Open the app and click "New Account Activation". Enter your registered email, create a password, and click "Locate License File" to select the license key provided by the Blueskeye AI team. Finally, click "Encrypt & Save Profile".
You will need a valid license key file to complete activation. Please contact the Blueskeye AI support team to request your license file.
This occurs when macOS Gatekeeper flags an un-notarized application package downloaded from the web. To bypass this, open Terminal on your Mac and run:
sudo xattr -rd com.apple.quarantine /Applications/bskai-desktop-data-analyser.app
Enter your system password when prompted, then re-open the app normally.
Supports standard video files, including .mp4, .mov, and .avi. For optimal performance and real-time SDK throughput, 1080p (1920x1080) at 15 fps is recommended for processing speed. SDK fps processing can be manually configured via the settings page, to either keep the native FPS (slower processing) or use the SDK optimised 15 fps processing (faster processing).
Bulk processing processes videos sequentially through the underlying SDK. Do not navigate away from the Analysis Workspace while processing is active. If the progress bar halts for longer than 60 seconds, verify your license key status or check that the input files are not corrupted.
If a video contains multiple subjects tracked by the SDK, click the Selected Subject dropdown menu at the top of the workspace report. Selecting a different subject ID will automatically update the Valence-Arousal graphs, Action Unit timelines, and summary reports for that ID.
Yes. You can export a full .csv file directly from the report view, or use the Splicing Tool to select a specific timestamp window and export a localised .csv subset.
Any standard UVC-compliant web camera or integrated webcam will work. Ensure your camera is selected under System Settings → Hardware Input before starting a Live Analysis session.
The Splicing Tool allows you to crop and export a specific timestamp segment of analysed data. Click the Splicing Tool icon in the top bar, set your start and end points on the timeline, edit the default file name if needed, and click Export CSV. This generates a new CSV for the selected time window without modifying your original video file.
Live Analysis displays real-time AI metrics (gaze, pose, emotions) from your camera feed on-screen without saving logs. Live Session Recording lets you record a live stream session, capture the real-time data, and generate exportable CSV or PDF reports once you click Stop.
You need a standard UVC-compliant webcam or integrated camera. If you are analysing audio-visual voice activity (AVAD), you will also need an active microphone. Select your preferred camera and microphone under System Settings -> Hardware Input before starting a session.
The system requires a brief warm-up window to connect to the hardware camera driver, initialise video frame buffers, and lock onto initial facial landmark meshes.
No. All computer vision processing, landmark tracking, and metric evaluations run 100% on your local machine. Video streams and audio feeds never leave local memory, and no raw media is sent to external cloud servers.
In the session report view or analysis view, click the Export as PDF button in the top right corner and select "Save as PDF" to generate a shareable visual summary.
Exports contain frame-by-frame numerical and categorical metrics generated by the SDK. No raw video files or facial imagery are ever saved.
The exported CSV includes:
Frame & Bounding Box Data: Frame indices, image dimensions, face detection coordinates, and face image quality indicators.
Spatial & Orientation Metrics: Head position (X, Y, Z coordinates), 3D head orientation (Yaw, Pitch, Roll), 68-point facial landmark coordinates (X0–X67, Y0–Y67), and Face ID tracking numbers.
Gaze & Attention Tracking: Gaze angles, 3D gaze vectors, eye occlusion/tracking statuses, and mapped attention intersection points.
Facial Action Units (AUs): Confidence scores and intensity values across 15 individual Action Units (e.g., AU01, AU06, AU12, AU45).
Affect & Emotion Metrics: Continuous Valence, Arousal, and Dominance scores (with confidence ratings), alongside categorical Emotion Zone probabilities (e.g., Happy, Neutral, Sad, Angry).
High-Level Operational Models: Real-time scores and status flags for Audio-Visual Voice Activity Detection (AVAD), Visual Voice Activity Detection (VVAD), Net Confusion, and Interaction Engagement.
Running on Tauri v2. The React 19 frontend dispatches asynchronous Inter-Process Communication (IPC) commands to the Rust core, which interfaces directly with the native C++ OpenCV engine over a C-FFI layer.
The system requires a brief warm-up period (typically 1–2 seconds) to allocate camera device handles, initialise OpenCV frame buffers, and establish the initial facial landmark tracking mesh.
Performing a Factory Reset permanently purges local user profiles, cached analysis sessions, and stored application settings. This action cannot be undone.