Recover Faces & Select Faces
This filter detects faces in your image and increases face quality.
Autopilot Settings
Autopilot face detection and strength are controlled from the following menu:
Open Topaz Photo AI and go to the Edit > Preferences > Autopilot menu.
Open Topaz Photo AI and go to the Topaz Photo AI > Preferences > Autopilot menu.
The Face Detection menu will let you control:
- which faces are selected to recover automatically
- strength of face recovery
- parts of the face that are recovered

The face selection settings are:
- Auto: low quality faces are selected improved
- All: all faces are selected and improved
- Subject Only: faces in the subject mask are selected and improved
- None: face detection is turned off and no faces are improved
Set the default strength to your preference.
Parts to Enhance lets you deselect areas around the face to control processing. Deselect:
- detected hair
- detected necks
Select Faces
Hover over the "Faces" text in the autopilot to view the faces detected.
- faces selected for processing are outlined in yellow.
- deselected faces are outlined in grey.
- low confidence faces are outlined in dim grey.
By default, only high confidence, low quality are selected for recovery. Low confidence faces must be added manually for recovery.

To select/deselect faces for processing, click the Select button in Autopilot section or in the Recover Faces filter and open the Select Faces menu.

Decide which faces to recover in this menu. In general, pick important faces to improve and leave the rest to become part of the background.

Click on a face to select it. The buttons in the right menu let you quickly select or deselect faces.
- Select All faces enables every face for recovery, including low confidence faces
- Quick Select lets you easily select groups of detected faces.
Apply your changes by clicking the Apply button at the bottom.
Face Recovery
This filter improves faces in the image by detecting details in the face and constructing a high quality version that is blended with the current face. For best results, control the strength to blend the generated face with the original face and the rest of the image. This creates a realistic, high quality result.