map

semantic cluster view. each dot is an image; position reflects caption-embedding similarity. projected via UMAP.

scroll = zoom · drag = pan

234 of 408 points · stale · last computed 7/6/2026, 8:36:21 AM