Although recent generative models have achieved impressive performance in image editing, they still exhibit difficulties in producing precise and consistent results. On the other hand, image matting methods can be used to extract objects from images for precise and consistent image editing, but they face challenges when the objects are occluded. In this paper, we propose a high-performance image matting model based on vision foundation models and generative models to predict alpha mattes and colors of objects and backgrounds. It can use both visual and text prompts to specify which object to extract. A training strategy is designed to enable the model to handle occlusions and improve the model’s robustness to varying prompt qualities. We demonstrate how our model excels in image editing with its capabilities. On the Adobe Image Matting and Distinctions-646 datasets, our model significantly outperforms state-of-the-art models in both image matting and image editing.The code is available at .