Content area
This thesis investigates how procedural, highly flexible workflows often come with a steep technical learning curve for artists who lack experience with complex node-based systems exploring how Artificial Intelligence (AI), in the guise of Natural Language Processing (NLP) and a Model Context Protocol (MCP), can bridge this gap by enabling users to define and modify procedural networks using natural language inputs. Some platforms, such as Unreal Engine, Nuke, Houdini, and Maya, have pipelines through which AI technologies may be integrated. Houdini was chosen as the preferred environment for initial testing, allowing thorough experimentation, as it is node-based and supports Python integrations. This experimentation led to the development of several prototype tools, including terrain, color ramp, shape, and curve generators, which were interlinked through Python scripting with large language models (LLMs) such as GPT-3.5-Turbo. The results of the qualitative trial showed that small asset creation tasks were relatively easy to automate and yielded consistent results. However, more complex node manipulations still required multiple steps and iterative adjustments, which were drastically streamlined with the implementation of MCP. The results confirm that AI-assisted systems, when well balanced with procedural design principles, can make node-based workflows more reproducible, accessible, and efficient without sacrificing user control over creativity. Future directions include broader integration across platforms like Blender, Maya, and Unreal Engine, where hybrid AI-procedure pipelines could further improve digital content creation.