How to cover a face with an emoji that follows it
Tell Vizard Agent whose face to cover and what to cover it with. It tracks the head through the clip and places the sticker on it frame by frame, which is the difference between an emoji that belongs to the person and one that sits in the middle of the screen while they walk out of it.
What do you type into Vizard Agent?
Say it simply and name the person. "Put a smiley on the face so you cannot see it" is a complete instruction. Which person is the only detail that's regularly missing: the adult, the one on the left, the one holding the camera. Mention the end of the clip too, because that's where tracking fails.
Prompt
What does Vizard Agent actually do?
Nearly all of the work goes into establishing exactly where the face is at every moment, which takes far more effort than placing the sticker. Vizard Agent also has to tell the real face from any emoji already in the edit, because a face detector may lock onto it and leave the real face visible. In the session behind this article it:
- Checked the already-edited file, made a contact sheet and looked at frames enlarged to see whether a sticker was already present and how the head moved.
- Fixed the face's coordinates at different moments, then ran built-in face detection on a local copy to track it frame by frame.
- Visualised the detected positions to tell the real face from a sticker already in the edit, and reviewed the tracking map across the whole video.
- Built a per-second overview to check the head's path by eye before placing anything.
- Concentrated on the final section, where the head shifts: it enlarged the top of the frame where the adult's head sits and isolated the hardest frames at seconds 8, 9, 10, 11, 12 and 13.
- Checked the last frame for the sticker's closing position.
When does this not work well?
Some shots defeat tracking. A face that turns fully away, is hidden behind an object, or moves very fast in low light will lose the sticker for a few frames, and a few frames is all it takes. Very small faces in a crowd are hard too, and existing stickers confuse automatic detection.
Common questions
Can it cover several faces? Yes. Vizard Agent tracks each one separately with its own sticker.
Can I use my own image? Yes. Upload it and Vizard Agent tracks that instead of an emoji.
What if the face turns away? Then there is nothing to cover, and Vizard Agent holds the last position or lets it go.