasting, the iPhone keyboard inserts it, and Mac can insert it into the active app. What you do with the text afterward is governed by the destination app and system.
The optional AI cleanup feature (removing filler words, fixing punctuation, reformatting) uses on-device Apple Intelligence where available, or a separately downloaded local model on Mac. The transcript being cleaned up is processed on the device and is not sent to VisionSpeech servers.
What is stored on your device
VisionSpeech keeps a local history of your most recent transcripts (up to 100, plus any you pin) so you can revisit, edit, or copy them again. You can delete an item or clear the entire local history from the app.
Transcript history, My Words, Text Replacements, and snippets are local by default. On iPhone and Vision Pro, if you turn on iCloud Sync in Settings, their contents are mirrored through Apple's iCloud Key-Value Storage so they can appear on your devices. VisionSpeech cannot access that data through a developer server. Turning sync off stops new syncing but does not delete data already stored in your iCloud account. The current Mac distribution keeps transcript and writing data local.
Model downloads
During setup, the app downloads the speech model you choose (roughly 500 to 625 MB) so transcription can run offline afterward. Mac users can also choose to download a larger local model for cleanup and autocomplete. Models are fetched from their model host (FluidAudio or WhisperKit, via Hugging Face). As with any download, the host can see standard network request information such as your IP address; we do not control or receive that information, and it is not tied to any account or identity within VisionSpeech. Installed on-device features work offline. Optional iCloud Sync naturally requires an iCloud connection.
Permissions we ask for
- Microphone. Required to record the speech you want transcribed. Audio is used only for transcription on the device, as described above.
- Accessibility (Mac). Used to read the focused text field and insert dictation or autocomplete text when you enable those features.
- Screen Recording (Mac, optional). Used only when you enable screen-context assistance for autocomplete. Captured pixels and recognized context remain in memory.
We do not request contacts, location, photos, or advertising identifiers.
Data we collect and share
None. VisionSpeech contains no analytics, advertising, or tracking SDKs. We do not collect personal data, we do not build a profile of you, and we have nothing to sell or share. If enabled, transcript sync is provided by your iCloud account as described above; VisionSpeech does not operate a sync server. The app's privacy practices are also declared in its App Store privacy details and in the bundled privacy manifest.
Children's privacy
VisionSpeech is a general utility app and is not directed at children. It does not knowingly collect any personal information from anyone, including children.
Changes to this policy
If this policy changes, we will update this page and revise the effective date above.
Contact
Questions about privacy? Email support@visionspeech.app. See also the support page.