In the era of big data, real-time multimodal content retrieval has become a key technology for processing diverse information. Existing retrieval algorithms often face the problem of low accuracy in personalized matching and cannot effectively meet the diverse needs of users. This paper first constructs a multimodal feature extraction model to comprehensively extract features from different types of data such as text, images, and audio. Next, a personalized matching algorithm based on deep learning is proposed, which can adjust the matching strategy according to user preferences. Finally, an optimization mechanism based on user feedback is established to continuously improve the performance of the algorithm. The experimental results show that the proposed algorithm significantly improves the personalized matching efficiency in real-time multimodal content retrieval and finally stabilizes at approximately 70% to 80%. This shows that the cross-modal matching of images and texts is relatively stable under multiple tests.