Face Comparison
Neural face embeddings, 478-point landmarks, periocular (eye-region) matching for masked faces, age / gender / expression estimation, head pose, image quality scoring and one-to-many matching across group photos.
Runs 100% in your browser
Files are never uploaded
Preparing the on-device AI engine…
-
Subject A · Reference / known person
Subject B · Questioned / unknown person
Getting the most reliable result
Image quality
Use the sharpest, highest-resolution source. Faces smaller than about 80 pixels wide give weak results. Use the Video Frame Extractor to grab the best CCTV frame, and the Forensic Enhancer for exposure.
Use the sharpest, highest-resolution source. Faces smaller than about 80 pixels wide give weak results. Use the Video Frame Extractor to grab the best CCTV frame, and the Forensic Enhancer for exposure.
Masks & occlusion
When a face wears a mask the tool automatically compares the upper face only: eyes, brows, forehead and nose bridge. It applies the same virtual mask to both faces for a fair comparison. Reliability is lower, so treat results as a lead.
When a face wears a mask the tool automatically compares the upper face only: eyes, brows, forehead and nose bridge. It applies the same virtual mask to both faces for a fair comparison. Reliability is lower, so treat results as a lead.
Pose & age gap
Frontal faces work best. Differences over about 30° of head turn, heavy make-up, glasses, strong lighting or many years between photos all reduce similarity, even for the same person.
Frontal faces work best. Differences over about 30° of head turn, heavy make-up, glasses, strong lighting or many years between photos all reduce similarity, even for the same person.
Results are automated estimates for investigative reference only and are not evidence. Always verify independently. Read the disclaimer.