FAQ

SafeOS Guardian is a free, open-source monitoring tool that runs entirely in your browser. It uses your device's camera and microphone to detect motion and sound, alerting you when activity is detected. It's designed for monitoring babies, pets, elderly family members, or any space that needs attention.

Setting up is simple: 1) Visit safeos.sh on your device, 2) Allow camera and microphone permissions when prompted, 3) Choose a monitoring scenario (baby, pet, or elderly), 4) Adjust sensitivity settings to your preference, 5) Enable browser notifications for alerts. That's it—you're ready to start monitoring!

Yes! SafeOS Guardian is designed to work offline. Once the app is loaded in your browser, all detection and alerting happens locally on your device. No internet connection is required for core monitoring functionality. The app can even be installed as a Progressive Web App (PWA) for offline access.

SafeOS Guardian works on all modern browsers that support the MediaDevices API. This includes Chrome, Firefox, Safari, and Edge on desktop, as well as Chrome and Safari on mobile devices. For the best experience, we recommend using the latest version of Chrome or Firefox.

Yes, SafeOS Guardian is completely free and always will be. There are no subscriptions, no hidden fees, and no premium tiers. We believe everyone deserves access to monitoring tools for their loved ones, regardless of their financial situation. The project is supported by the Frame.dev team and optional donations.

No. Your video and audio data never leaves your device. All processing happens locally in your browser using TensorFlow.js. We don't have servers receiving or storing your camera feed. The only data stored is your settings and alert history, which remain in your browser's IndexedDB storage.

Absolutely not. Frame.dev cannot see, access, or control your camera in any way. SafeOS Guardian runs entirely in your browser—we don't operate servers that receive your video feed. Even if we wanted to, we couldn't see what your camera captures. Your privacy is architecturally protected.

Local-first means all computation happens on your device. When you start monitoring, TensorFlow.js (a machine learning library) runs directly in your browser. It analyzes video frames and audio samples without sending them anywhere. This approach ensures maximum privacy and eliminates the need for expensive cloud infrastructure.

Yes! Once you've loaded SafeOS Guardian in your browser (and optionally installed it as a PWA), you can use it completely offline. Motion detection, audio analysis, and alerts all work without internet. The only features requiring internet are optional cloud sync and receiving notifications on other devices.

We collect zero personal data. No analytics, no tracking, no telemetry. The only data stored is your local settings (sensitivity, preferences) and alert history—all kept in your browser's IndexedDB, never sent to us. If you enable optional cloud sync, only encrypted metadata syncs; video/audio stays local.

Motion detection uses frame differencing—we compare consecutive video frames to detect changes. When pixels change beyond a threshold, motion is detected. For sleep monitoring, we use "pixel detection" mode which is ultra-sensitive, detecting even tiny movements of 3-10 pixels. You can adjust sensitivity based on your needs.

Pixel detection is our ultra-sensitive monitoring mode designed for sleeping babies and pets. Instead of looking for large movements, it detects tiny pixel changes (configurable from 3-10 pixels). This catches subtle breathing movements or when a sleeping child stirs, giving you peace of mind during nap time.

Alert sensitivity is fully customizable. You can adjust motion sensitivity (how much movement triggers an alert), audio sensitivity (how loud sounds need to be), and pixel threshold (for sleep monitoring). We also offer presets like "Infant Sleep Monitor" (ultra-sensitive) and "Pet Watch" (balanced sensitivity).

Yes! SafeOS Guardian includes multiple alert sounds—gentle chimes for routine alerts, urgent alarms for high-priority events, and custom sound options. You can also adjust volume, enable vibration on mobile devices, and configure quiet hours to prevent alerts during specified times.

To reduce false alerts: 1) Adjust sensitivity sliders lower, 2) Use appropriate presets for your scenario, 3) Ensure stable lighting (sudden light changes can trigger motion), 4) Position the camera to avoid windows or moving objects in the background, 5) Use the alert delay setting to require sustained motion before alerting.

We use TensorFlow.js models optimized for in-browser inference. For motion detection, we use efficient frame differencing algorithms. For audio analysis (cry detection), we use frequency band analysis and pattern matching. All models are designed to run smoothly on consumer devices without requiring a GPU.

TensorFlow.js is a JavaScript library that brings machine learning to the browser. It uses WebGL for GPU acceleration when available, or falls back to CPU computation. This means AI inference runs directly on your device, enabling real-time analysis without sending data to external servers.

Yes. In offline/local mode you can download a local bundle from History → Offline Export (settings + alerts + saved frames). If you run the optional SafeOS monitoring server (API), you can also export server-side alerts/analysis/streams from the Export page. Everything remains under your control (local IndexedDB or your own server).

If SafeOS Guardian ever generates revenue through optional donations or premium features, we pledge to donate 10% to humanitarian organizations focused on child safety, missing persons recovery, and family welfare. This commitment is permanent and reflects our belief that technology should serve humanity.

We welcome feedback! Report bugs and request features on our GitHub repository at github.com/framersai/safeos. You can also reach the Frame.dev team directly at team@frame.dev. We actively review all submissions and prioritize based on community impact.