Photoshop Skin Smoothing Neural Filter Explained

This article explores how Adobe Photoshop’s Skin Smoothing Neural Filter works to automatically retouch portraits. Powered by Adobe Sensei machine learning, the filter identifies facial skin regions and performs non-destructive transformations to soften blemishes, diminish wrinkles, and balance skin tones while preserving essential pore details and facial features.

Automated Facial and Skin Segmentation

The primary transformation begins with semantic segmentation. Rather than requiring users to manually select faces or create alpha masks, the neural network analyzes the image to automatically detect human faces and map out skin areas. It creates an internal, edge-aware mask that isolates the skin surface while entirely excluding key structural features such as the eyes, eyebrows, eyelashes, lips, teeth, nostrils, hair, and clothing.

Surface Blurring and Blemish Reduction

Once the skin is identified, the filter applies an intelligent blurring transformation to the mid-frequency details of the skin. This targets and softens:

• Acne, blemishes, and minor surface imperfections
• Fine lines and subtle wrinkles
• Uneven skin blotchiness and minor color variations

Unlike traditional Gaussian blur, which flattens all pixel information indiscriminately, the neural filter functions similarly to an automated frequency separation workflow. It targets the tonal inconsistencies on the skin's surface without degrading structural edges.

Texture and Pore Retention

A critical transformation performed by the filter is edge-preserving detail retention. The algorithm separates fine, high-frequency details—such as microscopic skin pores and fine hair—from the underlying skin tone. While the surface coloration is smoothed out, the original skin texture is mathematically blended back over the smoothed base layer. This prevents the "plastic" or over-airbrushed appearance common in standard blur filters.

Adjustable Parameter Controls

The filter applies transformations through two primary user-controlled adjustments:

• Blur: Adjusts the radius and strength of the surface smoothing applied to the skin tone. Higher values produce a more even, matte finish across the face.
• Smoothness: Regulates the preservation of high-frequency details. This slider balances the intensity of the smoothing against the natural texture of the skin, dictating how much original pore structure remains visible.

Non-Destructive Output Transformations

The final transformation can be output to several formats depending on the editor's workflow. It can modify the current layer directly, output as a duplicated layer with a layer mask, generate an entirely new layer, or apply non-destructively as a Smart Filter. When output as a Smart Filter, the underlying pixel data remains intact, allowing editors to readjust the smoothing parameters or paint on the filter mask at any time.