A company's logo is one of the most effective elements to identify a company's image at a glance. A well-crafted logo is given a very important role as a company's promotional and advertising tool that continuously enhances the company's image. However, producing a logo through a professional organization requires a lot of effort and expense. In fact, most companies invest a lot of money to create a brand logo that represents the company. However, it is a very big burden for new start-up companies to spend the cost of brand logo production. In order to reduce this burden, this study deals with the method of automatically generating a new logo image. When a logo is simply generated using an image generating network, it is difficult to expect a logo that reflects a specific shape or company's characteristics to be created. In order to solve this problem, this research proposes an AI-based logo generation method that can create a logo that reflects the sketch which can represent the characteristics of the company. For data generation, GAN, an artificial intelligence generation model, was used. Based on the existing GAN, a model of network structure that can be generated by reflecting the sketch image together was constructed. To generate a logo, a sketch is input from the user, and a logo is generated by using the input sketch as data input for creating a new logo. For stable learning of the proposed generation model, Wasserstein distance was used as the cost function, and 48,000 images of 32x32 size of the Large-Logo Dataset were uses as the dataset. In addition, by using a GUI based on Python, users can easily create sketch images. Through extensive experiments, it was confirmed that a new logo image with various changes was generated while having the characteristics of the sketch by reflecting the learning results on the existing logo data.