Enter one or more are.na channels separated by commas to map up to 300 image blocks. Images are embedded with SigLIP, positioned with t-SNE, and coloured by k-means clusters from the original embedding space.