Skip to main content
Face Blur

Face Blur · screenshot coming soon

​
What this node does

Face Blur runs a face detector over your video and covers each detected face with a chosen obscuring effect. On ▶ Run it detects faces (re-detecting periodically to follow movement), then blurs, pixelates, or boxes them out, producing a new video. There is no drawing — detection is automatic. Video in/out is COMFYTV_VIDEO. Bridge to native ComfyUI nodes via ComfyTV/Bridge — see https://github.com/jtydhr88/ComfyTV/blob/main/docs/bridges.md.

​
When to use it

  • Anonymize bystanders for privacy before publishing
  • Redact identities in review or reference footage
  • Quick censor pass on faces without rotoscoping

​
Parameters

​
mode

How to obscure each face. Options: blur, pixelate, box. Default blur.

​
shape

The region shape covered per face. Options: rect, ellipse. Default ellipse.

​
strength

Intensity of the effect (blur amount / pixel block size). Range 4–64, step 1, default 24. Higher hides more detail.

​
recheck

Re-detect faces every N frames. Range 1–120, default 12. Lower values follow fast movement more closely but cost more; higher values are faster but can lag on quick motion.

​
search_scale

Detector scale factor. Range 1.05–2.0, default 1.2. Smaller values scan more thoroughly (finding more/smaller faces) but slower; larger values are faster but may miss faces.

​
neighbors

Detector min-neighbors threshold. Range 1–10, default 4. Higher rejects weak detections (fewer false positives, but may miss borderline faces); lower is more permissive.

​
min_size

Minimum face size in pixels to detect. Range 8–400, default 24. Raise it to ignore tiny background faces.

​
Outputs

​
Tips

  • If faces flicker in and out, lower recheck and/or search_scale, or lower min_size for smaller faces.
  • Getting false detections on non-faces? Raise neighbors.
  • Detection follows motion only between rechecks, so very fast movement may briefly reveal a face — use a low recheck.
  • Spot Remover — cover a fixed spot/object rather than faces
  • Paint Strokes — manually obscure regions the detector misses