The course encompassed seven lessons with practical "homework" assignments meticulously reviewed by our AI specialist. The program featured:
Lesson 1: An overview of essential neural networks, including Midjorney and the online version of Stable Diffusion.
Lesson 2: Key principles of prompt engineering demonstrated through practical examples.
Lessons 3-6: Practical applications using Stable Diffusion, recognized as the most capable and cost-effective platform. This section included a guide to initiating Stable Diffusion, covering two methods of accessing the neural network: installation on a powerful computer or utilizing Google Colab. We delved into the interface differences based on the installation method, and crucially, explored methodologies for image generation and resizing. Topics covered ranged from basic syntax and photo processing to essential plugins application, and more.
A critical aspect addressed was the methodology for generating images and adjusting their size, covering basic plugins and techniques to instruct the neural network to generate faster or in higher quality.
The most crucial part for our client was model retraining based on the game assets. We instructed the client’s team on generating AI images of the specific game character, situating it in different scenarios and locations. These images could serve as sketches, be animated, and seamlessly incorporated into the actual mobile game.
Lesson 7: The final bonus lesson offered an overview of neural networks, industry news, and insights into copyright restrictions.
The clients acquired proficiency in installing and utilizing leading AI-generated tools. They gained a comprehensive understanding of the core principles of prompt engineering and studied crucial syntax for image generation and editing. Consequently, the client is now adept at training neural networks to generate characters for their game, saving substantial time and effort. This skill empowers them to test new concepts efficiently and bring innovative ideas to life.