Blur Face Video - TrackBlur

What happens the second you open TrackBlur

The app launches directly into a video picker. No login screen, no welcome tour, no permission requests upfront. The interface shows a large "Select Video" button at the center. Below it, a small settings gear icon. That is it. The app does not ask for storage permission until you actually pick a video. When you tap the button, the system file picker appears. You select a video, and the app immediately begins analyzing it. A progress bar shows detection status. For a 30-second clip, this takes about four to six seconds on a mid-range device. The app uses on-device AI for face and object detection. No data leaves the phone. The detection results appear as colored bounding boxes around faces and objects. Each detected item gets a track on the timeline at the bottom. The timeline shows thumbnails at regular intervals. You can scrub through it to see how the tracking holds up. The app does not auto-apply any blur. It shows you what it found and waits for your input. This is a deliberate design choice. The developer assumes you want control over what gets blurred and what does not. The interface remains responsive during detection. You can still interact with the settings while the analysis runs in the background. The app does not show any ads at this stage. Ads only appear when you attempt to export the final video.

How auto-detection and manual keyframes cover each other's gaps

Auto-detection works well for clear, front-facing shots with good lighting. It finds faces and common objects like vehicles and screens. But the app's own description admits that detection can miss targets in complex footage. This is where manual keyframes become essential. The workflow goes like this: you let the AI run first. It creates tracks for everything it finds. You review the timeline. If the AI missed a face or an object, you draw a manual region over it. The app then lets you add keyframes to animate that region's position over time. For example, if a person walks across the frame and the AI lost track halfway, you can place keyframes at the start and end of the movement. The app interpolates the position between them. This means you do not have to manually adjust every single frame. The combination of AI detection plus manual keyframe correction covers most real-world scenarios. The app also offers adjustable detection quality and face padding. Higher detection quality takes longer but finds more targets. Face padding adds extra space around detected faces, which helps when the subject moves slightly outside the bounding box. The app does not explain these settings in detail. You have to experiment to see what works for your specific video. The timeline shows all tracks stacked vertically. You can toggle each track on or off. This is useful when you want to blur only certain people and leave others untouched. The app does not offer batch processing. Each video must be edited individually.

The real-world performance: ads, tracking drift, and export consistency

Exporting a video triggers a short ad for free users. The ad plays before the export process begins. The export itself runs entirely on the device. It uses the device's GPU for rendering. The time depends on video length, resolution, and the number of effects applied. A 1080p clip with three blur tracks and one text overlay takes about two minutes to export on a Snapdragon 888 device. The app shows a progress bar during export. It does not allow background processing. The screen must stay on. Tracking drift remains the most common issue. The AI can lose lock when the subject moves quickly or when the background has similar colors. The app provides a stabilization setting that smooths out tracking jitter. But this setting can also introduce a slight lag. The tracked blur region may trail behind the actual subject by a few pixels. The app's preview window shows the real-time effect. You can scrub through the timeline to check every frame before exporting. The "What's New" section mentions improved tracking progress feedback. This suggests the developer is actively working on tracking accuracy. The app also includes a HEX color input for precise color selection in text and overlays. This is a niche feature for users who need exact brand colors. The app does not include any social sharing buttons. You export the video to your device's gallery and share it manually.

The permission that determines whether offline editing stays offline

TrackBlur requests storage permission when you first select a video. This is required for reading the source file and writing the exported video. The app does not request any other permissions. No camera, no microphone, no location. The app's description emphasizes that 100% of processing happens on the device. No data is sent to the cloud. This is verified by the fact that the app works in airplane mode. The detection and rendering engine is entirely local. The app does not use any external libraries that require network access. The APK size is around 30 MB. The app does not download additional model files after installation. All AI models are bundled in the APK. This means the app works offline from the first launch. The only network activity is for ad loading when exporting. If you block network access via the system firewall, the export still works but the ad may not load. The app handles this by showing a fallback message. The storage permission is the only gatekeeper for the entire editing workflow. Without it, the app cannot read or write videos. The app does not explain this clearly on first launch. It simply asks for permission and continues. If you deny storage, the app shows an error and asks again. There is no way to use the app without granting storage access.