This thesis presents a new image editing approach with convolutional networks to automatically alter the image content with a desired attribute and still keep the image photo-realistic. The proposed image editing approach effectively combines the strengths of two prominent images editing algorithms, conditional Generative Adversarial Networks and Deep Feature Interpolation, to be time-efficient, memory-efficient, and user-controllable. We also present an inverted deep convolutional network to facilitate the proposed image editing approach. Lastly, we describe the implementation of this image editing approach in an iOS application and demonstrate that this approach is feasible and practical in real-world applications.