Building the Word-Driven Variation System in Blender
Following last week’s first attempt to connect the conversations with the 3D structures, this week I focused on building a complete variation system in Blender, a system where changing a single word in the conversation automatically generates a new architectural variation.
I wrote a custom script that links each word to a specific parameter set inside Blender. I used 42 words, selected from the top words across all the conversations of my dataset. I then grouped these words into three categories, based on their narrative and emotional impact:
Soft: subtle changes in rotation, movement, scattering, duplication, and scale
Medium: more noticeable spatial transformations
Chaotic: the strongest, most disruptive changes with visible shifts in form and density
By switching from one word to another, the structure reformulates itself through these rules. The result is a fully automated, word-driven mechanism that translates linguistic content into spatial behavior.
