How PBR Transforms Mobile 3D Makeup: From Theory to Real‑World Apps

Overview

With the continuous improvement of mobile CPU and GPU performance, physically based rendering (PBR) for high‑realism 3D makeup is gaining increasing attention. This article uses two concrete cases (glossy lipstick and water‑light eye shadow) to analyze how to apply the PBR lighting model to develop mobile 3D makeup effects.

PBR Basics

PBR is a collection of rendering techniques based on the physical principles of the real world, producing more realistic results than traditional Phong or Blinn‑Phong shading. Because it aligns closely with physical properties, artists can author materials directly from physical parameters without tedious tweaking, and the materials remain correct under any lighting condition.

In a PBR workflow three conditions must be satisfied: a micro‑facet surface model, energy conservation, and a physically based BRDF. Four core parameters are required: Albedo (color), Metallic, Roughness, and Reflectance. These parameters, together with the BRDF formula, determine the final appearance. Normal maps are optional.

2.1 Color

Color (or color map) defines the base perceptible color of a material. For non‑metallic surfaces it represents the diffuse color; for metallic surfaces it represents the specular color. Typical brightness ranges are 10‑240 for dielectrics and 170‑255 for metals, as shown in the figures.

2.2 Metallic

Metallic defines whether a material is a dielectric (non‑metal) or a conductor (metal). A metallic map is usually binary: 0.0 for non‑metal, 1.0 for metal, as illustrated.

2.3 Roughness

Roughness (or roughness map) defines surface smoothness: 0.0 is perfectly smooth, 1.0 is fully rough. Roughness influences the diffusion of reflections; smoother surfaces produce clearer reflections. The figure shows roughness effects on both dielectric and metallic materials.

2.4 Reflectance

Reflectance indicates the intensity of reflected light for non‑metallic materials (typically 2‑16%). Metals have much higher reflectance, around 90%. The figures illustrate typical ranges.

2.5 Direct Lighting PBR

Using a non‑metallic sphere and a metallic sphere under a single light source, we set the non‑metal reflectance to 0.04 and metal reflectance to 0.9. The four parameters (Albedo, Metallic, Normal, Roughness) are shown for the non‑metallic plastic sphere, which renders as a realistic plastic material.

Similarly, the metallic sphere uses a yellowish base color, metallic = 1.0, and renders as shown.

Direct lighting PBR captures the basic appearance but lacks realism for metallic objects because indirect lighting is omitted.

2.6 Indirect Lighting PBR

Indirect lighting uses HDR environment maps where each pixel acts as a light source, dramatically increasing computation. Pre‑filtering techniques reduce the cost while preserving real‑time performance. An additional Ambient Occlusion (AO) map adds subtle shadowing based on geometry.

3D Makeup Applications

After reviewing PBR fundamentals, we examine makeup effects commonly seen on major beauty platforms, focusing on lipstick and eye shadow. Traditional 2D approaches cannot capture lighting nuances, so a 3D workflow with PBR lighting is required for realism.

3.1 Glossy Lipstick

The glossy lipstick case demonstrates a high‑gloss effect that follows camera or face movement, achievable with direct‑lighting PBR. Required textures include Albedo, Metallic (≈0), Roughness (low), Reflectance (0.04), and optional normal map. The final render is shown in the accompanying video.

3.2 Water‑Light Eye Shadow

The water‑light eye shadow uses indirect‑lighting PBR to create a flowing light effect that reacts to camera or face motion. Textures are similar to the lipstick case, with a black base color, white AO (1.0), and an HDR environment map to drive reflections.

The indirect lighting algorithm pre‑computes filtered environment maps and BRDF lookup tables (see LearnOpenGL for details), enabling real‑time performance. The final eye‑shadow result is composited onto the face image using an inverse‑screen blend, as shown in the second video.

Conclusion

Ideal makeup effects must be physically accurate yet sometimes transcend realism. Achieving such results requires expertise in computer vision, computer graphics, performance optimization, technical art, machine learning, and computational aesthetics. Kuaishou Y‑Tech will continue to innovate in makeup technology to deliver more realistic, attractive, and stable experiences.