Six segmentation models, picked automatically from what your browser can do.
Drop a photo.
The subject is cut out,
the text goes behind it.
Segmentation, repair and video encoding all run on your own hardware. The models are downloaded once and cached, and after that the page works with the network off.
Four things, and the fourth is the one most tools get wrong: what you export is the full-resolution frame, not the preview scaled up.
Open the editorSix models, picked from what your GPU actually reports.
Eleven faces, plus any font file you own.
The preview and the 4K export run identical draw calls.
Answers "what is the subject". Right for a portrait or a product shot.
Answers "what is near". Everything closer than the cutoff occludes the text.
Answers neither question. You mark exactly what goes in front — and it needs no download.
Ready to put
text behind
your subject?