Pokémon, Picasso, and the Language of Comics: Can AI Map Art Styles to Purpose?

We put Pokémon and Picasso on McCloud’s Big Triangle to ask: can comic language describe vision — and can AI map art styles to purpose? Inspired by Understanding Comics.
TLDR
Scott McCloud’s Big Triangle places pictures between Reality, Language, and Picture. The Style Atlas maps 365 looks there. The published map uses the tagging results of the with Gemini 3.7 Flash. Typical miss on the is ~8% on ten comics (McCloud’s whole printed set ~13%; ~15%). Franchises skew Language; artists stay more Reality-led. Gemini 3.7 Flash is the balanced winner of that pass. Try Pokémon and Picasso.
Published Labels default
Why Gemini 3.7 Flash is the balanced winner
9.3% mean
Lowest average miss
Best even score of the six-model pass.
~1.7% band
Inside digitization noise
Sits with Qwen and Kimi; several models can still share “#1.”
14% worst
No blow-ups
Qwen’s tail exceeded 24%; Kimi hit 31% on Little Orphan Annie.
Not every column
Median and style wins
Qwen had the tightest typical miss (9.3% median). Gemini 3.5 Flash won the most styles, but sat outside the cluster.
Explore again
Jump to Pokémon
Published Labels (Gemini 3.7 Flash tagging results from the gated repair v3 prompt), Big Triangle diamonds on, Pokémon selected.
Open Pokémon →Flip to Trapezoid
Same Pokémon pin, calculated sister map for pictures that carry a concept.
Open Trapezoid →
* Sample images are for style analysis; we claim no rights to them (fair use). Contact me to request removal.