In the information age, short videos have become an integral part of daily life, leading to an exponential increase in the volume of video data on short video platforms. Without effective organization and classification of these video resources, management becomes a significant challenge. Therefore, the development of efficient video categorization techniques is crucial. This study proposes an automatic video tag extraction method based on Large Language Model text content parsing (VTE-LLM). The method integrates OCR (Optical Character Recognition) and CRNN (Convolutional Recurrent Neural Network) models to extract subtitle information from videos, and utilizes Large Language Models for text parsing to automatically generate video classifications and tags. This approach effectively addresses the cost and efficiency issues associated with traditional methods. Experimental results demonstrate that the proposed method achieves high accuracy across multiple video types.