Blurzy documentation
Blurzy is an under-development desktop application for detecting and tracking subjects in video, reviewing annotations, and exporting video with selected subjects blurred.
What you can do
- Open video and review it with playback and frame-level controls.
- Run detection with supported model providers and track subjects.
- Review and edit bounding-box annotations.
- Import and export annotation data and export blurred video.
- Monitor long-running detection, tracking, and export tasks without blocking the desktop interface.
Model availability depends on the installed providers, model weights, and hardware/runtime support. Blurzy does not yet claim a stable release or a model-independent detection experience. Review annotations and exported video before relying on the result; automated detection can miss subjects.
Explore the project
- Getting started: install and launch the desktop app.
- Development: run tests, build packages, and preview this site.
- Architecture: understand the Clean Architecture layers, four data layers, and processing workflows.
- Engineering conventions: follow the project's documentation style.
- Planning notes: read proposals, not implemented features or a committed roadmap.
The current workflow includes detection with YOLO, TorchVision, or MTCNN providers; Hungarian, ByteTrack, DeepSORT, and BoT-SORT tracking; editable annotation layers; project save/load; JSON/CSV layer interchange; and separate preview and rendered-video blur controls. Provider availability and hardware requirements vary. Review annotations and exported video before relying on automated redaction.