What makes a photo lip sync well
A talking photo moves the mouth, jaw and eyes of the face in the picture, so the model needs to see them clearly. Percify’s own guidance is a clear, front facing portrait. The checker turns that into things it can measure:
| Check | Good | Why it matters |
|---|---|---|
| Face size | Face fills 15% or more of the photo’s height | A small face leaves few pixels for the lips and eyes |
| Detail on the face | 160 pixels or more across the face | Low resolution faces come out soft when they move |
| Facing the camera | Nose centred between the eyes | A turned head has one side of the mouth hidden |
| Head level | Eye line within 10° of level | A tilted head reads as unnatural when it starts talking |
| Whole face in frame | Space around the head and chin | A cut off chin or forehead limits movement |
| Lighting | Face neither dark nor blown out | Shadows and glare hide the mouth |
| Sharpness | Face in focus | Blur makes the lip movement smeary |
- These thresholds are our guidance from testing, not hard limits: a photo marked could be better still works, just less cleanly.
How it works
The face is found with Google’s MediaPipe face detector running in your browser. The detector is downloaded once, from a pinned version, and checked against a fingerprint before it runs. Light is measured on the face itself, and focus from how much fine detail the face has.
Your photo is never uploaded. The closer crop is cut from your original file on your device, at full resolution.
Checked on 17 September 2026 against Percify’s photo guidance and tests on sample portraits.