Original
Blackout result
Output appears here
v6 single-model (SegFormer-B2 trained on DviantDses) — 3 classes: background, skin, other. v1 limitation: no face/hair distinction, so faces and skin both get blacked. v2 in progress to add face/hair/clothing as separate classes. This is a research preview; do not upload images you wouldn't want logged server-side.