- Are the photos or video uploaded?
- No. Reference photos, the source video, and face embeddings stay in this browser tab. Refreshing the page clears the session.
- What makes a reference photo valid?
- The photo must contain exactly one clear face. Multiple faces, a missing face, or a blurry face are rejected so each photo represents one person.
- Which videos can be analyzed?
- MP4, MOV, and MKV files that the browser can decode. If playback fails, convert the file to H.264 MP4 on your computer first.
- What are the model terms?
- Buffalo L uses InsightFace's official pretrained models, which are limited to non-commercial research. The page links to the model source and license.
- Which browser should I use?
- Desktop Chrome or Edge is the practical choice. The model is large, and sampling a long video once per second depends on the device.
