Skin Enhancer

Skin Enhancer is a portrait-refinement tool that removes synthetic, waxy smoothness from AI-generated faces and replaces it with realistic micro-pores, fine grain, and natural facial imperfections. Built on Alibaba Tongyi-MAI's Z-Image architecture with a specialized skin-detailing LoRA, it operates as an image-to-image post-processing pass to turn airbrushed-looking outputs into photorealistic editorial stills without altering underlying facial likeness.

Set the image strength parameter between 0.75 and 0.95 to preserve your subject's exact identity while applying texture details. Keeping LoRA strength between 0.8 and 1.5 delivers balanced, organic micro-pores and fine epidermal grain. If you want the model to adjust facial expressions or lighting according to your text prompt, lower image strength toward 0.25.

High-resolution close-ups and cropped portraits where the subject's face occupies at least 30% to 50% of the frame produce the best results. The model requires clear facial geometry to semantically map pores and epidermal details. Avoid low-resolution, heavily compressed images or wide-angle group shots where faces appear as small background elements.

Avoid applying Skin Enhancer to non-photorealistic styles such as 2D anime, flat vector illustrations, and stylized digital paintings, as it will aggressively project human skin textures onto stylized surfaces. You should also avoid using it for geometric recomposition, severe character re-posing, or complex background modifications.

The underlying Z-Image model family was developed by Alibaba Tongyi-MAI, introducing the Scalable Single-Stream Diffusion Transformer (S3-DiT) architecture. Utilizing Decoupled-DMD distillation, Z-Image Turbo generates high-resolution portrait enhancements in 1 to 8 steps, enabling rapid, cost-effective single-pass retouching.