Image inpainting is widely used to restore the originality of the image by filling in the missing parts of the image, removing the mask or text in the picture, and clearing tiny scratches that cause damage to the old image and create a visually appealing image. In this work a generative adversarial network (GAN)-based image inpainting system is proposed with free-form masking. The proposed productive inpainting system uses gated convolutions to extract shallow features learned from many images and uses the weight normalization technique, spectral normalization with Wasserstein GAN + Gradient Penalty (WGAN-GP) loss functions in the discriminator to provide fast and stable training. The suggested model generates higher-quality images and more flexible extensions due to automatic image inpainting and user-guided extension results compared to earlier techniques.