Sentiment style transfer (SST) has recently attracted extensive interest from researchers, specifically, prompt learning and large language models (LLMs) have attained outstanding performance for transferring text styles such as formality and politeness. However, SST is still a challenging problem as the sentiment style is intertwined with the content of the text compared with other styles. In this paper, we investigateF the effects of dynamic prompts for SST by leveraging LLMs compared with static prompts which have been explored well in previous works. Our dynamic prompts utilize the results of an aspect-based sentiment analysis (ABSA) model. Through massive experiments on two datasets, we found that dynamic prompts can help five open-source LLMs improve the performance of sentiment transfer. Our source codes are available online.