Wikimedia Commons · live cross-modal search
Describe what
you want to see.
Layer turns a text query into the same CLIP space as thousands of freely licensed images. Both towers run inside the public gateway on CPU—no GPU worker, embedding pool, or client model.
Try a scene
Serving leg
local CLIP
Compute
gateway CPU
Model
ViT-B/32 · 512d
Gateway embed
waiting
Round trip
waiting
Images indexed
loading…
GPU workers
0
Nearest images
Results
No close images yet.
Try a broader visual description.