imgoji renders a string of emoji into a composite image, encodes arbitrary images into an imgoji string or into a grid of emoji.
Processed in your browser. The image is not uploaded.
Short link / public gallery
Short links and gallery submissions send your imgoji to the imgoji server. Short links expire in 90 days; gallery entries are public.
examples
Encode the image as a blended quadtree string.
Pins detected objects (faces, clothing, etc.) as the encode's base layers via CLIP + OWL-ViT β downloads hundreds of MB from huggingface.co and runs them on your device. Slow first run; needs a capable machine.
Advanced
Threshold 0 lets the byte budget pick the prune point; raise it to prune harder. Compare is match resolution: higher is sharper but slower and rebuilds the palette.
Flatten the image into a grid of emoji β color, plus semantic picks where objects are detected.
Semantic mode downloads CLIP + OWL-ViT models (hundreds of MB) from huggingface.co and runs them on your device β a large download; the first run is slow and needs a capable machine.
Advanced
Detail sets the resolution and autotunes blend and box threshold. Specificity favors smaller foreground objects over larger background ones in overlaps. Tune manually here; Reset restores the autotuned values.