Help centre

Any stable position facing oncoming or passing traffic works. A few practical notes: • Window mount (interior, facing street): the most common setup. Clean the glass, avoid reflections. Keep the phone portrait or landscape — the app handles any rotation. • Dashboard: works for roads in front of you. Avoid dashboard glare on the windshield. • Tripod (exterior, pointing at a cross-street): best accuracy because the camera doesn't move. Use a phone mount rated for outdoor use. • Height: roughly 1–2 metres above road level at a slight downward angle. This gives the vanishing-point estimator enough perspective to converge quickly. • Angle to road: anywhere from 15° to nearly perpendicular is fine. The homography solver compensates. Avoid: airbag deployment zones, pointed directly at the sun, or any position where the phone vibrates continuously (engine vibration confuses the tilt sensor).

SpeedCam AI offers two calibration paths: Automatic (default): The vanishing-point solver collects contact-point data from passing vehicles. At the Standard level, expect roughly 15 vehicles before the calibration indicator turns green. This typically takes 3–8 minutes on a moderately busy street. Manual 2-point line (faster): Open the Tool Palette (upper-left), tap the line icon, and drag two endpoints onto any road feature of known real-world length — a painted crosswalk stripe, a lane-width marker, or parking bay lines work well. Enter the distance in metres or feet and tap Done. The app immediately uses this ground-truth reference. If auto-calibration stalls: • Make sure the camera is stable — brief phone movement resets the VP solver. • Vehicles need to be in motion. Parked cars do not contribute contact-point data. • Switch to the manual line if fewer than ~5 vehicles pass per minute. • Tap the calibration indicator and choose "Start Fresh" to discard a bad calibration.

Camera-based speed estimation is not radar-gun accuracy and does not pretend to be. SpeedCam AI is designed for trend detection and comparative studies — "how fast is traffic on this street this week versus last month," "does the new speed bump actually slow cars down" — not for absolute ticket-grade measurement. Under good conditions (Standard or High calibration level, stable mount, a moderate camera angle, consistent lighting) the app's readings are directionally reliable for traffic studies. Accuracy degrades with camera vibration, very shallow angles under 10° from road level, poor lighting, edge-of-frame passages, or use before calibration has converged. Two quality gates built in: • Two independent estimators — closed-form velocity and homography projection — must agree within 20% before a reading displays. • A 30-frame rolling median filters per-frame outliers. Readings below 5 mph are suppressed (parked-vehicle noise). Use SpeedCam AI for what it's good at: advocacy-grade data, before/after studies, per-session distributions, and flagging outlier events for context. Do not use readings as sole evidence in formal legal or enforcement proceedings. This is not certified calibration equipment.

Five export formats are available from the Incidents tab or Analytics screen: • PDF — formatted report with summary statistics, speed distribution chart, and incident thumbnails. • JSON — raw incident and observation records, machine-readable. • CSV — spreadsheet-compatible; one row per vehicle observation with speed, class, confidence, timestamp, and GPS coordinates. • TMAS — Traffic Monitoring Analysis System format, compatible with FHWA traffic-study tooling. • (A fifth format, reserved for future release, appears greyed out.) Exports are filtered by whatever campaign, session, or confidence level you have selected. Files are shared via the iOS share sheet — AirDrop, Mail, Files, and any other extension you have installed. No data is uploaded to a server during export. Files are generated entirely on-device.

Very little, and none of it is footage. On-device only (never transmitted): • Camera frames • Incident images and burst stills • Bounding-box coordinates • GPS coordinates of incidents • Chat queries to "Ask About Your Data" Transmitted to Firebase (aggregate counts only): • Anonymous session-started events • Incident-recorded counts (no images, no GPS) • Crash reports via Crashlytics (stack traces only) • App performance metrics • Subscription status via RevenueCat (no personal payment data — that stays with Apple) Firebase App Attest hardware-attests the app without identifying you. No Firestore or Cloud Storage is used; the Firestore and Storage rules files in the repo are unused scaffolding for a potential future feature. Full details are in the Privacy Policy.

Tap the Tool Palette (upper-left radial menu on the Watch tab), then tap the video import icon. The iOS photo picker opens — select any video from your library, including security-camera exports. Next, choose a camera FOV preset matching the camera that recorded the video: • Nest: 130° • Ring: 140° • Wyze: 110° • Arlo: 130° • iPhone: 77° • GoPro: 170° If your camera is not listed, choose the closest match or select Custom and enter the horizontal field of view. The app processes every frame at the video's native framerate (native encoding is preserved — no re-transcoding). All vehicles travelling ≥5 mph are recorded as a speed survey, not just threshold violators. Results appear as a session in the Incidents tab tagged "Video Import."

SpeedCam AI supports RTSP streaming from Reolink, Amcrest, Tapo, Hikvision, Dahua, Axis, Ubiquiti, and any other RTSP-capable camera. Video decoding uses VideoToolbox on-device — no cloud round-trips. This feature requires the RTSP add-on subscription. Tap "Connect Camera" in the Tool Palette to connect: • ONVIF auto-discovery: the app scans your local network and lists compatible cameras automatically. Tap a camera to connect. • Manual RTSP URL: if your camera doesn't appear, enter the RTSP URL and credentials from your camera's admin panel directly. A 15-second connect timeout applies. H.264 and H.265 streams are both supported.

Incident images are saved to the app's private Documents directory at Documents/SpeedCamEvidence/{YYYY-MM-DD}/. They do not appear in your Photos library unless you explicitly share them. To back up: • Connect to a Mac and use Finder → iPhone → Files → SpeedCam AI to copy the SpeedCamEvidence folder. • Or use the Export function (PDF or JSON) to create a portable record. Clearing data in Settings → Clear All removes all incidents, sessions, and evidence files permanently. This action cannot be undone.

SpeedCam AI requires iPhone 12 or newer running iOS 26. The on-device object detector runs on the Apple Neural Engine. Some advanced analytics views require iOS 26.

Pricing and subscription details are listed on the App Store product page. Subscriptions are managed through your Apple ID. To cancel, go to Settings → Apple ID → Subscriptions on your iPhone.

SpeedCam AI measures and records vehicle speed from your own property or a public right-of-way, using your iPhone camera. This is generally lawful for personal, research, and advocacy purposes in most jurisdictions — the same legal framework that applies to filming from a public space. Laws vary by location, particularly around data retention, sharing footage of licence plates, and using data in formal proceedings. If you intend to share data publicly or present it to authorities, consult local laws. SpeedCam AI does not provide legal advice.