Face Recovery

Utilize Face Recovery to improve images where facial attributes could be enhanced. Low resolution, blurry, and unclear details of human faces are best used for this feature.

• Face recovery detects and recovers human faces. It works best on blurry, low-resolution, or poor-quality faces, but may remove detail from high-quality faces
• Select faces - detects high and low confidence faces to enhance

Understanding Face Recovery

After importing an image, analysis scans the entire image, searching for the eyes, nose, and mouth of a human face.

It determines the number of faces recognized and categorizes them into two types: high confidence and low confidence.

High Confidence: AI analysis distinguishes clear details of the eyes, nose, and mouth, recognizing a face.

Low Confidence: faces that lack detail, are out of focus, or partial faces where the head is turned away.

Using Face Recovery

Face Recovery appears under the model settings in the right panel. Toggle the switch to activate the enhancement.

Click the button labeled Select faces to recover to see a selection of the desired faces.

In the preview, the image shows boxes around faces that are detected. Click on the desired boxes that need enhancement.

A dark gray box indicates a High Confidence result.
A light gray box indicates a Low Confidence result.
A blue box indicates the face is selected for enhancement.

After faces have been selected, click Apply.

After the AI model finishes enhancing, use Face recovery strength by adjusting the slider to change the effect on the face. Increasing or decreasing the value will adjust the intensity of the enhancement. Skin can appear smoother or clearer depending on the strength.

Face Recovery 3

Face Recovery 3 sharpens and recovers soft detail on faces of any size, producing sharp, portrait-quality output regardless of how large the face is in the frame.

The new model is a generative recover faces model that processes each face separately. Each additional face selected will increase processing time. For example, 4 similarly sized faces will take 4x the time compared to 1 face of the same size.

Recover Faces 3 runs on NeuroServer and processes locally on NVIDIA, AMD, and Apple Silicon (macOS 14 or later).

Face Recovery 2

Our models are designed to preserve the original image, but since Face Recovery is specifically designed to reconstruct damaged faces, the model will sometimes change expressions slightly or may smooth skin, hair, or small facial details.

This version offers two modes:

• Realistic mode closely mirrors the original Face Recovery model, but with improved overall quality. For scenarios with very low resolution or tiny faces, Creative mode might offer better results.
• Creative mode generates additional details that might be missing from the original image, producing high-quality results even from very low-resolution inputs. It’s incredible what can be achieved with just a few pixels.