Blurr Help Center
Deep dive into the tools and technologies that power secure on-device video redaction.
Feature Guides
Master Blurr's automatic ML-powered detection modes.
Face Redaction
Blurr uses Apple's Vision framework to detect and blur faces across thousands of frames in seconds. Once enabled in the sidebar, the Neural Engine scans the entire duration of the video prior to rendering.
- Configurable blur intensity or solid shape fill.
- Operates at high speeds, capable of tracking profiles dynamically.
- Generates specific bounding box logs in the JSON audit report.
Screen & Text Redaction
Our Core ML monitor detection model automatically identifies digital displays, laptops, and tablets. It also utilizes OCR to obscure dense, readable text blocks in the environment before saving the final clip.
- Ideal for office recordings and healthcare compliance.
- Identifies text regardless of screen presence if legible.
- Mitigates manual frame-by-frame blurring in standard video editors.
Audio Filter Pipeline
Blurr integrates a multi-step audio pipeline using Whisper transcription locally. We process the audio track, locate sensitive names/entities via NLP, and generate FFmpeg timestamps to selectively mute.
- Does not destroy ambient audio, only spoken PII.
- Operates entirely offline without transcription APIs.
- Redacted segments are precisely zero-padded in the audio stream.
License Plate Detection
Identify vehicle license plates to maintain anonymity. Our system tracks moving vehicles using high-speed heuristics to ensure comprehensive redaction across intersection feeds and dashcams.
- Regional formatting logic matches alpha-numeric rules.
- Protects driver privacy without impacting background movement.
Frequently Asked Questions
Important details about Blurr's strict on-device limitations and update functionality.
No. Nothing leaves your device. Blurr is built using Apple's Core ML and Vision frameworks. All detection, transcription, and blurring processes run locally on your Mac's Neural Engine. Your raw footage is never uploaded anywhere, making Blurr completely secure for highly sensitive legal, medical, or law enforcement footage.
There are no hard limits to batch sizes. You can drag in a folder containing hundreds of clips and let Blurr queue them. Videos are processed entirely in the background, dependent only on your Mac's available RAM and storage.
No. Blurr reads your original files and exports redacted versions into a designated sibling directory (by default, a /Redacted subfolder alongside your originals). Your source footage remains completely untouched to maintain chain of custody.
Blurr uses the industry-standard Sparkle framework for seamless over-the-air updates. The app checks for new versions in the background and prompts you when an update is available.
Blurr requires a Mac with an Apple Silicon processer (M1, M2, M3, or M4 series) running macOS Sonoma (14.0) or later. The intensive machine learning models require the Neural Engine exclusive to Apple Silicon hardware, so Intel Macs are not supported.
Yes, Blurr automatically generates a structured JSON log alongside the redacted files, detailing the exact start/end constraints, ML confidence scores, and items redacted. You can also generate a human-readable PDF report for compliance reviews directly from the batch summary window.