This project focuses on generating unique game data by using a variety of learned techniques. Four techniques are used in the whole process, including text database production, image synthesis, image capture and text synthesis. The purpose of this experiment is to design some unique monster illustrations based on the Dungeons and Dragons series board games. There are more than two hundred monsters in the Dungeons & Dragons official monster manual. However, players often design monster settings that are more in line with their own game background during actual play. Therefore, the monster illustrations designed in this project are more in line with the requirements of players' custom games.

The first step is to collect the illustrations of the 254 monsters in the monster manual and use the image capture function to create a different captions text for each monster. 

The next step is to make a text database of monsters. Make an excel sheet and enter the monster’s picture number, caption, name, hit points, armor class, speed, six attributes and their descriptions. Make these texts into a text database after AI training. 

After that, an image database was created and 600 transition pictures were output, each of which contained different characteristics of multiple monsters.

The last step is to combine all the previous preparations, enter the transition picture into the software, and generate its caption. Input the caption into the text database to generate text synthesis which creates name, hit points, armor class, speed, six attributes and their descriptions. From this, the image and text description of each newly created monster are automatically generated. Then I used the Photoshop to make a few monster illustrations similar in style to the monster manual as the final work display.