This supplemental material provides a comprehensive overview of the implementation details and evaluation of FFGLC, our proposed method for fashion social media popularity prediction. To enhance reproducibility, the implementation section includes detailed prompts used for generating captions, descriptions of the newly constructed WEAR-PD dataset, the integration process of FFGLC into existing popularity prediction frameworks, and specifications of hyperparameters and computational resources. In the evaluation section, both quantitative and qualitative analyses are presented. The quantitative evaluation demonstrates, using similarity metrics, that the generated captions effectively extract complementary information from fashion images. The qualitative evaluation provides 100 examples of captions generated for social media posts, showcasing their ability to accurately and comprehensively describe visual features. These results underline FFGLC's effectiveness in capturing intricate fashion-related information, ultimately improving the accuracy of popularity prediction.