The short version
Earshot processes audio entirely on your iPhone. We never collect or transmit your speech, your transcripts, or anything else that identifies you. Anything the app remembers — your transcripts, settings, conversation history — stays on your device. No accounts, no analytics, no ads.
What we process and where
- Microphone audio is captured only while you are on a Live screen. It is fed directly to an on-device transcription model (OpenAI Whisper Small, running locally via the WhisperKit framework), an on-device speaker-separation model (pyannote, used in group mode), and an on-device sound classifier (Apple SoundAnalysis). Raw audio buffers are processed and discarded — they are never written to disk.
- Transcripts are saved locally on your device using Apple's SwiftData framework. They show up under Settings → Transcripts and in History, and you can delete them at any time by swiping left on a transcript in the History list. Transcripts never leave your device unless you explicitly share one via the iOS share sheet — which sends plain text to the app you choose (Notes, Mail, Messages, etc.).
- Sound detections (e.g., smoke alarm) are stored with the transcript they occurred during, only for display in that conversation's summary.
- Your settings (palette, text size, armed sounds, etc.) are stored locally in iOS UserDefaults. They stay on your device.
Network use
Earshot uses the network only to download its on-device models — the speech model (OpenAI Whisper) and the speaker-separation (diarization) models — from Hugging Face (huggingface.co). That happens on first launch, and again only if you switch transcription models or install an app update that ships newer models. Those downloads only ever come to your device — your audio and transcripts are never uploaded. Once the models are on your device, the app works fully offline. We do not contact any analytics, advertising, or telemetry servers — there are none.
What we do not do
- No accounts or sign-in.
- No advertising, no advertising identifiers, no cross-app tracking.
- No analytics SDKs, no crash-reporting SDKs.
- No selling, sharing, or otherwise transferring personal information to third parties — we have none to share.
Permissions
Microphone access is required for transcription and sound recognition. You can grant, revoke, or check it at any time in iOS Settings → Privacy & Security → Microphone → Earshot.
Children
Earshot is suitable for general audiences. Because we do not collect personal information from anyone, we do not knowingly collect personal information from children under 13 in the United States, or equivalent ages in other jurisdictions.
Changes
If we ever change how the app handles data, this page and the in-app policy will be updated together, and the effective date above will change.
Contact
Questions, requests, or concerns: [email protected]