NEWS

---

rembg 0.1.1 (2026-07-22)

• Ask for confirmation before downloading models into the cache, per CRAN policy: an interactive prompt confirms the location on first use, and non-interactive use opts in with options(rembg.download = TRUE) or REMBG_DOWNLOAD=1.

rembg 0.1.0

• First release. An R port of the Python rembg background-removal package, running pre-trained ONNX segmentation models through onnxr (no Python), with jpeg/png and base R for image IO, resize and compositing.
• 16 models plus 3 bring-your-own-model presets: the u2net family, silueta, isnet (general/anime), birefnet (7 variants), bria-rmbg, u2net_cloth_seg, sam, and u2net_custom / dis_custom / ben_custom.
• rembg() outputs a transparent RGBA cutout, the mask only, a solid background colour, raw PNG bytes, or a written file.
• Optional closed-form alpha matting (a port of pymatting) for soft hair/fur edges, JPEG EXIF orientation handling, per-garment clothing segmentation, and SAM click-to-segment point prompts.