The detection of insulator defects in power transmission systems is crucial for ensuring the reliability and safety of electrical grids. However, the development of effective defect detection models is often hindered by the scarcity of annotated defect data. To address this challenge, we propose to use the controllable image editing framework i.e. DesignEdit that enables the generation of synthetic insulator defect images, thereby enriching the training dataset and enhancing the model’s ability to detect various defect types. Specifically, the DesignEdit combines foreground defects and background normal insulator images by manipulating the latent layers and then harmonizes the synthetic images. Extensive experiments demonstrate that our method not only generates high-quality synthetic defect images but also significantly boosts the performance of defect detection models on real-world datasets.