piep

piep is an iOS app for detecting bird calls on device. It uses bundled BirdNET model files through TensorFlow Lite, stores listening sessions locally, can optionally sync session data through iCloud, and can download freely licensed bird photos from Wikimedia Commons for local cache use.

The app is currently a personal, free, non-commercial project.

Features

• Live bird call detection from the microphone
• Overlapping 3 second analysis windows, processed once per second
• On-device BirdNET inference with location/date filtering
• Local listening sessions with timestamp, duration, coordinates, and species
• Optional iCloud Sync for sessions and detections
• Bird overview with search, sorting, day counts, and session references
• Map view with detected species by location
• Wikimedia Commons image loading with local cache and license attribution
• Configurable confidence threshold and expert audio profiles
• Manual benchmark sample recording/import for comparing audio profiles
• Alternate app icons

Privacy

piep is designed to keep user data local:

• Audio is analyzed on device and is not uploaded by the app.
• Location is used on device for BirdNET filtering and session coordinates.
• Sessions and bird detections are stored locally on the device.
• Optional iCloud Sync can synchronize sessions, detections, and coordinates through Apple's CloudKit service in the user's private iCloud database.
• Bird images are fetched from Wikimedia Commons and cached locally.

See PRIVACY.md for the full privacy policy draft.

Requirements

• Xcode 26 or newer
• CocoaPods
• An Apple developer team for device builds or App Store distribution

Setup

Install dependencies:

pod install

Build without a local signing configuration:

./build.sh

Install on a local device:

cp xcbuild.example.conf xcbuild.conf
# edit xcbuild.conf with your local device ID
./build.sh
./deploy.sh

xcbuild.conf is intentionally ignored because it contains local device and signing configuration.

App Store Notes

Review APP_STORE_REVIEW_NOTES.md before submitting to App Review. The most important point is that the bundled BirdNET model resources are licensed for non-commercial use.

Licensing

The app source code is licensed under the MIT License. See LICENSE.

The bundled BirdNET model files and third-party resources have separate license terms. See THIRD_PARTY_NOTICES.md.

Bird photos are loaded at runtime from freely licensed Wikimedia Commons sources. The app displays per-image author, license, and source information.