Using coarse meshes with textures and/or normal maps to represent detailed meshes often results in poor visual quality along silhouettes. To tackle this problem, we introduce silhouette texture, a new data structure for capturing and reconstructing the silhouettes of detailed meshes. In addition to the recording of color and normal fields in traditional methods, we sample information that represents the original silhouettes and pack it into a three dimensional texture. In the rendering stage, our algorithm extracts relevant information from the texture to rebuild the silhouettes for any perspective view. Unlike previous work, our approach is based on GPU and could achieve high rendering performance. Moreover, both exterior and interior silhouettes are processed for better approximation quality. In addition to rendering acceleration, our algorithm also enables detailed silhouette visualization with minimum geometric complexity.