Generative Artificial Intelligence for the Visualization of Source Code as Comics

Data comics offer an innovative and accessible approach to visualizing abstract data, like source code. However, creating these comics is very challenging, as it requires an artist who can conceive and draw the comic while having a deep knowledge of the abstract data. This work explores the application of state-of-the art generative AI models, specifically GPT-4 and DALL \(\cdot \) E 3, to generate a complete comic using a zero-shot approach with three different prompts. Our experiment focuses on generating comics from Python source code. Through a qualitative evaluation, we observed that chain-of-thought prompting could enhance the quality of the generated comics, showcasing the potential advantages and limitations of current generative AI models in creating comics aimed at software comprehension.