Accurate seasonal precipitation forecasts, especially for extreme events, are crucial to preventing meteorological hazards and its potential impacts on national development, social stability, and security. However, the intensity of summer precipitation is often significantly underestimated in many current dynamical models. This study uses a deep learning method called Cycle-Consistent Generative Adversarial Networks (CycleGAN) to enhance the seasonal forecast skill of the Nanjing University of Information Science & Technology Climate Forecast System (NUIST-CFS1.0) in predicting June-July-August precipitation in southeastern China. The results suggest that the CycleGAN-based model significantly improves the accuracy in predicting the spatial-temporal distribution of summer precipitation than traditional quantile mapping (QM) method. Due to the use of unpaired day-to-day correction models, we can pay more attention to the frequency, intensity, and duration of extreme precipitation events in the climate dynamical model forecast. This study expands the potential applications of deep learning models to improving seasonal precipitation forecasts.