What Is Occlusion Handling in AI Lip-Sync?

Occlusion handling is the ability of a lip-sync model to keep producing accurate mouth movements when part of the speaker's face is covered — by a hand, a phone, a microphone, a coffee cup, or the product they are holding up to the camera. "Occlusion" simply means something is blocking the model's view of the mouth.

Why occlusion breaks lip-sync

Most lip-sync systems learn from the visible mouth area and paint new mouth shapes back into the same region. When an object crosses that region, the model has nothing reliable to work from, so it generates over the obstruction: the hand blurs, the microphone warps, or the mouth appears to bleed through whatever is in front of it. The result is a distorted frame in the middle of your ad.

This is not an edge case in performance creative. In UGC-style footage, speakers routinely:

• Hold the product next to their face while describing it
• Touch their chin, mouth or hair mid-sentence
• Drink, eat or gesture across the frame during a casual review
• Use a handheld or clip-on microphone that sits near the jaw

How GoSync 3.0 handles it

GoSync 3.0 detects when the mouth region is partly or fully obstructed and predicts the lip movement behind the obstacle using the surrounding facial context — jaw position, cheek movement and head angle — while keeping the occluding object itself intact. GeckoDub describes the goal as maintaining facial integrity with near-zero visual artifacts.

What you can do to help

Occlusion handling reduces the problem; it does not make framing irrelevant. Clips where the mouth is visible for most of the shot, filmed with steady lighting and limited motion blur, will always sync more cleanly than ones where the product covers the face for seconds at a time.