Skip to main content
Toolzum

Bulk Face Anonymizer

Detect and blur or pixelate faces across whole photo batches with on-device BlazeFace (one shared model download for the batch). Faces need confidence above 0.5 and size above 8px; faceless files are skipped with an error. Outputs PNG as <name>-anonymized.png — local processing, 2 free batches a day.

Runs on TensorFlow.jsCanvasJSZip

100% Local Browser Native

How to Use the Bulk Face Anonymizer

1. Upload the batch

Front-facing, well-lit photos detect best. The model loads once, then runs per image — slower on constrained devices.

2. Pick blur or pixelate and strength

Gaussian Blur (strength = blur px, default 20) or Pixelate (block = face/strength×2, min 8px). Strength runs 5–50. No solid-cover mode exists.

3. Download ZIP

Only anonymized copies export as PNGs; originals stay untouched. Faceless files are skipped, not exported. One batch save counts as one Pro download.

Similar Tools You Might Need

See all

Frequently Asked Questions

How accurate is batch face detection?

Faces pass only above 0.5 confidence and 8px size — front-facing, well-lit faces pass reliably; profile angles, sunglasses, and tiny background faces get skipped. For 100 event photos, re-run misses individually rather than trusting blind automation for publication.

Blur or pixelate — which is safest?

Heavy pixelation is hardest to reverse; light blur can sometimes be enhanced back, so use high strength (up to 50px blur) where identity protection is legally required. No solid-bar mode exists — pixelate at low strength values for the blockiest cover.

Are my photos uploaded for detection?

No — detection runs on-device in your browser with TensorFlow.js; images never leave your device. Originals stay untouched in memory; only anonymized copies export.

Can it anonymize video frames?

Still images only in this tool — export video frames first, anonymize the batch, and rebuild. For full video redaction, use a dedicated video anonymization workflow instead.

Last updated October 2026