Skip to main content
Use face and license-plate detection for UGC moderation, testimonials, vehicle photos, and privacy-safe previews. Region params hide explicit coordinates when you already know what to obscure.
Face detection debug overlay and pixelated face privacy result

Group portrait validation with face-debug=1 (left), then ?pixelateFaces=1 (right).

Vehicle license plate before and after automatic plate blur

License plate privacy - ?blurPlates=1&plate-debug=1 while validating. Remove debug before production.

Image with a blurred rectangular region

Region blur - ?blurRegion=120,320,220,80.

Face blur and pixelation

Face privacy runs after resize/crop/text overlay and before final encoding, so detected boxes line up with the delivered image. Face detection is best-effort - review sensitive workflows manually. For crop behaviour centred on faces (gravity=face), see Resize, crop & fit and the deeper Smart cropping guide. Use face-debug=1 with gravity=face while validating crop behaviour. Debug boxes are rendered into the image response, so do not ship debug URLs in production.

License plate blur and pixelation

Automatic plate privacy currently applies to image CDN outputs. Video URLs can be transformed by the CDN, but frame-by-frame automatic plate redaction for full videos is a separate pipeline.
Plate detection is best-effort and depends on angle, distance, plate design, and source resolution. For legal or high-risk privacy workflows, combine automatic detection with plateRegion or regions=plate:... when you know the exact area. plate-debug=1 draws detection boxes into the returned image. It is useful in the Transform Lab and QA environments, but should be removed from production URLs.

Region privacy

Use output-pixel coordinates when you already know what should be hidden:
Coordinate-based plate privacy remains the most deterministic option when your app already knows the plate bounds.
Last modified on July 17, 2026