Recommendation systems play a crucial role in attracting and maintaining relationships with users by providing personalized product recommendations that meet their needs. Currently, many advanced methods in the field of recommendation systems can accurately predict ratings to suggest potential products. To improve the caliber of product recommendations in the fashion e-commerce industry, we propose a recommendation method in this study that combines textual and visual clues. Sentence Transformer models are used to encode customer reviews and textual descriptions, successfully capturing their semantic content, while the system uses image inputs processed by pre-trained Convolutional Neural Networks (CNNs) to produce rich visual embedding. After converting each modality into a feature vector, similarity scores are determined separately using both content-based and image-based features. To create a single relevance score for every item, these similarity values are then merged using a weighted fusion technique. Items are ranked according to how similar they are to the target product overall to produce the final suggestion list. The hybrid HCFIF model not only increases accuracy but also lowers prediction error (RMSE), according to the results, demonstrating the improved efficacy of integrating textual and visual content information in product recommendation systems.