Mapping Art Styles with Multimodal Models Using Scott McCloud’s Big Triangle

This project maps visual styles onto Scott McCloud’s Big Triangle (Reality – Language – Picture Plane) from Understanding Comics using multimodal models.
It plots 365 different looks on an interactive Style Atlas and compares several models on how closely their placements match expected positions. Gemini 3.7 Flash came out as the strongest overall on the validation set. One clear pattern: franchises tend to land closer to the Language corner, while individual artists stay more Reality-oriented.
You can explore the map directly here:
https://www.alexxg.com/posts/mapping-the-visual-triangle?reading=tldr