Vision Model
nomic-embed-vision (general · 380 MB)
Sapiens2-0.1B fp16 (human-centric · 229 MB)
Sapiens2-0.1B int8 (human-centric · 116 MB)
Sapiens2-0.1B fp32 (human-centric · 458 MB)
Chrome AI · Gemini Nano (built-in · 0 MB + 134 MB text)
Remote endpoint · OpenAI-compatible (uploads thumbnails)
Sapiens2 is pretrained on 1B human images — better for photos of people. Chrome AI uses the browser's built-in Gemini Nano to describe images (Chrome 138+ only, requires chrome://flags/#prompt-api-for-gemini-nano). Takes effect on next model load. Switching requires a page reload.
Image budget (EmbeddingGemma 2)
Auto (light on phones and Chromebooks)
Light — faster, less detail, gentler on the GPU
Full — more detail, needs a stronger GPU
How many image tokens each photo gets. Switch to Light if the GPU resets while embedding. Each choice keeps its own cache, and changing it reloads the page.
Chrome AI Describe Prompt
Reset
The prompt sent to Gemini Nano to describe each image. Changes apply to new embeddings only. Only shown when Chrome AI is selected.
Search: show top 20 % of photos
Slideshow: 3 s per photo
Slideshow cuts videos at the timer (off: waits for each video to end)
Enable Text Search (requires 134MB model)
Generate captions when viewing images (requires Chrome AI)
Loop Videos by Default
Saved Embeddings
Export file
Import file
Save into folder…
Moves embeddings and captions between devices, e.g. an overnight run on a desktop to your phone. Importing adds to what is here. A folder that holds a media-clusterer.mcemb is imported when you open it. Files are matched by SHA-256, so copies with a new modified date (phones, sync tools) still match.
Projection Method
UMAP
t-SNE
PCA
Isomap
LLE
MDS
Sammon
TriMap
Settings persist automatically in your browser cache.