Effectively and accurately portraying the short-term output uncertainty of renewable energy is an important foundation for the stable operation of power systems. In recent years, researchers conduct a lot of studies on generative models, among which flow-based generative models show great potential. However, flow models are rarely used to portray the uncertainty of renewable energy. Therefore, a short-term output scenario generation method of renewable energy based on conditional flow model is proposed. Firstly, a series of invertible functions are trained using real output data to map the probability distribution of the data to a standard Gaussian distribution. After the training is completed, the new output curve can be generated by inputting random noise obeying Gaussian distribution into the trained flow model. The method can be trained directly by maximum likelihood estimation and employs a series of specially designed invertible transformations, which can efficiently compute the training target. In addition, a generated scene evaluation index system is constructed to assess the quality of generated scenes, achieving interpretability of the results generated by the artificial intelligence black box model. The examples verify the effectiveness and sophistication of the proposed method.